jmlr jmlr2010 jmlr2010-35 knowledge-graph by maker-knowledge-mining

35 jmlr-2010-Error-Correcting Output Codes Library

Source: pdf

Author: Sergio Escalera, Oriol Pujol, Petia Radeva

Abstract: In this paper, we present an open source Error-Correcting Output Codes (ECOC) library. The ECOC framework is a powerful tool to deal with multi-class categorization problems. This library contains both state-of-the-art coding (one-versus-one, one-versus-all, dense random, sparse random, DECOC, forest-ECOC, and ECOC-ONE) and decoding designs (hamming, euclidean, inverse hamming, laplacian, β-density, attenuated, loss-based, probabilistic kernel-based, and lossweighted) with the parameters deﬁned by the authors, as well as the option to include your own coding, decoding, and base classiﬁer. Keywords: error-correcting output codes, multi-class classiﬁcation, coding, decoding, open source, matlab, octave 1. Error-Correcting Output Codes The Error-Correcting Output Codes (ECOC) framework (Dietterich and Bakiri, 1995) is a simple but powerful framework to deal with the multi-class categorization problem based on the embedding of binary classiﬁers. Given a set of Nc classes, the basis of the ECOC framework consists of designing a codeword for each of the classes. These codewords encode the membership information of each class for a given binary problem. Arranging the codewords as rows of a matrix, we obtain a ”coding matrix” Mc , where Mc ∈ {−1, 0, 1}Nc ×n , being n the length of the codewords codifying each class. From the point of view of learning, Mc is constructed by considering n binary problems, each one corresponding to a column of the matrix Mc . Each of these binary problems (or dichotomizers) splits the set of classes in two partitions (coded by +1 or -1 in Mc according to their class set membership, or 0 if the class is not considered by the current binary problem). Then, at the decoding step, applying the n trained binary classiﬁers, a code is obtained for each data point in the test set. This code is compared to the base codewords of each class deﬁned in the matrix Mc , and the data point is assigned to the class with the ”closest” codeword. Several decoding strategies have been proposed in literature. The reader is referred to Escalera et al. (2008) for a more detailed review. An example of an ECOC design is described in Fig. 1. The ECOC designs are independent of the base classiﬁer applied. They involve error-correcting properties (Dietterich and Bakiri, 1995) and have shown to be able to reduce the bias and variance produced by the learning algorithm (Kong and Dietterich, 1995). Because of these reasons, ECOCs have been widely used to deal with multi-class categorization problems. c 2010 Sergio Escalera, Oriol Pujol and Petia Radeva. E SCALERA , P UJOL AND R ADEVA ECOC coding design for a 4-class problem. White, black, and grey positions corresponds to the symbols +1, -1, and 0, respectively. Once the four binary problems are learnt, at the decoding step a new test sample X is tested by the n classiﬁers. Then, the new codeword x = {x1 , .., xn } is compared with the class codewords {C1 , ..C4 }, classifying the new sample by the class ci which codeword minimizes the decoding measure. Figure 1: ECOC design example. 2. Library Algorithms The ECOCs library is a Matlab/Octave code under the open source GPL license (gpl) with the implementation of the state-of-the-art coding and decoding ECOC designs. A main function deﬁnes the multi-class data, coding, decoding, and base classiﬁer. A list of parameters are also included in order to tune the different strategies. In addition to the implemented coding and decoding designs, which are described in the following section, the user can include his own coding, decoding, and base classiﬁer as deﬁned in the user guide. 2.1 Implemented Coding Designs The ECOC designs of the ECOC library cover the state-of-the-art of coding strategies, mainly divided in two main groups: problem-independent approaches, which do not take into account the distribution of the data to deﬁne the coding matrix, and the problem-dependent designs, where information of the particular domain is used to guide the coding design. 2.1.1 P ROBLEM - INDEPENDENT ECOC D ESIGNS • One-versus-all (Rifkin and Klautau, 2004): Nc dichotomizers are learnt for Nc classes, where each one splits one class from the rest of classes. • One-versus-one (Nilsson, 1965): n = Nc (Nc − 1)/2 dichotomizers are learnt for Nc classes, splitting each possible pair of classes. • Dense Random (Allwein et al., 2002): n = 10 · log Nc dichotomizers are suggested to be learnt for Nc classes, where P(−1) = 1 − P(+1), being P(−1) and P(+1) the probability of the symbols -1 and +1 to appear, respectively. Then, from a set of deﬁned random matrices, the one which maximizes a decoding measure among all possible rows of Mc is selected. • Sparse Random (Escalera et al., 2009): n = 15 · log Nc dichotomizers are suggested to be learnt for Nc classes, where P(0) = 1 − P(−1) − P(+1), deﬁning a set of random matrices Mc and selecting the one which maximizes a decoding measure among all possible rows of Mc . 662 E RROR -C ORRECTING O UPUT C ODES L IBRARY 2.1.2 P ROBLEM - DEPENDENT ECOC D ESIGNS • DECOC (Pujol et al., 2006): problem-dependent design that uses n = Nc − 1 dichotomizers. The partitions of the problem are learnt by means of a binary tree structure using exhaustive search or a SFFS criterion. Finally, each internal node of the tree is embedded as a column in Mc . • Forest-ECOC (Escalera et al., 2007): problem-dependent design that uses n = (Nc − 1) · T dichotomizers, where T stands for the number of binary tree structures to be embedded. This approach extends the variability of the classiﬁers of the DECOC design by including extra dichotomizers. • ECOC-ONE (Pujol et al., 2008): problem-dependent design that uses n = 2 · Nc suggested dichotomizers. A validation sub-set is used to extend any initial matrix Mc and to increase its generalization by including new dichotomizers that focus on difﬁcult to split classes. 2.2 Implemented Decoding Designs The software comes with a complete set of ECOC decoding strategies. The notation used refers to that used in (Escalera et al., 2008): j • Hamming decoding: HD(x, yi ) = ∑n (1 − sign(x j · yi ))/2, being x a test codeword and yi a j=1 codeword from Mc corresponding to class Ci . • Inverse Hamming decoding: IHD(x, yi ) = max(∆−1 DT ), where ∆(i1 , i2 ) = HD(yi1 , yi2 ), and D is the vector of Hamming decoding values of the test codeword x for each of the base codewords yi . • Euclidean decoding: ED(x, yi ) = j ∑n (x j − yi )2 . j=1 • Attenuated Euclidean decoding: AED(x, yi ) = j j ∑n | yi || x j | (x j − yi )2 . j=1 • Loss-based decoding: LB(ρ, yi ) = ∑n L(yi · f j (ρ)), where ρ is a test sample, L is a lossj=1 function, and f is a real-valued function f : R n → R . j • Probabilistic-based decoding: PD(yi , x)=−log ∏ j∈[1,..,n]:Mc (i, j)=0 P(x j = Mc (i, j)| f j ) + K , where K is a constant factor that collects the probability mass dispersed on the invalid codes, and the probability P(x j = Mc (i, j)| f j ) j 1 , where vectors υ and ω are obtained by is estimated by means of P(x j = yi | f j ) = j j j j 1+eyi (υ f +ω ) solving an optimization problem (Passerini et al., 2004). αi +1 • Laplacian decoding: LAP(x, yi ) = αi +βi +K , where αi is the number of matched positions between x and yi , βi is the number of miss-matches without considering the positions coded by 0, and K is an integer value that codiﬁes the number of classes considered by the classiﬁer. νi 1 • Pessimistic β-Density Distribution decoding: accuracy si : νi −si ψi (ν, αi , βi )dν = 3 , where 1 ψi (ν, αi , βi ) = K ναi (1 − ν)βi , ψi is the β-Density Distribution between a codeword x and a class codeword yi for class ci , and ν ∈ R : [0, 1]. R j • Loss-Weighted decoding: LW (ρ, i) = ∑n MW (i, j)L(yi · f (ρ, j)), where MW (i, j) = ∑nH(i, j) j) , j=1 H(i, j=1 j 1, if x j = yi , 1 H(i, j) = mi ∑mi ϕ(h j (ρi ), i, j), ϕ(x j , i, j) = , mi is the number of training k k=1 0, otherwise. samples from class Ci , and ρi is the kth sample from class Ci . k 663 E SCALERA , P UJOL AND R ADEVA 3. Implementation Details The ECOCs Library comes with detailed documentation. A user guide describes the usage of the software. All the strategies and parameters used in the functions and ﬁles are described in detail. The user guide also presents examples of variable setting and execution, including a demo ﬁle. About the computational complexity, the training and testing time depends on the data size, coding and decoding algorithms, as well as the base classiﬁer used in the ECOC design. Acknowledgments This work has been supported in part by projects TIN2009-14404-C02 and CONSOLIDER-INGENIO CSD 2007-00018. References URL http://www.gnu.org/licences/. E. Allwein, R. Schapire, and Y. Singer. Reducing multiclass to binary: A unifying approach for margin classiﬁers. Journal of Machine Learning Research, 1:113–141, 2002. T. Dietterich and G. Bakiri. Solving multiclass learning problems via error-correcting output codes. Journal of Artiﬁcial Intelligence Research, 2:263–282, 1995. S. Escalera, Oriol Pujol, and Petia Radeva. Boosted landmarks of contextual descriptors and ForestECOC: A novel framework to detect and classify objects in clutter scenes. Pattern Recognition Letters, 28(13):1759–1768, 2007. S. Escalera, O. Pujol, and P. Radeva. On the decoding process in ternary error-correcting output codes. IEEE Transactions in Pattern Analysis and Machine Intelligence, 99, 2008. S. Escalera, O. Pujol, and P. Radeva. Separability of ternary codes for sparse designs of errorcorrecting output codes. Pattern Recognition Letters, 30:285–297, 2009. E. B. Kong and T. G. Dietterich. Error-correcting output coding corrects bias and variance. International Conference of Machine Learning, pages 313–321, 1995. N. J. Nilsson. Learning Machines. McGraw-Hill, 1965. A. Passerini, M. Pontil, and P. Frasconi. New results on error correcting output codes of kernel machines. IEEE Transactions on Neural Networks, 15(1):45–54, 2004. O. Pujol, P. Radeva, , and J. Vitri` . Discriminant ECOC: A heuristic method for application depena dent design of error correcting output codes. IEEE Transactions in Pattern Analysis and Machine Intelligence, 28:1001–1007, 2006. O. Pujol, S. Escalera, and P. Radeva. An incremental node embedding technique for error-correcting output codes. Pattern Recognition, 4:713–725, 2008. R. Rifkin and A. Klautau. In defense of one-vs-all classiﬁcation. The Journal of Machine Learning Research, 5:101–141, 2004. 664

Reference: text

Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 EDU Computer Vision Center Ediﬁci O, Campus UAB, 08193, Bellaterra, Barcelona, Spain Editor: Cheng Soon Ong Abstract In this paper, we present an open source Error-Correcting Output Codes (ECOC) library. [sent-7, score-0.021]

2 The ECOC framework is a powerful tool to deal with multi-class categorization problems. [sent-8, score-0.075]

3 Keywords: error-correcting output codes, multi-class classiﬁcation, coding, decoding, open source, matlab, octave 1. [sent-10, score-0.063]

4 Error-Correcting Output Codes The Error-Correcting Output Codes (ECOC) framework (Dietterich and Bakiri, 1995) is a simple but powerful framework to deal with the multi-class categorization problem based on the embedding of binary classiﬁers. [sent-11, score-0.127]

5 Given a set of Nc classes, the basis of the ECOC framework consists of designing a codeword for each of the classes. [sent-12, score-0.238]

6 These codewords encode the membership information of each class for a given binary problem. [sent-13, score-0.238]

7 Arranging the codewords as rows of a matrix, we obtain a ”coding matrix” Mc , where Mc ∈ {−1, 0, 1}Nc ×n , being n the length of the codewords codifying each class. [sent-14, score-0.38]

8 From the point of view of learning, Mc is constructed by considering n binary problems, each one corresponding to a column of the matrix Mc . [sent-15, score-0.027]

9 Each of these binary problems (or dichotomizers) splits the set of classes in two partitions (coded by +1 or -1 in Mc according to their class set membership, or 0 if the class is not considered by the current binary problem). [sent-16, score-0.117]

10 Then, at the decoding step, applying the n trained binary classiﬁers, a code is obtained for each data point in the test set. [sent-17, score-0.375]

11 This code is compared to the base codewords of each class deﬁned in the matrix Mc , and the data point is assigned to the class with the ”closest” codeword. [sent-18, score-0.23]

12 An example of an ECOC design is described in Fig. [sent-22, score-0.029]

13 The ECOC designs are independent of the base classiﬁer applied. [sent-24, score-0.18]

14 Because of these reasons, ECOCs have been widely used to deal with multi-class categorization problems. [sent-26, score-0.056]

15 E SCALERA , P UJOL AND R ADEVA ECOC coding design for a 4-class problem. [sent-28, score-0.243]

16 White, black, and grey positions corresponds to the symbols +1, -1, and 0, respectively. [sent-29, score-0.079]

17 Once the four binary problems are learnt, at the decoding step a new test sample X is tested by the n classiﬁers. [sent-30, score-0.375]

18 , xn } is compared with the class codewords {C1 , . [sent-33, score-0.179]

19 C4 }, classifying the new sample by the class ci which codeword minimizes the decoding measure. [sent-35, score-0.635]

20 Library Algorithms The ECOCs library is a Matlab/Octave code under the open source GPL license (gpl) with the implementation of the state-of-the-art coding and decoding ECOC designs. [sent-38, score-0.667]

21 A main function deﬁnes the multi-class data, coding, decoding, and base classiﬁer. [sent-39, score-0.051]

22 In addition to the implemented coding and decoding designs, which are described in the following section, the user can include his own coding, decoding, and base classiﬁer as deﬁned in the user guide. [sent-41, score-0.677]

23 1 P ROBLEM - INDEPENDENT ECOC D ESIGNS • One-versus-all (Rifkin and Klautau, 2004): Nc dichotomizers are learnt for Nc classes, where each one splits one class from the rest of classes. [sent-46, score-0.318]

24 • One-versus-one (Nilsson, 1965): n = Nc (Nc − 1)/2 dichotomizers are learnt for Nc classes, splitting each possible pair of classes. [sent-47, score-0.298]

25 , 2002): n = 10 · log Nc dichotomizers are suggested to be learnt for Nc classes, where P(−1) = 1 − P(+1), being P(−1) and P(+1) the probability of the symbols -1 and +1 to appear, respectively. [sent-49, score-0.342]

26 Then, from a set of deﬁned random matrices, the one which maximizes a decoding measure among all possible rows of Mc is selected. [sent-50, score-0.37]

27 , 2009): n = 15 · log Nc dichotomizers are suggested to be learnt for Nc classes, where P(0) = 1 − P(−1) − P(+1), deﬁning a set of random matrices Mc and selecting the one which maximizes a decoding measure among all possible rows of Mc . [sent-52, score-0.688]

28 , 2006): problem-dependent design that uses n = Nc − 1 dichotomizers. [sent-56, score-0.029]

29 The partitions of the problem are learnt by means of a binary tree structure using exhaustive search or a SFFS criterion. [sent-57, score-0.158]

30 Finally, each internal node of the tree is embedded as a column in Mc . [sent-58, score-0.021]

31 , 2007): problem-dependent design that uses n = (Nc − 1) · T dichotomizers, where T stands for the number of binary tree structures to be embedded. [sent-60, score-0.077]

32 This approach extends the variability of the classiﬁers of the DECOC design by including extra dichotomizers. [sent-61, score-0.029]

33 , 2008): problem-dependent design that uses n = 2 · Nc suggested dichotomizers. [sent-63, score-0.049]

34 A validation sub-set is used to extend any initial matrix Mc and to increase its generalization by including new dichotomizers that focus on difﬁcult to split classes. [sent-64, score-0.209]

35 2 Implemented Decoding Designs The software comes with a complete set of ECOC decoding strategies. [sent-66, score-0.348]

36 , 2008): j • Hamming decoding: HD(x, yi ) = ∑n (1 − sign(x j · yi ))/2, being x a test codeword and yi a j=1 codeword from Mc corresponding to class Ci . [sent-68, score-0.728]

37 • Inverse Hamming decoding: IHD(x, yi ) = max(∆−1 DT ), where ∆(i1 , i2 ) = HD(yi1 , yi2 ), and D is the vector of Hamming decoding values of the test codeword x for each of the base codewords yi . [sent-69, score-0.984]

38 • Euclidean decoding: ED(x, yi ) = j ∑n (x j − yi )2 . [sent-70, score-0.168]

39 j=1 • Attenuated Euclidean decoding: AED(x, yi ) = j j ∑n | yi || x j | (x j − yi )2 . [sent-71, score-0.252]

40 j=1 • Loss-based decoding: LB(ρ, yi ) = ∑n L(yi · f j (ρ)), where ρ is a test sample, L is a lossj=1 function, and f is a real-valued function f : R n → R . [sent-72, score-0.084]

41 αi +1 • Laplacian decoding: LAP(x, yi ) = αi +βi +K , where αi is the number of matched positions between x and yi , βi is the number of miss-matches without considering the positions coded by 0, and K is an integer value that codiﬁes the number of classes considered by the classiﬁer. [sent-77, score-0.308]

42 νi 1 • Pessimistic β-Density Distribution decoding: accuracy si : νi −si ψi (ν, αi , βi )dν = 3 , where 1 ψi (ν, αi , βi ) = K ναi (1 − ν)βi , ψi is the β-Density Distribution between a codeword x and a class codeword yi for class ci , and ν ∈ R : [0, 1]. [sent-78, score-0.609]

43 R j • Loss-Weighted decoding: LW (ρ, i) = ∑n MW (i, j)L(yi · f (ρ, j)), where MW (i, j) = ∑nH(i, j) j) , j=1 H(i, j=1 j 1, if x j = yi , 1 H(i, j) = mi ∑mi ϕ(h j (ρi ), i, j), ϕ(x j , i, j) = , mi is the number of training k k=1 0, otherwise. [sent-79, score-0.14]

44 A user guide describes the usage of the software. [sent-83, score-0.075]

45 All the strategies and parameters used in the functions and ﬁles are described in detail. [sent-84, score-0.023]

46 The user guide also presents examples of variable setting and execution, including a demo ﬁle. [sent-85, score-0.075]

47 About the computational complexity, the training and testing time depends on the data size, coding and decoding algorithms, as well as the base classiﬁer used in the ECOC design. [sent-86, score-0.613]

48 Reducing multiclass to binary: A unifying approach for margin classiﬁers. [sent-95, score-0.023]

49 Boosted landmarks of contextual descriptors and ForestECOC: A novel framework to detect and classify objects in clutter scenes. [sent-104, score-0.089]

50 On the decoding process in ternary error-correcting output codes. [sent-110, score-0.437]

51 Separability of ternary codes for sparse designs of errorcorrecting output codes. [sent-116, score-0.341]

52 New results on error correcting output codes of kernel machines. [sent-134, score-0.201]

53 Discriminant ECOC: A heuristic method for application depena dent design of error correcting output codes. [sent-140, score-0.125]

54 An incremental node embedding technique for error-correcting output codes. [sent-146, score-0.068]

similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('ecoc', 0.477), ('decoding', 0.348), ('escalera', 0.298), ('pujol', 0.268), ('mc', 0.241), ('codeword', 0.238), ('coding', 0.214), ('dichotomizers', 0.209), ('nc', 0.2), ('codewords', 0.179), ('designs', 0.129), ('codes', 0.123), ('oriol', 0.119), ('petia', 0.119), ('decoc', 0.089), ('ecocs', 0.089), ('learnt', 0.089), ('hamming', 0.088), ('yi', 0.084), ('sergio', 0.076), ('library', 0.063), ('dietterich', 0.062), ('adeva', 0.06), ('bakiri', 0.06), ('maia', 0.06), ('passerini', 0.06), ('radeva', 0.06), ('scalera', 0.06), ('ujol', 0.06), ('attenuated', 0.051), ('esigns', 0.051), ('base', 0.051), ('ub', 0.05), ('ci', 0.049), ('coded', 0.046), ('roblem', 0.046), ('mw', 0.046), ('allwein', 0.046), ('gpl', 0.046), ('ternary', 0.046), ('output', 0.043), ('guide', 0.043), ('categorization', 0.037), ('rifkin', 0.037), ('kong', 0.037), ('positions', 0.036), ('hd', 0.035), ('correcting', 0.035), ('membership', 0.032), ('user', 0.032), ('design', 0.029), ('mi', 0.028), ('binary', 0.027), ('classi', 0.027), ('laplacian', 0.026), ('rror', 0.025), ('lap', 0.025), ('clutter', 0.025), ('arranging', 0.025), ('ihd', 0.025), ('invalid', 0.025), ('ivanova', 0.025), ('embedding', 0.025), ('symbols', 0.024), ('campus', 0.023), ('dispersed', 0.023), ('eyi', 0.023), ('klautau', 0.023), ('corrects', 0.023), ('ibrary', 0.023), ('landmarks', 0.023), ('multiclass', 0.023), ('strategies', 0.023), ('rows', 0.022), ('classes', 0.022), ('tree', 0.021), ('barcelona', 0.021), ('descriptors', 0.021), ('boosted', 0.021), ('license', 0.021), ('partitions', 0.021), ('euclidean', 0.021), ('source', 0.021), ('dense', 0.02), ('splits', 0.02), ('suggested', 0.02), ('odes', 0.02), ('contextual', 0.02), ('nh', 0.02), ('lw', 0.02), ('nilsson', 0.02), ('octave', 0.02), ('powerful', 0.019), ('deal', 0.019), ('lb', 0.019), ('defense', 0.019), ('separability', 0.019), ('pessimistic', 0.019), ('grey', 0.019), ('dent', 0.018)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 1.0000002 35 jmlr-2010-Error-Correcting Output Codes Library

Author: Sergio Escalera, Oriol Pujol, Petia Radeva

2 0.047085088 29 jmlr-2010-Covariance in Unsupervised Learning of Probabilistic Grammars

Author: Shay B. Cohen, Noah A. Smith

Abstract: Probabilistic grammars offer great ﬂexibility in modeling discrete sequential data like natural language text. Their symbolic component is amenable to inspection by humans, while their probabilistic component helps resolve ambiguity. They also permit the use of well-understood, generalpurpose learning algorithms. There has been an increased interest in using probabilistic grammars in the Bayesian setting. To date, most of the literature has focused on using a Dirichlet prior. The Dirichlet prior has several limitations, including that it cannot directly model covariance between the probabilistic grammar’s parameters. Yet, various grammar parameters are expected to be correlated because the elements in language they represent share linguistic properties. In this paper, we suggest an alternative to the Dirichlet prior, a family of logistic normal distributions. We derive an inference algorithm for this family of distributions and experiment with the task of dependency grammar induction, demonstrating performance improvements with our priors on a set of six treebanks in different natural languages. Our covariance framework permits soft parameter tying within grammars and across grammars for text in different languages, and we show empirical gains in a novel learning setting using bilingual, non-parallel data. Keywords: dependency grammar induction, variational inference, logistic normal distribution, Bayesian inference

3 0.036241997 87 jmlr-2010-Online Learning for Matrix Factorization and Sparse Coding

Author: Julien Mairal, Francis Bach, Jean Ponce, Guillermo Sapiro

Abstract: Sparse coding—that is, modelling data vectors as sparse linear combinations of basis elements—is widely used in machine learning, neuroscience, signal processing, and statistics. This paper focuses on the large-scale matrix factorization problem that consists of learning the basis set in order to adapt it to speciﬁc data. Variations of this problem include dictionary learning in signal processing, non-negative matrix factorization and sparse principal component analysis. In this paper, we propose to address these tasks with a new online optimization algorithm, based on stochastic approximations, which scales up gracefully to large data sets with millions of training samples, and extends naturally to various matrix factorization formulations, making it suitable for a wide range of learning problems. A proof of convergence is presented, along with experiments with natural images and genomic data demonstrating that it leads to state-of-the-art performance in terms of speed and optimization for both small and large data sets. Keywords: basis pursuit, dictionary learning, matrix factorization, online learning, sparse coding, sparse principal component analysis, stochastic approximations, stochastic optimization, nonnegative matrix factorization

4 0.033639688 39 jmlr-2010-FastInf: An Efficient Approximate Inference Library

Author: Ariel Jaimovich, Ofer Meshi, Ian McGraw, Gal Elidan

Abstract: The FastInf C++ library is designed to perform memory and time efﬁcient approximate inference in large-scale discrete undirected graphical models. The focus of the library is propagation based approximate inference methods, ranging from the basic loopy belief propagation algorithm to propagation based on convex free energies. Various message scheduling schemes that improve on the standard synchronous or asynchronous approaches are included. Also implemented are a clique tree based exact inference, Gibbs sampling, and the mean ﬁeld algorithm. In addition to inference, FastInf provides parameter estimation capabilities as well as representation and learning of shared parameters. It offers a rich interface that facilitates extension of the basic classes to other inference and learning methods. Keywords: graphical models, Markov random ﬁeld, loopy belief propagation, approximate inference

5 0.028153433 61 jmlr-2010-Learning From Crowds

Author: Vikas C. Raykar, Shipeng Yu, Linda H. Zhao, Gerardo Hermosillo Valadez, Charles Florin, Luca Bogoni, Linda Moy

Abstract: For many supervised learning tasks it may be infeasible (or very expensive) to obtain objective and reliable labels. Instead, we can collect subjective (possibly noisy) labels from multiple experts or annotators. In practice, there is a substantial amount of disagreement among the annotators, and hence it is of great practical interest to address conventional supervised learning problems in this scenario. In this paper we describe a probabilistic approach for supervised learning when we have multiple annotators providing (possibly noisy) labels but no absolute gold standard. The proposed algorithm evaluates the different experts and also gives an estimate of the actual hidden labels. Experimental results indicate that the proposed method is superior to the commonly used majority voting baseline. Keywords: multiple annotators, multiple experts, multiple teachers, crowdsourcing 1. Supervised Learning From Multiple Annotators/Experts A typical supervised learning scenario consists of a training set D = {(xi , yi )}N containing N i=1 instances, where xi ∈ X is an instance (typically a d-dimensional feature vector) and yi ∈ Y is the corresponding known label. The task is to learn a function f : X → Y which generalizes well on unseen data. Speciﬁcally for binary classiﬁcation the supervision is from the set Y = {0, 1}, for multi-class classiﬁcation Y = {1, . . . , K}, for ordinal regression Y = {1, . . . , K} (with an ordering 1 < . . . < K), and Y = R for regression. c 2010 Vikas C. Raykar, Shipeng Yu, Linda H. Zhao, Gerardo H. Valadez, Charles Florin, Luca Bogoni and Linda Moy. R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY However, for many real life tasks, it may not be possible, or may be too expensive (or tedious) to acquire the actual label yi for training—which we refer to as the gold standard or the objective ground truth. Instead, we may have multiple (possibly noisy) labels y1 , . . . , yR provided by R i i different experts or annotators. In practice, there is a substantial amount of disagreement among the experts, and hence it is of great practical interest to address conventional supervised learning algorithms in this scenario. Our motivation for this work comes from the area of computer-aided diagnosis1 (CAD), where the task is to build a classiﬁer to predict whether a suspicious region on a medical image (like a X-ray, CT scan, or MRI) is malignant (cancerous) or benign. In order to train such a classiﬁer, a set of images is collected from hospitals. The actual gold standard (whether it is cancer or not) can only be obtained from a biopsy of the tissue. Since it is an expensive, invasive, and potentially dangerous process, often CAD systems are built from labels assigned by multiple radiologists who identify the locations of malignant lesions. Each radiologist visually examines the medical images and provides a subjective (possibly noisy) version of the gold standard.2 The radiologist also annotates various descriptors of the potentially malignant lesion, like the size (a regression problem), shape (a multiclass classiﬁcation problem), and also degree of malignancy (an ordinal regression problem). The radiologists come from a diverse pool including luminaries, experts, residents, and novices. Very often there is lot of disagreement among the annotations. For a lot of tasks the labels provided by the annotators are inherently subjective and there will be substantial variation among different annotators. The domain of text classiﬁcation offers such a scenario. In this context the task is to predict the category for a token of text. The labels for training are assigned by human annotators who read the text and attribute their subjective category. With the advent of crowdsourcing (Howe, 2008) services like Amazon’s Mechanical Turk,3 Games with a Purpose,4 and reCAPTCHA5 it is quite inexpensive to acquire labels from a large number of annotators (possibly thousands) in a short time (Sheng et al., 2008; Snow et al., 2008; Sorokin and Forsyth, 2008). Websites such as Galaxy Zoo6 allow the public to label astronomical images over the internet. In situations like these, the performance of different annotators can vary widely (some may even be malicious), and without the actual gold standard, it may not be possible to evaluate the annotators. In this work, we provide principled probabilistic solutions to the following questions: 1. How to adapt conventional supervised learning algorithms when we have multiple annotators providing subjective labels but no objective gold standard? 2. How to evaluate systems when we do not have absolute gold-standard? 3. A closely related problem—particularly relevant when there are a large number of annotators— is to estimate how reliable/trustworthy is each annotator. 1. See Fung et al. (2009) for an overview of the data mining issues in this area. 2. Sometimes even a biopsy cannot conﬁrm whether it is cancer or not and hence all we can hope to get is subjective ground truth. 3. Mechanical Turk found at https://www.mturk.com. 4. Games with a Purpose found at http://www.gwap.com. 5. reCAPTCHA found at http://recaptcha.net/. 6. Galaxy Zoo found at http://galaxyzoo.org. 1298 L EARNING F ROM C ROWDS 1.1 The Problem With Majority Voting When we have multiple labels a commonly used strategy is to use the labels on which the majority of them agree (or average for regression problem) as an estimate of the actual gold standard. For binary classiﬁcation problems this amounts to using the majority label,7 that is, j yi = ˆ 1 if (1/R) ∑R yi > 0.5 j=1 , j 0 if (1/R) ∑R yi < 0.5 j=1 as an estimate of the hidden true label and use this estimate to learn and evaluate classiﬁers/annotators. Another strategy is that of considering every pair (instance, label) provided by each expert as a separate example. Note that this amounts to using a soft probabilistic estimate of the actual ground truth to learn the classiﬁer, that is, R Pr[yi = 1|y1 , . . . , yR ] = (1/R) ∑ yi . i i j j=1 Majority voting assumes all experts are equally good. However, for example, if there is only one true expert and the majority are novices, and if novices give the same incorrect label to a speciﬁc instance, then the majority voting method would favor the novices since they are in a majority. One could address this problem by introducing a weight capturing how good each expert is. But how would one measure the performance of an expert when there is no gold standard available? 1.2 Proposed Approach and Organization To address the apparent chicken-and-egg problem, we present a maximum-likelihood estimator that jointly learns the classiﬁer/regressor, the annotator accuracy, and the actual true label. For ease of exposition we start with binary classiﬁcation problem in § 2. The performance of each annotator is measured in terms of the sensitivity and speciﬁcity with respect to the unknown gold standard (§ 2.1). The proposed algorithm automatically discovers the best experts and assigns a higher weight to them. In order to incorporate prior knowledge about each annotator, we impose a beta prior on the sensitivity and speciﬁcity and derive the maximum-a-posteriori estimate (§ 2.6). The ﬁnal estimation is performed by an Expectation Maximization (EM) algorithm that iteratively establishes a particular gold standard, measures the performance of the experts given that gold standard, and reﬁnes the gold standard based on the performance measures. While the proposed approach is described using logistic regression as the base classiﬁer (§ 2.2), it is quite general, and can be used with any black-box classiﬁer (§ 2.7), and can also handle missing labels (that is, each expert is not required to label all the instances). Furthermore, we extend the proposed algorithm to handle categorical (§ 3), ordinal (§ 4), and regression problems (§ 5). In § 6 section we extensively validate our approach using both simulated data and real data from different domains. 1.3 Related Work and Novel Contributions We ﬁrst summarize the novel contributions of this work in context of other related work in this emerging new area. There has been a long line of work in the biostatistics and epidemiology literature on latent variable models where the task is to get an estimate of the observer error rates based 7. When there is no clear majority among the multiple experts (that is, yi = 0.5) in CAD domain the ﬁnal decision is ˆ often made by an adjudicator or a super-expert. When there is no adjudicator a fair coin toss is used. 1299 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY on the results from multiple diagnostic tests without a gold standard (see Dawid and Skene, 1979, Hui and Walter, 1980, Hui and Zhou, 1998, Albert and Dodd, 2004 and references therein). In the machine learning community Smyth et al. (1995) ﬁrst addressed the same problem in the context of labeling volcanoes in satellite images of Venus. We differ from this previous body of work in the following aspects: 1. Unlike Dawid and Skene (1979) and Smyth et al. (1995) which just focused on estimating the ground truth from multiple noisy labels, we speciﬁcally address the issue of learning a classiﬁer. Estimating the ground truth and the annotator/classifer performance is a byproduct of our proposed algorithm. 2. In order to learn a classiﬁer Smyth (1995) proposed to ﬁrst estimate the ground truth (without using the features) and then use the probabilistic ground truth to learn a classiﬁer. In contrast, our proposed algorithm learns the classiﬁer and the ground truth jointly. Our experiments (§ 6.1.1) show that the classiﬁer learnt and ground truth obtained by the proposed algorithm is superior to that obtained by other procedures which ﬁrst estimates the ground truth and then learns the classiﬁer. 3. Our solution is more general and can be easily extended to categorical(§ 3), ordinal(§ 4), and continuous data(§ 5). It can also be used in conjunction with any supervised learning algorithm. A preliminary version of this paper (Raykar et al., 2009) mainly discussed the binary classiﬁcation problem. 4. Our proposed algorithm is also Bayesian—we impose a prior on the experts. The priors can potential capture the skill of different annotators. In this paper we refrain from doing a full Bayesian inference and use the mode of the posterior as a point estimate. A recent complete Bayesian generalization of these kind of models has been developed by Carpenter (2008). 5. The EM approach used in this paper is similar to that proposed by Jin and Ghahramani (2003). However their motivation is somewhat different. In their setting, each training example is annotated with a set of possible labels, only one of which is correct. There has been recent interest in the natural language processing (Sheng et al., 2008; Snow et al., 2008) and computer vision (Sorokin and Forsyth, 2008) communities where they use Amazon’s Mechanical Turk to collect annotations from many people. They show that it can be potentially as good as that provided by an expert. Sheng et al. (2008) analyzed when it is worthwhile to acquire new labels for some of the training examples. There is also some theoretical work (see Lugosi, 1992 and Dekel and Shamir, 2009a) dealing with multiple experts. Recently Dekel and Shamir (2009b) presented an algorithm which does not resort to repeated labeling, that is, each example does not have to be labeled by multiple teachers. Donmez et al. (2009) address the issue of active learning in this scenario—How to jointly learn the accuracy of labeling sources and obtain the most informative labels for the active learning task? There has also been some work in the medical imaging community (Warﬁeld et al., 2004; Cholleti et al., 2008). 2. Binary Classiﬁcation We ﬁrst describe our proposed noise model for the annotators. The performance of each annotator is measured in terms of the sensitivity and speciﬁcity with respect to the unknown gold standard. 1300 L EARNING F ROM C ROWDS 2.1 A Two-coin Model for Annotators Let y j ∈ {0, 1} be the label assigned to the instance x by the jth annotator/expert. Let y be the actual (unobserved) label for this instance. Each annotator provides a version of this hidden true label based on two biased coins. If the true label is one, she ﬂips a coin with bias α j (sensitivity). If the true label is zero, she ﬂips a coin with bias β j (speciﬁcity). In each case, if she gets heads she keeps the original label, otherwise she ﬂips the label. If the true label is one, the sensitivity (true positive rate) for the jth annotator is deﬁned as the probability that she labels it as one. α j := Pr[y j = 1|y = 1]. (1) On the other hand, if the true label is zero, the speciﬁcity (1−false positive rate) is deﬁned as the probability that she labels it as zero. β j := Pr[y j = 0|y = 0]. (2) The assumption introduced is that α j and β j do not depend on the instance x. For example, in the CAD domain, this means that the radiologist’s performance is consistent across different sub-groups of data.8 2.2 Classiﬁcation Model While the proposed method can be used for any classiﬁer, for ease of exposition, we consider the family of linear discriminating functions: F = { fw}, where for any x, w ∈ Rd , fw(x) = w⊤ x. The ﬁnal classiﬁer can be written in the following form: y = 1 if w⊤ x ≥ γ and 0 otherwise. The ˆ threshold γ determines the operating point of the classiﬁer. The Receiver Operating Characteristic (ROC) curve is obtained as γ is swept from −∞ to ∞. The probability for the positive class is modeled as a logistic sigmoid acting on fw, that is, Pr[y = 1|x, w] = σ(w⊤ x), where the logistic sigmoid function is deﬁned as σ(z) = 1/(1 + e−z ). This classiﬁcation model is known as logistic regression. 2.3 Estimation/Learning Problem Given the training data D consisting of N instances with annotations from R annotators, that is, D = {xi , y1 , . . . , yR }N , the task is to estimate the weight vector w and also the sensitivity α = i i i=1 [α1 , . . . , αR ] and the speciﬁcity β = [β1 , . . . , βR ] of the R annotators. It is also of interest to get an estimate of the unknown gold standard y1 , . . . , yN . 2.4 Maximum Likelihood Estimator Assuming the training instances are independently sampled, the likelihood function of the parameters θ = {w, α, β} given the observations D can be factored as N Pr[D |θ] = ∏ Pr[y1 , . . . , yR |xi , θ]. i i i=1 8. While this is a reasonable assumption, it is not entirely true. It is known that some radiologists are good at detecting certain kinds of malignant lesions based on their training and experience. 1301 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY j Conditioning on the true label yi , and also using the assumption yi is conditionally independent (of everything else) given α j , β j and yi , the likelihood can be decomposed as N Pr[D |θ] = ∏ Pr[y1 , . . . , yR |yi = 1, α]Pr[yi = 1|xi, w] i i i=1 + Pr[y1 , . . . , yR |yi = 0, β]Pr[yi = 0|xi, w] . i i Given the true label yi , we assume that y1 , . . . , yR are independent, that is, the annotators make their i i decisions independently.9 Hence, R R Pr[y1 , . . . , yR |yi = 1, α] = ∏ Pr[yi |yi = 1, α j ] = ∏ [α j ]yi [1 − α j ]1−yi . i i j j=1 j j j=1 Similarly, we have R Pr[y1 , . . . , yR |yi = 0, β] = ∏ [β j ]1−yi [1 − β j ]yi . i i j j j=1 Hence the likelihood can be written as N Pr[D |θ] = ∏ ai pi + bi (1 − pi ) , i=1 where we have deﬁned pi := σ(w⊤ xi ). R ai := ∏[α j ]y [1 − α j ]1−y . j i j i j=1 R bi := ∏[β j ]1−y [1 − β j ]y . j i j i j=1 The maximum-likelihood estimator is found by maximizing the log-likelihood, that is, ˆ θML = {α, β, w} = arg max{ln Pr[D |θ]}. ˆ ˆ ˆ θ 2.5 The EM Algorithm This maximization problem can be simpliﬁed a lot if we use the Expectation-Maximization (EM) algorithm (Dempster et al., 1977). The EM algorithm is an efﬁcient iterative procedure to compute the maximum-likelihood solution in presence of missing/hidden data. We will use the unknown hidden true label yi as the missing data. If we know the missing data y = [y1 , . . . , yN ] then the complete likelihood can be written as N ln Pr[D , y|θ] = ∑ yi ln pi ai + (1 − yi ) ln(1 − pi )bi . i=1 9. This assumption is not true in general and there is some correlations among the labels assigned by multiple annotators. For example in the CAD domain if the cancer is in advanced stage (which is very easy to detect) almost all the radiologists assign the same label. 1302 L EARNING F ROM C ROWDS Each iteration of the EM algorithm consists of two steps: an Expectation(E)-step and a Maximization(M)step. The M-step involves maximization of a lower bound on the log-likelihood that is reﬁned in each iteration by the E-step. 1. E-step. Given the observation D and the current estimate of the model parameters θ, the conditional expectation (which is a lower bound on the true likelihood) is computed as N E {ln Pr[D , y|θ]} = ∑ µi ln pi ai + (1 − µi ) ln(1 − pi )bi , (3) i=1 where the expectation is with respect to Pr[y|D , θ], and µi = Pr[yi = 1|y1 , . . . , yR , xi , θ]. Using i i Bayes’ theorem we can compute µi ∝ Pr[y1 , . . . , yR |yi = 1, θ] · Pr[yi = 1|xi, θ] i i ai pi = . ai pi + bi (1 − pi ) 2. M-step. Based on the current estimate µi and the observations D , the model parameters θ are then estimated by maximizing the conditional expectation. By equating the gradient of (3) to zero we obtain the following estimates for the sensitivity and speciﬁcity: j j αj = ∑N µi yi i=1 , ∑N µi i=1 βj = ∑N (1 − µi )(1 − yi ) i=1 . ∑N (1 − µi ) i=1 Due to the non-linearity of the sigmoid, we do not have a closed form solution for w and we have to use gradient ascent based optimization methods. We use the Newton-Raphson update given by wt+1 = wt − ηH −1 g, where g is the gradient vector, H is the Hessian matrix, and η is the step length. The gradient vector is given by N g(w) = ∑ µi − σ(w⊤ xi ) xi . i=1 The Hessian matrix is given by N H(w) = − ∑ σ(w⊤ xi ) 1 − σ(w⊤ xi ) xi x⊤ . i i=1 Essentially, we are estimating a logistic regression model with probabilistic labels µi . These two steps (the E- and the M-step) can be iterated till convergence. The log-likelihood increases monotonically after every iteration, which in practice implies convergence to a local maximum. The EM algorithm is only guaranteed to converge to a local maximum. In practice multiple restarts with different initializations can potentially mitigate the local maximum problem. In this j paper we use majority voting µi = 1/R ∑R yi as the initialization for µi to start the EM-algorithm. j=1 1303 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY 2.6 A Bayesian Approach In some applications we may want to trust a particular expert more than the others. One way to achieve this is by imposing priors on the sensitivity and speciﬁcity of the experts. Since α j and β j represent the probability of a binary event, a natural choice of prior is the beta prior. The beta prior is also conjugate to the binomial distribution. For any a > 0, b > 0, and δ ∈ [0, 1] the beta distribution is given by δa−1 (1 − δ)b−1 , Beta(δ|a, b) = B(a, b) where B(a, b) = 01 δa−1 (1 − δ)b−1 dδ is the beta function.We assume a beta prior10 for both the sensitivity and the speciﬁcity as R j j j j j j j j Pr[α j |a1 , a2 ] = Beta(α j |a1 , a2 ). Pr[β j |b1 , b2 ] = Beta(β j |b1 , b2 ). For sake of completeness we also assume a zero mean Gaussian prior on the weights w with inverse covariance matrix Γ, that is, Pr[w] = N (w|0, Γ−1 ). Assuming that {α j }, {β j }, and w have independent priors, the maximum-a-posteriori (MAP) estimator is found by maximizing the logposterior, that is, ˆ θMAP = arg max{ln Pr[D |θ] + ln Pr[θ]}. θ An EM algorithm can be derived in a similar fashion for MAP estimation by relying on the interpretation of Neal and Hinton (1998). The ﬁnal algorithm is summarized below: j 1. Initialize µi = (1/R) ∑R yi based on majority voting. j=1 2. Given µi , estimate the sensitivity and speciﬁcity of each annotator/expert as follows. j αj = j a1 − 1 + ∑N µi yi i=1 j j a1 + a2 − 2 + ∑N µi i=1 j β j = . j b1 − 1 + ∑N (1 − µi )(1 − yi ) i=1 j j b1 + b2 − 2 + ∑N (1 − µi ) i=1 . (4) The Newton-Raphson update for optimizing w is given by wt+1 = wt − ηH −1 g, with step length η, gradient vector N g(w) = ∑ µi − σ(w⊤ xi ) xi − Γw, i=1 and Hessian matrix N H(w) = − ∑ σ(w⊤ xi ) 1 − σ(w⊤ xi ) xi x⊤ − Γ. i i=1 10. It may be convenient to specify a prior in terms of the mean µ and variance σ2 . The mean and the variance for a beta prior are given by µ = a/(a + b) and σ2 = ab/((a + b)2 (a + b + 1)). Solving for a and b we get a = (−µ3 + µ2 − µσ2 )/σ2 and b = a(1 − µ)/µ. 1304 L EARNING F ROM C ROWDS 3. Given the sensitivity and speciﬁcity of each annotator and the model parameters, update µi as µi = ai pi , ai pi + bi (1 − pi ) (5) where pi = σ(w⊤ xi ). R ai = ∏[α j ]y [1 − α j ]1−y . j i j i j=1 R bi = ∏[β j ]1−y [1 − β j ]y . j i j i (6) j=1 Iterate (2) and (3) till convergence. 2.7 Discussions 1. Estimate of the gold standard The value of the posterior probability µi is a soft probabilistic estimate of the actual ground truth yi , that is, µi = Pr[yi = 1|y1 , . . . , yR , xi , θ]. The actual i i hidden label yi can be estimated by applying a threshold on µi , that is, yi = 1 if µi ≥ γ and zero otherwise. We can use γ = 0.5 as the threshold. By varying γ we can change the misclassiﬁcation costs and obtain a ground truth with large sensitivity or large speciﬁcity. Because of this in our experimental validation we can actually draw an ROC curve for the estimated ground truth. 2. Log-odds of µ A particularly revealing insight can be obtained in terms of the log-odds or the logit of the posterior probability µi . From (5) the logit of µi can be written as logit(µi ) = ln Pr[yi = 1|y1 , . . . , yR , xi , θ] µi i i = ln 1 − µi Pr[yi = 0|y1 , . . . , yR , xi , θ] i i R = w⊤ xi + c + ∑ yi [logit(α j ) + logit(β j )]. j j=1 j where c = ∑R log 1−α is a constant term which does not depend on i. This indicates that j=1 βj the estimated ground truth (in the logit form of the posterior probability) is a weighted linear combination of the labels from all the experts. The weight of each expert is the sum of the logit of the sensitivity and speciﬁcity. 3. Using any other classiﬁer For ease of exposition we used logistic regression. However, the proposed algorithm can be used with any generalized linear model or in fact with any classiﬁer that can be trained with soft probabilistic labels. In each step of the EM-algorithm, the classiﬁer is trained with instances sampled from µi . This modiﬁcation is easy for most probabilistic classiﬁers. For general black-box classiﬁers where we cannot tweak the training algorithm an alternate approach is to replicate the training examples according to the soft label. For example a probabilistic label µi = 0.8 can be effectively simulated by adding 8 training examples with deterministic label 1 and 2 examples with label 0. 1305 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY 4. Obtaining ground truth with no features In some scenarios we may not have features xi and we wish to obtain an estimate of the actual ground truth based only on the labels from multiple annotators. Here instead of learning a classiﬁer we estimate p which is the prevalence of the positive class, that is, p = Pr[yi = 1]. We further assume a beta prior for the prevalence, that is, Beta(p|p1 , p2 ). The algorithm simpliﬁes as follows. j (a) Initialize µi = (1/R) ∑R yi based on majority voting. j=1 (b) Given µi , estimate the sensitivity and speciﬁcity of each annotator using (4). The prevalence of the positive class is estimated as follows. p = p1 − 1 + ∑N µi i=1 . p1 + p2 − 2 + N (c) Given the sensitivity and speciﬁcity of each annotator and prevalence, reﬁne µi as follows. ai p µi = . ai p + bi (1 − p) Iterate (2) and (3) till convergence. This algorithm is similar to the one proposed by Dawid and Skene (1979) and Smyth et al. (1995). 5. Handling missing labels The proposed approach can easily handle missing labels, that is, when the labels from some experts are missing for some instances. Let Ri be the number of radiologists labeling the ith instance, and let N j be the number of instances labeled by the jth radiologist. Then in the EM algorithm, we just need to replace N by N j for estimating the sensitivity and speciﬁcity in (4), and replace R by Ri for updating µi in (6). 6. Evaluating a classiﬁer We can use the probability scores µi directly to evaluate classiﬁers. If zi are the labels obtained from any other classiﬁer, then sensitivity and speciﬁcity can be estimated as ∑N (1 − µi )(1 − zi ) ∑N µi zi , β = i=1 N . α = i=1 ∑N µi ∑i=1 (1 − µi ) i=1 7. Posterior approximation At the end of each EM iteration a crude approximation to the posterior is obtained as N N α j ∼ Beta α j |a1 + ∑ µi yi , a2 + ∑ µi (1 − yi ) , j j i=1 N j j i=1 N β j ∼ Beta β j |b1 + ∑ (1 − µi )(1 − yi ), b2 + ∑ (1 − µi )yi j j i=1 j j . i=1 3. Multi-class Classiﬁcation In this section we describe how the proposed approach for binary classiﬁcation can be extended to categorical data. Suppose there are K ≥ 2 categories. An example for categorical data from the CAD domain is in LungCAD, where the radiologist needs to label whether a nodule (known to be precursors of cancer) is a solid, a part-solid, or a ground glass opacity—which are three 1306 L EARNING F ROM C ROWDS different kinds on nodules. We can extend the previous model and introduce a vector of multinomial j j j parameters αc = (αc1 , . . . , αcK ) for each annotator, where αck := Pr[y j = k|y = c] j and ∑K αck = 1. Here αck denotes the probability that the annotator j assigns class k to an instance k=1 j j given the true class is c. When K = 2, α11 and α00 are sensitivity and speciﬁcity, respectively. A similar EM algorithm can be derived. In the E-step, we estimate j j R K Pr[yi = c|Y , α] ∝ Pr[yi = c|xi ] ∏ ∏ (αck )δ(yi ,k) , j j j=1 k=1 where δ(u, v) = 1 if u = v and 0 otherwise and in the M-step we learn a multi-class classiﬁer and update the multinomial parameter as j αck = j ∑N Pr[yi = c|Y , α]δ(yi , k) i=1 . ∑N Pr[yi = c|Y , α] i=1 One can also assign a Dirichlet prior for the multinomial parameters, and this results in a smoothing term in the above updates in the MAP estimate. 4. Ordinal Regression We now consider the situation where the outputs are categorical and have an ordering among the labels. In the CAD domain the radiologist often gives a score (for example, 1 to 5 from lowest to highest) to indicate how likely she thinks it is malignant. This is different from a multi-class setting in which we do not have any preference among the multiple class labels. j Let yi ∈ {1, . . . , K} be the label assigned to the ith instance by the jth expert. Note that there is an ordering in the labels 1 < . . . < K. A simple approach is to convert the ordinal data into a series of binary data (Frank and Hall, 2001). Speciﬁcally the K class ordinal labels are transformed into K − 1 binary class labels as follows: jc yi = j 1 if yi > c 0 otherwise c = 1, . . . , K − 1. Applying the same procedure used for binary labels we can estimate Pr[yi > c] for c = 1, . . . , K − 1. The probability of the actual class values can then be obtained as Pr[yi = c] = Pr[yi > c − 1 and yi ≤ c] = Pr[yi > c − 1] − Pr[yi > c]. The class with the maximum probability is assigned to the instance. 5. Regression In this section we develop a similar algorithm to learn a regression function using annotations from multiple experts. In the CAD domain as a part of the annotation process a common task for a radiologist is to measure the diameter of a suspicious lesion. 1307 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY 5.1 Model for Annotators j Let yi ∈ R be the continuous target value assigned to the ith instance by the jth annotator. Our model is that the annotator provides a noisy version of the actual true value yi . For the jth annotator we will assume a Gaussian noise model with mean yi (the true unknown value) and inverse-variance (precision) τ j , that is, j j Pr[yi |yi , τ j ] = N (yi |yi , 1/τ j ), (7) where the Gaussian distribution is deﬁned as N (z|m, σ2 ) = (2πσ2 )−1/2 exp(−(z − m)2 /2σ2 ). The unknown inverse-variance τ j measures the accuracy of each annotator—the larger the value of τ j the more accurate the annotator. We have assumed that τ j does not depend on the instance xi . For example, in the CAD domain, this means that the radiologist’s accuracy does not depend on the nodule she is measuring. While this a practical assumption, it is not entirely true. It is known that some nodules are harder to measure than others. 5.2 Linear Regression Model for Features As before we consider the family of linear regression functions: F = { fw}, where for any x, w ∈ Rd , fw(x) = w⊤ x. We assume that the actual target response yi is given by the deterministic regression function fw with additive Gaussian noise, that is, yi = w⊤ xi + ε, where ε is a zero-mean Gaussian random variable with inverse-variance (precision) γ. Hence Pr[yi |xi , w, γ] = N (yi |w⊤ xi , 1/γ). (8) 5.3 Combined Model Combining both the annotator (7) and the regressor (8) model we have Pr[yi |xi , w, τ j , γ] = j Z Pr[yi |yi , τ j ]Pr[yi |xi , w, γ]dyi = N (yi |w⊤ xi , 1/γ + 1/τ j ). j j Since the two precision terms (γ and τ j ) are grouped together they are not uniquely identiﬁable. Hence we will deﬁne a new precision term λ j as 1 1 1 = + j. j λ γ τ So we have the following model Pr[yi |xi , w, λ j ] = N (yi |w⊤ xi , 1/λ j ). j j (9) 5.4 Estimation/Learning Problem Given the training data D consisting of N instances with annotations from R experts, that is, D = {xi , y1 , . . . , yR }N , the task is to estimate the weight vector w and the precision λ = [λ1 , . . . , λR ] of i i=1 i all the annotators. 1308 L EARNING F ROM C ROWDS 5.5 Maximum-likelihood Estimator Assuming the instances are independent the likelihood of the parameters θ = {w, λ} given the observations D can be factored as N Pr[D |θ] = ∏ Pr[y1 , . . . , yR |xi , θ]. i i i=1 Conditional on the instance xi we assume that y1 , . . . , yR are independent, that is, the annotators i i provide their responses independently. Hence from (9) the likelihood can be written as N R Pr[D |θ] = ∏ ∏ N (yi |w⊤ xi , 1/λ j ). j i=1 j=1 The maximum-likelihood estimator is found by maximizing the log-likelihood θML = {λ, w} = arg max{ln Pr[D |θ]}. θ By equating the gradient of the log-likelihood to zero we obtain the following update equations for the precision and the weight vector. 1 λj = 1 N ∑ yij − w⊤ xi N i=1 N w = ∑ −1 2 . (10) N ∑R λ j yi j=1 i=1 ∑R λ j j=1 ∑ xi xi x⊤ i i=1 j . (11) As the parameters w and λ are coupled together we iterate these two steps till convergence. 5.6 Discussions 1. Is this standard least-squares? Deﬁne the design matrix X = [x1 , . . . , xN ]⊤ and the rej j sponse vector for each annotator as y j = [y1 , . . . , yN ]⊤ . Using matrix notation Equation 11 can be written as w = (X ⊤ X)−1 X ⊤ y where y = ∑R λ j y j j=1 ∑R λ j j=1 . (12) Equation 12 is essentially the solution to a standard linear regression model, except that we are training a linear regression model with y as the ground truth, which is a precision weighted mean of the response vectors from all the annotators. The variance of each annotator is estimated using (10). The ﬁnal algorithm iteratively establishes a particular gold standard (y), measures the performance of the annotators and learns a regressor given that gold standard, and reﬁnes the gold standard based on the performance measures. 2. Are we better than the best annotator? If we assume λ is ﬁxed (i.e., we ignore the variability and assume that it is well estimated) then w is an unbiased estimator of w and the ˆ covariance matrix is given by Cov(w) = Cov(y) X ⊤ X ˆ 1309 −1 = 1 ∑R λ j j=1 X ⊤X −1 . R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY Since ∑R λ j > max j (λ j ) the proposed method has a lower variance than the regressor learnt j=1 with the best annotator (i.e., the one with the minimum variance). 3. Are we better than the average? For a ﬁxed X the error in w depends only on the variance of y j . If we know the true λ j then yi is the best linear unbiased estimator for yi which minij mizes the variance. To see this consider any linear estimator of the form yi = ∑ j a j (yi − b j ). The variance is given by Var[yi ] = ∑ j (a j )2 /λ j . Since E[yi ] = yi ∑ j a j , for the bias of this estimator to be zero we require that ∑ j a j = 1. Solving the constrained minimization problem we see that a j = λ j / ∑ j λ j minimizes the variance. 4. Obtaining a consensus without features When no features are available the same algorithm can be simpliﬁed to get a consensus estimate of the actual ground truth and also evaluate the annotators. Essentially we have to iterate the following two updates till convergence ∑R λ j yi j=1 1 ∑R λ j j=1 λj j yi = = 1 N ∑ y j − yi N i=1 i 2 . 6. Experimental Validation We now experimentally validate the proposed algorithms on both simulated and real data. 6.1 Classiﬁcation Experiments We use two CAD and one text data set in our experiments. The CAD data sets include a digital mammography data set and a breast MRI data set, both of which are biopsy proven, that is, the gold standard is available. For the digital mammography data set we simulate the radiologists in order to validate our methods. The breast MRI data has annotations from four radiologists. We also report results on a Recognizing Textual Entailment data collected by Snow et al. (2008) using the Amazon’s Mechanical Turk which has annotations from 164 annotators. 6.1.1 D IGITAL M AMMOGRAPHY WITH S IMULATED R ADIOLOGISTS Mammograms are used as a screening tool to detect early breast cancer. CAD systems search for abnormal areas (lesions) in a digitized mammographic image. These lesions generally indicate the presence of malignant cancer. The CAD system then highlights these areas on the images, alerting the radiologist to the need for a further diagnostic mammogram or a biopsy. In classiﬁcation terms, given a set of descriptive morphological features for a region on a image, the task is to predict whether it is potentially malignant (1) or not (0). In order to train such a classiﬁer, a set of mammograms is collected from hospitals. The ground truth (whether it is cancer or not) is obtained from biopsy. Since biopsy is an expensive, tedious, and an invasive process, very often CAD systems are built from labels collected from multiple expert radiologists who visually examine the mammograms and mark the lesion locations—this constitutes our ground truth (multiple labels) for learning. In this experiment we use a proprietary biopsy-proven data set (Krishnapuram et al., 2008) containing 497 positive and 1618 negative examples. Each instance is described by a set of 27 morphological features. In order to validate our proposed algorithm, we simulate multiple radiologists according to the two-coin model described in § 2.1. Based on the labels from multiple radiologists, 1310 L EARNING F ROM C ROWDS we can simultaneously (1) learn a logistic-regression classiﬁer, (2) estimate the sensitivity and speciﬁcity of each radiologist, and (3) estimate the golden ground truth. We compare the results with the classiﬁer trained using the biopsy proved ground truth as well as the majority-voting baseline. For the ﬁrst set of experiments we use 5 radiologists with sensitivity α = [0.90 0.80 0.57 0.60 0.55] and speciﬁcity β = [0.95 0.85 0.62 0.65 0.58]. This corresponds to a scenario where the ﬁrst two radiologists are experts and the last three are novices. Figure 1 summarizes the results. We compare on three different aspects: (1) How good is the learnt classiﬁer? (2) How well can we estimate the sensitivity and speciﬁcity of each radiologist? (3) How good is the estimated ground truth? The following observations can be made. 1. Classiﬁer performance Figure 1(a) plots the ROC curve of the learnt classiﬁer on the training set. The dotted (black) line is the ROC curve for the classiﬁer learnt using the actual ground truth. The solid (red) line is the ROC curve for the proposed algorithm and the dashed (blue) line is for the classiﬁer learnt using the majority-voting scheme. The classiﬁer learnt using the proposed method is as good as the one learnt using the golden ground truth. The area under the ROC curve (AUC) for the proposed algorithm is around 3.5% greater than that learnt using the majority-voting scheme. 2. Radiologist performance The actual sensitivity and speciﬁcity of each radiologist is marked as a black × in Figure 1(b). The end of the solid red line shows the estimates of the sensitivity and speciﬁcity from the proposed method. We used a uniform prior on all the parameters. The ellipse plots the contour of one standard deviation as obtained from the beta posterior estimates. The end of the dashed blue line shows the estimate obtained from the majorityvoting algorithm. We see that the proposed method is much closer to the actual values of sensitivity and speciﬁcity. 3. Actual ground truth Since the estimates of the actual ground truth are probabilistic scores, we can also plot the ROC curves of the estimated ground truth. From Figure 1(b) we can see that the ROC curve for the proposed method dominates the majority voting ROC curve. Furthermore, the area under the ROC curve (AUC) is around 3% higher. The estimate obtained by majority voting is closer to the novices since they form a majority (3/5). It does not have an idea of who is an expert and who is a novice. The proposed algorithm appropriately weights each radiologist based on their estimated sensitivity and speciﬁcity. The improvement obtained is quite large in Figure 2 which corresponds a situation where we have only one expert and 7 novices. 4. Joint Estimation To learn a classiﬁer, Smyth et al. (1995) proposed to ﬁrst estimate the golden ground truth and then use the probabilistic ground truth to learn a classiﬁer. In contrast, our proposed algorithm learns the classiﬁer and the ground truth jointly as a part of the EM algorithm. Figure 3 shows that the classiﬁer and the ground truth learnt obtained by the proposed algorithm is superior than that obtained by other procedures which ﬁrst estimates the ground truth and then learns the classiﬁer. 6.1.2 B REAST MRI In this example, each radiologist reviews the breast MRI data and assesses the malignancy of each lesion on a BIRADS scale of 1 to 5. The BIRADS scale is deﬁned as follows: 1 Negative, 2 Benign, 1311 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY Majority Voting Estimated 1 Estimated 2 Estimated 3 Estimated 4 Estimated 5 EM algorithm Estimated 1 Estimated 2 Estimated 3 Estimated 4 Estimated 5 True 1 True 2 True 3 True 4 True 5 x x x x x 0.0217 0.5869 0.2391 0.1521 0.0000 0 0 0 1 0 x x x x x 0.0000 0.1785 0.1071 0.2500 0.4642 True 1 True 2 True 3 True 4 True 5 x x x x x 0.0000 0.6957 0.1304 0.1739 0.0000 0 0 0 1 0 x x x x x 0.0000 0.1428 0.0000 0.3214 0.5357 Table 1: The confusion matrix for the estimate obtained using majority voting and the proposed EM algorithm. The x indicates that there was no such category in the true labels (the gold standard). The gold-standard is obtained by the biopsy which can conﬁrm whether it is benign (BIRADS=2) or malignant (BIRADS=5). 3 Probably Benign, 4 Suspicious abnormality, and 5 Highly suggestive of malignancy. Our data set comprises of 75 lesions with annotations from four radiologists, and the true labels from biopsy. Based on eight morphological features, we have to predict whether a lesion is malignant or not. For the ﬁrst experiment we reduce the BIRADS scale to a binary one: any lesion with a BIRADS > 3 is considered malignant and benign otherwise. The set included 28 malignant and 47 benign lesions. Figure 4 summarizes the results. We show the leave-one-out cross validated ROC for the classiﬁer. The cross-validated AUC of the proposed method is approximately 6% better than the majority voting baseline. We also consider the BIRADS labels as a set of ordinal measurements since there is an ordering among the BIRADS label. The confusion matrix in Table 1 shows that the EM algorithm is signiﬁcantly superior than the majority voting in estimating the true BIRADS. 6.1.3 R ECOGNIZING T EXTUAL E NTAILMENT Finally we report results on Recognizing Textual Entailment data collected by Snow et al. (2008) using the Amazon’s Mechanical Turk. In this task, the annotator is presented with two sentences and given a choice of whether the second sentence can be inferred from the ﬁrst. The data has 800 tasks and 164 distinct readers, with 10 annotations per task along with the golden ground truth. The majority of the entries (94 %) in the 800x164 matrix are missing. There is one annotator who has labeled all the tasks. We use this data set to obtain an estimate of the actual ground truth. Figure 5 plots the accuracy of the estimated ground truth as a function of the number of annotators. The proposed EM algorithm achieves a higher accuracy than majority voting. In other words to achieve a desired accuracy the proposed algorithm needs fewer annotators than the majority voting scheme. 1312 L EARNING F ROM C ROWDS ROC Curve for the classifier 1 True Positive Rate (sensitivity) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 Golden ground truth AUC=0.915 Proposed EM algorithm AUC=0.913 Majority voting baseline AUC=0.882 0.1 0 0 0.2 0.4 0.6 False Positive Rate (1−specifcity) 0.8 1 (a) ROC Curve for the estimated true labels 1 True Positive Rate (sensitivity) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 Proposed EM algorithm AUC=0.991 Majority voting baseline AUC=0.962 0.2 0.1 0 0 0.2 0.4 0.6 False Positive Rate (1−specifcity) 0.8 1 (b) Figure 1: Results for the digital mammography data set with annotations from 5 simulated radiologists. (a) The ROC curve of the learnt classiﬁer using the golden ground truth (dotted black line), the majority voting scheme (dashed blue line), and the proposed EM algorithm (solid red line). (b) The ROC curve for the estimated ground truth. The actual sensitivity and speciﬁcity of each of the radiologists is marked as a ×. The end of the dashed blue line shows the estimates of the sensitivity and speciﬁcity obtained from the majority voting algorithm. The end of the solid red line shows the estimates from the proposed method. The ellipse plots the contour of one standard deviation. 1313 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY ROC Curve for the classifier 1 True Positive Rate (sensitivity) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 Golden ground truth AUC=0.915 Proposed EM algorithm AUC=0.906 Majority voting baseline AUC=0.884 0.1 0 0 0.2 0.4 0.6 False Positive Rate (1−specifcity) 0.8 1 (a) ROC Curve for the estimated true labels 1 True Positive Rate (sensitivity) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 Proposed EM algorithm AUC=0.967 Majority voting baseline AUC=0.872 0.2 0.4 0.6 False Positive Rate (1−specifcity) 0.8 1 (b) Figure 2: Same as Figure 1 except with 8 different radiologist annotations. 1314 L EARNING F ROM C ROWDS ROC Curve for the classifier 1 True Positive Rate (sensitivity) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 Proposed EM algorithm [Joint Estimation] AUC=0.905 Decoupled Estimation AUC=0.884 0.2 0.4 0.6 False Positive Rate (1−specifcity) 0.8 1 (a) ROC Curve for the estimated true labels 1 True Positive Rate (sensitivity) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 Proposed EM algorithm [Joint Estimation] AUC=0.972 Decoupled Estimation AUC=0.921 0.2 0.4 0.6 False Positive Rate (1−specifcity) 0.8 1 (b) Figure 3: ROC curves comparing the proposed algorithm (solid red line) with the Decoupled Estimation procedure (dotted blue line), which refers to the algorithm where the ground truth is ﬁrst estimated using just the labels from the ﬁve radiologists and then a logistic regression classiﬁer is trained using the soft probabilistic labels. In contrast the proposed EM algorithm estimates the ground truth and learns the classiﬁer simultaneously during the EM algorithm. 1315 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY Leave−One−Out ROC Curve for the classifier 1 True Positive Rate (sensitivity) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 Golden ground truth AUC=0.909 Majority voting baseline AUC=0.828 Proposed EM algorithm AUC=0.879 0.2 0.1 0 0 0.2 0.4 0.6 False Positive Rate (1−specifcity) 0.8 1 (a) ROC Curve for the estimated true labels 1 0.9 True Positive Rate (sensitivity) 0.8 0.7 0.6 0.5 0.4 Proposed EM algorithm AUC=0.944 Majority voting baseline AUC=0.937 0.3 0.2 0.1 0 0 0.2 0.4 0.6 False Positive Rate (1−specifcity) 0.8 1 (b) Figure 4: Breast MRI results. (a) The leave-one-out cross validated ROC. (b) ROC for the estimated ground truth. 1316 L EARNING F ROM C ROWDS 0.95 0.9 0.85 Accuracy 0.8 0.75 0.7 0.65 Majority Voting EM Algorithm 0.6 0.55 0.5 20 40 60 80 100 120 Number of Annotators 140 160 Figure 5: The mean and the one standard deviation error bars for the accuracy of the estimated ground truth for the Recognizing Textual Entailment task as a function of the number of annotators. The plot was generated by randomly sampling the annotators 100 times. 6.2 Regression Experiments We ﬁrst illustrate the algorithm on a toy dataset and then present a case study for automated polyp measurements. 6.2.1 I LLUSTRATION Figure 6 illustrates the the proposed algorithm for regression on a one-dimensional toy data set with three annotators. The actual regression model (shown as a blue dotted line) is given by y = 5x − 2. We simulate 20 samples from three annotators with precisions 0.01, 0.1, and 1.0. The data are shown by the annotators’s number. While we can ﬁt a regression model using each annotators’s response, we see that only the model for annotator three (with highest precision) is close to the true regression model. The green dashed line shows the model learnt using the average response from all the three annotators. The red line shows the model learnt by the proposed algorithm. 6.2.2 AUTOMATED P OLYP M EASUREMENTS Colorectal polyps are small colonic ﬁndings that may develop into cancer at a later stage. The diameter of the polyp is one of the key factors which decides the malignancy of a suspicious polyp. Hence accurate size estimation is crucial to decide the action to be taken on a polyp. We have developed various algorithms to segment a polyp. Multiple segmentation algorithms give rise to a set of features which are correlated with the diameter of the polyp. We want to learn a regression function which can predict the diameter of a polyp as a function of these features. In order to learn a regression function we collect our ground truth by asking many radiologists to manually measure the the diameter of the polyps from the three-dimensional images. In practice there is a lot of disagreement among the radiologists as to the actual size of the polyp. 1317 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY 2 1 N=50 examples R=3 annotators 2 2 5 Actual regression model Model with annotator 1 (Precision=0.01) 2 1 Model with annotator 2 (Precision=0.10) 2 Model with annotator 3 (Precision=1.00) 3 2 3 2 Model with average response 2 Proposed algorithm 2 3 1 1 4 3 2 2 y (response) 2 2 1 3 2 3 3 2 3 1 −1 3 3 2 2 3 3 1 2 2 2 3 33 3 1 1 3 3 3 3 2 3 3 2 33 33 2 2 2 3 3 3 32 1 2 3 2 1 3 3 2 1 2 1 2 2 1 1 2 3 2 3 2 3 1 3 3 −3 2 3 3 2 3 3 3 2 3 3 2 1 3 3 2 0 −2 2 2 3 2 3 2 3 1 2 2 1 2 3 −4 1 2 2 1 −5 0 2 0.1 0.2 1 0.3 0.4 0.5 0.6 10.8 2 0.7 x (feature) 0.9 1 1 Figure 6: Illustration of the proposed algorithm on a one-dimensional toy data set. The actual regression model (shown as a blue dotted line) is given by y = 5x − 2. We simulate 50 samples from three annotators with precisions 0.01, 0.1, and 1.0. The data are shown by the annotators’s number. While we can ﬁt a regression model using each annotators’s response, we see that only the model for annotator three (with highest precision) is close to the true regression model. The green dashed line shows the model learnt using the average response from all the three annotators. The red line shows the model learnt by the proposed algorithm. 5 5 10 Actual Polyp Diameter (mm) (a) Gold standard model 15 10 5 0 0 Pearson Correlation Coefficient=0.554966 RMSE=2.887740 15 Estimated Polyp Diameter (mm) 10 0 0 Pearson Correlation Coefficient=0.706558 RMSE=1.991576 15 Estimated Polyp Diameter (mm) Estimated Polyp Diameter (mm) Pearson Correlation Coefficient=0.714720 RMSE=1.970815 15 5 10 Actual Polyp Diameter (mm) (b) Proposed model 15 10 5 0 0 5 10 Actual Polyp Diameter (mm) 15 (c) Average model Figure 7: Scatter plot of the actual polyp diameter vs the diameter predicted by the models learnt using (a) the actual gold standard, (b) the proposed algorithm with annotations from ﬁve radiologists, and (c) the average of the radiologist’s annotations. (See § 6.2.2 for a description of the experimental setup.) 1318 L EARNING F ROM C ROWDS We use a proprietary data set containing 393 examples (which point to 285 distinct polyps— the segmentation algorithms generally return multiple marks on the same polyp.) along with the measured diameter (ranging from 2mm to 15mm) as our training set. Each example is described by a set of 60 morphological features which are correlated to the diameter of the polyp. In order to validate the feasibility of our proposed algorithm, we simulate ﬁve radiologists according to the noisy model described in § 5.1 with τ = [0.001 0.01 0.1 1 10]. This corresponds to a situation where the ﬁrst three radiologists are extremely noisy and the last two are quite accurate. Based on the measurements from multiple radiologists, we can simultaneously (1) learn a linear regressor and (2) estimate the precision of each radiologist. We compare the results with the classiﬁer trained using the actual golden ground truth as well as the regressor learnt using the average of the radiologists measurements. The results are validated on an independent test set containing 397 examples (which point to 298 distinct polyps). Figure 7 shows the scatter plot of the actual polyp diameter vs the diameter predicted by the three different models. We compare the performance based on the root mean squared error (RMSE) and also the Pearson’s correlation coefﬁcient. The regressor learnt using the proposed iterative algorithm (Figure 7(b)) is almost as good as the one learnt using the golden ground truth (Figure 7(a)). The correlation coefﬁcient for the proposed algorithm is signiﬁcantly larger than that learnt using the average of the radiologists response. The estimate obtained by averaging is closer to the novices since they form a majority (3/5). The proposed algorithm appropriately weights each radiologist based on their estimated precisions. 7. Conclusions and Future Work In this paper we proposed a probabilistic framework for supervised learning with multiple annotators providing labels but no absolute gold standard. The proposed algorithm iteratively establishes a particular gold standard, measures the performance of the annotators given that gold standard, and then reﬁnes the gold standard based on the performance measures. We speciﬁcally discussed binary/categorical/ordinal classiﬁcation and regression problems. We made two key assumptions: (1) the performance of each annotator does not depend on the feature vector for a given instance and (2) conditional on the truth the experts are independent, that is, they make their errors independently. As we pointed out earlier these assumptions are not true in practice. The annotator performance depends on the instance he is labeling and there is some degree of correlation among the annotators. We brieﬂy discuss some strategies to relax these two assumptions. 7.1 Instance Difﬁculty One drawback of the current model is that it doesn’t estimate difﬁculty of items. It is often observed that for the easy instances all the annotators agree on the labels—thus violating our conditional independence assumption. The difﬁculty of annotating an item can be captured by another latent variable γi for each instance—which modulates the annotators performance. Models for this have been developed in the area of item-response theory (Baker and Kim, 2004) and also in epidemiology (Uebersax and Grove, 1993)—see also Whitehill et al. (2009) for a recent paper in the machine learning community. While these models do not take into account the available features our pro1319 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY posed model for sensitivity and speciﬁcity can be extended as follows (in place of (1) and (2)): α j (γi ) := Pr[yi = 1|yi = 1, γi ] = σ(a j1 + b j1 γi ). j β j (γi ) := Pr[yi = 0|yi = 0, γi ] = σ(a j0 + b j0 γi ). j Here the parameters a j1 and a j0 are related to the sensitivity and speciﬁcity of the jth annotator, while the latent term γi captures the difﬁculty of the instance. The key assumption here is that the annotators are independent conditional on both yi and γi . Various assumptions can be made on two parameters b j1 and b j0 to simplify these models further—for example we could set b j1 = b1 and b j0 = b0 for all the annotators. 7.2 Annotators Actually Look at the Data In our model we made the assumption that the sensitivity α j and the speciﬁcity β j of the jth annotator does not depend on the feature vector xi . For example, in the CAD domain, this meant that the radiologist’s performance is consistent across different sub-groups of data—which is not entirely true. It is known that some radiologists are good at detecting certain kinds of malignant lesions based on their training and experience. We can extend the previous model such that the sensitivity and the speciﬁcity depends on the feature vector xi explicitly as follows j⊤ α j (γi , xi ) := Pr[yi = 1|yi = 1, γi , xi ] = σ(a j1 + b j1 γi + wα xi ). j j⊤ α j (γi , xi ) := Pr[yi = 0|yi = 0, γi , xi ] = σ(a j0 + b j0 γi + wβ xi ). j However this change increases the number of parameters to be learned. References P. S. Albert and L. E. Dodd. A cautionary note on the robustness of latent class models for estimating diagnostic error without a gold standard. Biometrics, 60:427–435, 2004. F. B. Baker and S. Kim. Item Response Theory: Parameter Estimation Techniques. CRC Press, 2 edition, 2004. B. Carpenter. Multilevel bayesian models of categorical data annotation. Technical Report available at http://lingpipe-blog.com/lingpipe-white-papers/, 2008. S. R. Cholleti, S. A. Goldman, A. Blum, D. G. Politte, and S. Don. Veritas: Combining expert opinions without labeled data. In Proceedings of the 2008 20th IEEE international Conference on Tools with Artiﬁcial intelligence, 2008. A. P. Dawid and A. M. Skene. Maximum likeihood estimation of observer error-rates using the EM algorithm. Applied Statistics, 28(1):20–28, 1979. O. Dekel and O. Shamir. Vox Populi: Collecting high-quality labels from a crowd. In COLT 2009: Proceedings of the 22nd Annual Conference on Learning Theory, 2009a. O. Dekel and O. Shamir. Good learners for evil teachers. In ICML 2009: Proceedings of the 26th International Conference on Machine Learning, pages 233–240, 2009b. 1320 L EARNING F ROM C ROWDS A. P. Dempster, N. M. Laird, and D. B. Rubin. Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society: Series B, 39(1):1–38, 1977. P. Donmez, J. G. Carbonell, and J. Schneider. Efﬁciently learning the accuracy of labeling sources for selective sampling. In KDD 2009: Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 259–268, 2009. E. Frank and M. Hall. A simple approach to ordinal classiﬁcation. Lecture Notes in Computer Science, pages 145–156, 2001. G. Fung, B. Krishnapuram, J. Bi, M. Dundar, V. C. Raykar, S. Yu, R. Rosales, S. Krishnan, and R. B. Rao. Mining medical images. In Fifteenth Annual SIGKDD International Conference on Knowledge Discovery and Data Mining: Third Workshop on Data Mining Case Studies and Practice Prize, 2009. J. Howe. Crowd sourcing: Why the Power of the Crowd Is Driving the Future of Business. 2008. S. L. Hui and S. D. Walter. Estimating the error rates of diagnostic tests. Biometrics, 36:167–171, 1980. S. L. Hui and X. H. Zhou. Evaluation of diagnostic tests without gold standards. Statistical Methods in Medical Research, 7:354–370, 1998. R. Jin and Z. Ghahramani. Learning with multiple labels. In Advances in Neural Information Processing Systems 15, pages 897–904. 2003. B. Krishnapuram, J. Stoeckel, V. C. Raykar, R. B. Rao, P. Bamberger, E. Ratner, N. Merlet, I. Stainvas, M. Abramov, and A. Manevitch. Multiple-instance learning improves CAD detection of masses in digital mammography. In IWDM 2008: Proceedings of the 9th international workshop on Digital Mammography, pages 350–357. 2008. G. Lugosi. Learning with an unreliable teacher. Pattern Recognition, 25(1):79–87, 1992. R. M. Neal and G. E. Hinton. A view of the EM algorithm that justiﬁes incremental, sparse, and other variants. In Learning in Graphical Models, pages 355–368. Kluwer Academic Publishers, 1998. V. C. Raykar, S. Yu, L .H. Zhao, A. Jerebko, C. Florin, G. H. Valadez, L. Bogoni, and L. Moy. Supervised learning from multiple experts: Whom to trust when everyone lies a bit. In ICML 2009: Proceedings of the 26th International Conference on Machine Learning, pages 889–896, 2009. V. S. Sheng, F. Provost, and P. G. Ipeirotis. Get another label? improving data quality and data mining using multiple, noisy labelers. In Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 614–622, 2008. P. Smyth. Learning with probabilistic supervision. In Computational Learning Theory and Natural Learning Systems 3, pages 163–182. MIT Press, 1995. 1321 R AYKAR , Y U , Z HAO , VALADEZ , F LORIN , B OGONI AND M OY P. Smyth, U. Fayyad, M. Burl, P. Perona, and P. Baldi. Inferring ground truth from subjective labelling of venus images. In Advances in Neural Information Processing Systems 7, pages 1085– 1092. 1995. R. Snow, B. O’Connor, D. Jurafsky, and A. Ng. Cheap and fast - But is it good? Evaluating nonexpert annotations for natural language tasks. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 254–263, 2008. A. Sorokin and D. Forsyth. Utility data annotation with Amazon Mechanical Turk. In Proceedings of the First IEEE Workshop on Internet Vision at CVPR 08, pages 1–8, 2008. J. S. Uebersax and W. M. Grove. A latent trait ﬁnite mixture model for the analysis of rating agreement. Biometrics, 49:823–835, 1993. S. K. Warﬁeld, K. H. Zou, and W. M. Wells. Simultaneous truth and performance level estimation (STAPLE): an algorithm for the validation of image segmentation. IEEE Transactions on Medical Imaging, 23(7):903–921, 2004. J. Whitehill, P. Ruvolo, T. Wu, J. Bergsma, and J. Movellan. Whose vote should count more: Optimal integration of labels from labelers of unknown expertise. In Advances in Neural Information Processing Systems 22, pages 2035–2043. 2009. 1322

6 0.025305398 52 jmlr-2010-Incremental Sigmoid Belief Networks for Grammar Learning

7 0.024004176 45 jmlr-2010-High-dimensional Variable Selection with Sparse Random Projections: Measurement Sparsity and Statistical Efficiency

8 0.0236995 89 jmlr-2010-PAC-Bayesian Analysis of Co-clustering and Beyond

9 0.022028178 91 jmlr-2010-Posterior Regularization for Structured Latent Variable Models

10 0.021858536 90 jmlr-2010-Permutation Tests for Studying Classifier Performance

11 0.019693354 99 jmlr-2010-Restricted Eigenvalue Properties for Correlated Gaussian Designs

12 0.019006427 42 jmlr-2010-Generalized Expectation Criteria for Semi-Supervised Learning with Weakly Labeled Data

13 0.018880343 15 jmlr-2010-Approximate Tree Kernels

14 0.018866677 107 jmlr-2010-Stacked Denoising Autoencoders: Learning Useful Representations in a Deep Network with a Local Denoising Criterion

15 0.018764382 22 jmlr-2010-Classification Using Geometric Level Sets

16 0.01772207 7 jmlr-2010-A Streaming Parallel Decision Tree Algorithm

17 0.017052012 92 jmlr-2010-Practical Approaches to Principal Component Analysis in the Presence of Missing Values

18 0.016961522 12 jmlr-2010-Analysis of Multi-stage Convex Relaxation for Sparse Regularization

19 0.016956795 103 jmlr-2010-Sparse Semi-supervised Learning Using Conjugate Functions

20 0.015155269 41 jmlr-2010-Gaussian Processes for Machine Learning (GPML) Toolbox

similar papers computed by lsi model

lsi for this paper:

topicId topicWeight

[(0, -0.076), (1, 0.003), (2, -0.046), (3, 0.02), (4, 0.004), (5, -0.031), (6, 0.029), (7, -0.033), (8, -0.013), (9, 0.035), (10, 0.017), (11, -0.011), (12, -0.003), (13, 0.081), (14, -0.016), (15, 0.004), (16, -0.016), (17, -0.038), (18, -0.049), (19, -0.052), (20, -0.026), (21, 0.029), (22, -0.05), (23, -0.029), (24, -0.007), (25, 0.035), (26, -0.019), (27, -0.03), (28, 0.015), (29, 0.0), (30, 0.035), (31, 0.094), (32, -0.115), (33, 0.003), (34, -0.121), (35, -0.063), (36, 0.069), (37, 0.254), (38, 0.016), (39, 0.154), (40, -0.425), (41, 0.134), (42, 0.358), (43, 0.094), (44, -0.102), (45, 0.267), (46, 0.093), (47, 0.062), (48, -0.039), (49, -0.08)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.97251832 35 jmlr-2010-Error-Correcting Output Codes Library

Author: Sergio Escalera, Oriol Pujol, Petia Radeva

2 0.30891585 61 jmlr-2010-Learning From Crowds

Author: Vikas C. Raykar, Shipeng Yu, Linda H. Zhao, Gerardo Hermosillo Valadez, Charles Florin, Luca Bogoni, Linda Moy

3 0.25242347 87 jmlr-2010-Online Learning for Matrix Factorization and Sparse Coding

Author: Julien Mairal, Francis Bach, Jean Ponce, Guillermo Sapiro

4 0.21390863 93 jmlr-2010-PyBrain

Author: Tom Schaul, Justin Bayer, Daan Wierstra, Yi Sun, Martin Felder, Frank Sehnke, Thomas Rückstieß, Jürgen Schmidhuber

Abstract: PyBrain is a versatile machine learning library for Python. Its goal is to provide ﬂexible, easyto-use yet still powerful algorithms for machine learning tasks, including a variety of predeﬁned environments and benchmarks to test and compare algorithms. Implemented algorithms include Long Short-Term Memory (LSTM), policy gradient methods, (multidimensional) recurrent neural networks and deep belief networks. Keywords: Python, neural networks, reinforcement learning, optimization

5 0.20259123 110 jmlr-2010-The SHOGUN Machine Learning Toolbox

Author: Sören Sonnenburg, Gunnar Rätsch, Sebastian Henschel, Christian Widmer, Jonas Behr, Alexander Zien, Fabio de Bona, Alexander Binder, Christian Gehl, Vojtěch Franc

Abstract: We have developed a machine learning toolbox, called SHOGUN, which is designed for uniﬁed large-scale learning for a broad range of feature types and learning settings. It offers a considerable number of machine learning models such as support vector machines, hidden Markov models, multiple kernel learning, linear discriminant analysis, and more. Most of the speciﬁc algorithms are able to deal with several different data classes. We have used this toolbox in several applications from computational biology, some of them coming with no less than 50 million training examples and others with 7 billion test examples. With more than a thousand installations worldwide, SHOGUN is already widely adopted in the machine learning community and beyond. SHOGUN is , implemented in C++ and interfaces to MATLABTM R, Octave, Python, and has a stand-alone command line interface. The source code is freely available under the GNU General Public License, Version 3 at http://www.shogun-toolbox.org. Keywords: support vector machines, kernels, large-scale learning, Python, Octave, R

6 0.18780243 32 jmlr-2010-Efficient Algorithms for Conditional Independence Inference

7 0.18434121 39 jmlr-2010-FastInf: An Efficient Approximate Inference Library

8 0.16472635 29 jmlr-2010-Covariance in Unsupervised Learning of Probabilistic Grammars

9 0.16233888 2 jmlr-2010-A Convergent Online Single Time Scale Actor Critic Algorithm

10 0.15941535 90 jmlr-2010-Permutation Tests for Studying Classifier Performance

11 0.1593264 98 jmlr-2010-Regularized Discriminant Analysis, Ridge Regression and Beyond

12 0.13152304 89 jmlr-2010-PAC-Bayesian Analysis of Co-clustering and Beyond

13 0.12234975 22 jmlr-2010-Classification Using Geometric Level Sets

14 0.12067005 94 jmlr-2010-Quadratic Programming Feature Selection

15 0.11802264 8 jmlr-2010-A Surrogate Modeling and Adaptive Sampling Toolbox for Computer Based Design

16 0.1171981 38 jmlr-2010-Expectation Truncation and the Benefits of Preselection In Training Generative Models

17 0.11702673 95 jmlr-2010-Rademacher Complexities and Bounding the Excess Risk in Active Learning

18 0.11403228 88 jmlr-2010-Optimal Search on Clustered Structural Constraint for Learning Bayesian Network Structure

19 0.11309998 10 jmlr-2010-An Exponential Model for Infinite Rankings

20 0.11095291 71 jmlr-2010-Matched Gene Selection and Committee Classifier for Molecular Classification of Heterogeneous Diseases

similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(4, 0.02), (21, 0.014), (32, 0.028), (36, 0.011), (37, 0.053), (75, 0.057), (79, 0.609), (85, 0.041), (96, 0.03)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.80021805 35 jmlr-2010-Error-Correcting Output Codes Library

Author: Sergio Escalera, Oriol Pujol, Petia Radeva

2 0.38434681 40 jmlr-2010-Fast and Scalable Local Kernel Machines

Author: Nicola Segata, Enrico Blanzieri

Abstract: A computationally efﬁcient approach to local learning with kernel methods is presented. The Fast Local Kernel Support Vector Machine (FaLK-SVM) trains a set of local SVMs on redundant neighbourhoods in the training set and an appropriate model for each query point is selected at testing time according to a proximity strategy. Supported by a recent result by Zakai and Ritov (2009) relating consistency and localizability, our approach achieves high classiﬁcation accuracies by dividing the separation function in local optimisation problems that can be handled very efﬁciently from the computational viewpoint. The introduction of a fast local model selection further speeds-up the learning process. Learning and complexity bounds are derived for FaLK-SVM, and the empirical evaluation of the approach (with data sets up to 3 million points) showed that it is much faster and more accurate and scalable than state-of-the-art accurate and approximated SVM solvers at least for non high-dimensional data sets. More generally, we show that locality can be an important factor to sensibly speed-up learning approaches and kernel methods, differently from other recent techniques that tend to dismiss local information in order to improve scalability. Keywords: locality, kernel methods, local learning algorithms, support vector machines, instancebased learning

3 0.15118216 42 jmlr-2010-Generalized Expectation Criteria for Semi-Supervised Learning with Weakly Labeled Data

Author: Gideon S. Mann, Andrew McCallum

Abstract: In this paper, we present an overview of generalized expectation criteria (GE), a simple, robust, scalable method for semi-supervised training using weakly-labeled data. GE ﬁts model parameters by favoring models that match certain expectation constraints, such as marginal label distributions, on the unlabeled data. This paper shows how to apply generalized expectation criteria to two classes of parametric models: maximum entropy models and conditional random ﬁelds. Experimental results demonstrate accuracy improvements over supervised training and a number of other stateof-the-art semi-supervised learning methods for these models. Keywords: generalized expectation criteria, semi-supervised learning, logistic regression, conditional random ﬁelds

4 0.14408335 57 jmlr-2010-Iterative Scaling and Coordinate Descent Methods for Maximum Entropy Models

Author: Fang-Lan Huang, Cho-Jui Hsieh, Kai-Wei Chang, Chih-Jen Lin

Abstract: Maximum entropy (Maxent) is useful in natural language processing and many other areas. Iterative scaling (IS) methods are one of the most popular approaches to solve Maxent. With many variants of IS methods, it is difﬁcult to understand them and see the differences. In this paper, we create a general and uniﬁed framework for iterative scaling methods. This framework also connects iterative scaling and coordinate descent methods. We prove general convergence results for IS methods and analyze their computational complexity. Based on the proposed framework, we extend a coordinate descent method for linear SVM to Maxent. Results show that it is faster than existing iterative scaling methods. Keywords: maximum entropy, iterative scaling, coordinate descent, natural language processing, optimization

5 0.14208084 91 jmlr-2010-Posterior Regularization for Structured Latent Variable Models

Author: Kuzman Ganchev, João Graça, Jennifer Gillenwater, Ben Taskar

Abstract: We present posterior regularization, a probabilistic framework for structured, weakly supervised learning. Our framework efﬁciently incorporates indirect supervision via constraints on posterior distributions of probabilistic models with latent variables. Posterior regularization separates model complexity from the complexity of structural constraints it is desired to satisfy. By directly imposing decomposable regularization on the posterior moments of latent variables during learning, we retain the computational efﬁciency of the unconstrained model while ensuring desired constraints hold in expectation. We present an efﬁcient algorithm for learning with posterior regularization and illustrate its versatility on a diverse set of structural constraints such as bijectivity, symmetry and group sparsity in several large scale experiments, including multi-view learning, cross-lingual dependency grammar induction, unsupervised part-of-speech induction, and bitext word alignment.1 Keywords: posterior regularization framework, unsupervised learning, latent variables models, prior knowledge, natural language processing

6 0.14181703 104 jmlr-2010-Sparse Spectrum Gaussian Process Regression

7 0.14146203 103 jmlr-2010-Sparse Semi-supervised Learning Using Conjugate Functions

8 0.14038789 110 jmlr-2010-The SHOGUN Machine Learning Toolbox

9 0.13833678 114 jmlr-2010-Unsupervised Supervised Learning I: Estimating Classification and Regression Errors without Labels

10 0.13760349 89 jmlr-2010-PAC-Bayesian Analysis of Co-clustering and Beyond

11 0.13742356 74 jmlr-2010-Maximum Relative Margin and Data-Dependent Regularization

12 0.13726278 109 jmlr-2010-Stochastic Composite Likelihood

13 0.13725817 46 jmlr-2010-High Dimensional Inverse Covariance Matrix Estimation via Linear Programming

14 0.13691674 59 jmlr-2010-Large Scale Online Learning of Image Similarity Through Ranking

15 0.1361575 18 jmlr-2010-Bundle Methods for Regularized Risk Minimization

16 0.13582501 102 jmlr-2010-Semi-Supervised Novelty Detection

17 0.13557434 49 jmlr-2010-Hubs in Space: Popular Nearest Neighbors in High-Dimensional Data

18 0.13501757 92 jmlr-2010-Practical Approaches to Principal Component Analysis in the Presence of Missing Values

19 0.13496922 118 jmlr-2010-libDAI: A Free and Open Source C++ Library for Discrete Approximate Inference in Graphical Models

20 0.13456161 62 jmlr-2010-Learning Gradients: Predictive Models that Infer Geometry and Statistical Dependence