nips nips2009 nips2009-229 knowledge-graph by maker-knowledge-mining

229 nips-2009-Statistical Analysis of Semi-Supervised Learning: The Limit of Infinite Unlabelled Data

Source: pdf

Author: Boaz Nadler, Nathan Srebro, Xueyuan Zhou

Abstract: We study the behavior of the popular Laplacian Regularization method for SemiSupervised Learning at the regime of a ﬁxed number of labeled points but a large number of unlabeled points. We show that in Rd , d 2, the method is actually not well-posed, and as the number of unlabeled points increases the solution degenerates to a noninformative function. We also contrast the method with the Laplacian Eigenvector method, and discuss the “smoothness” assumptions associated with this alternate method. 1 Introduction and Setup In this paper we consider the limit behavior of two popular semi-supervised learning (SSL) methods based on the graph Laplacian: the regularization approach [15] and the spectral approach [3]. We consider the limit when the number of labeled points is ﬁxed and the number of unlabeled points goes to inﬁnity. This is a natural limit for SSL as the basic SSL scenario is one in which unlabeled data is virtually inﬁnite. We can also think of this limit as “perfect” SSL, having full knowledge of the marginal density p(x). The premise of SSL is that the marginal density p(x) is informative about the unknown mapping y(x) we are trying to learn, e.g. since y(x) is expected to be “smooth” in some sense relative to p(x). Studying the inﬁnite-unlabeled-data limit, where p(x) is fully known, allows us to formulate and understand the underlying smoothness assumptions of a particular SSL method, and judge whether it is well-posed and sensible. Understanding the inﬁnite-unlabeled-data limit is also a necessary ﬁrst step to studying the convergence of the ﬁnite-labeled-data estimator. We consider the following setup: Let p(x) be an unknown smooth density on a compact domain Ω ⊂ Rd with a smooth boundary. Let y : Ω → Y be the unknown function we wish to estimate. In case of regression Y = R whereas in binary classiﬁcation Y = {−1, 1}. The standard (transductive) semisupervised learning problem is formulated as follows: Given l labeled points, (x1 , y1 ), . . . , (xl , yl ), with yi = y(xi ), and u unlabeled points xl+1 , . . . , xl+u , with all points xi sampled i.i.d. from p(x), the goal is to construct an estimate of y(xl+i ) for any unlabeled point xl+i , utilizing both the labeled and the unlabeled points. We denote the total number of points by n = l + u. We are interested in the regime where l is ﬁxed and u → ∞. 1 2 SSL with Graph Laplacian Regularization We ﬁrst consider the following graph-based approach formulated by Zhu et. al. [15]: y (x) = arg min In (y) ˆ subject to y(xi ) = yi , i = 1, . . . , l y where 1 n2 In (y) = Wi,j (y(xi ) − y(xj ))2 (1) (2) i,j is a Laplacian regularization term enforcing “smoothness” with respect to the n×n similarity matrix W . This formulation has several natural interpretations in terms of, e.g. random walks and electrical circuits [15]. These interpretations, however, refer to a ﬁxed graph, over a ﬁnite set of points with given similarities. In contrast, our focus here is on the more typical scenario where the points xi ∈ Rd are a random sample from a density p(x), and W is constructed based on this sample. We would like to understand the behavior of the method in terms of the density p(x), particularly in the limit where the number of unlabeled points grows. Under what assumptions on the target labeling y(x) and on the density p(x) is the method (1) sensible? The answer, of course, depends on how the matrix W is constructed. We consider the common situation where the similarities are obtained by applying some decay ﬁlter to the distances: xi −xj σ Wi,j = G (3) where G : R+ → R+ is some function with an adequately fast decay. Popular choices are the 2 Gaussian ﬁlter G(z) = e−z /2 or the ǫ-neighborhood graph obtained by the step ﬁlter G(z) = 1z<1 . For simplicity, we focus here on the formulation (1) where the solution is required to satisfy the constraints at the labeled points exactly. In practice, the hard labeling constraints are often replaced with a softer loss-based data term, which is balanced against the smoothness term In (y), e.g. [14, 6]. Our analysis and conclusions apply to such variants as well. Limit of the Laplacian Regularization Term As the number of unlabeled examples grows the regularization term (2) converges to its expectation, where the summation is replaced by integration w.r.t. the density p(x): lim In (y) = I (σ) (y) = n→∞ G Ω Ω x−x′ σ (y(x) − y(x′ ))2 p(x)p(x′ )dxdx′ . (4) In the above limit, the bandwidth σ is held ﬁxed. Typically, one would also drive the bandwidth σ to zero as n → ∞. There are two reasons for this choice. First, from a practical perspective, this makes the similarity matrix W sparse so it can be stored and processed. Second, from a theoretical perspective, this leads to a clear and well deﬁned limit of the smoothness regularization term In (y), at least when σ → 0 slowly enough1 , namely when σ = ω( d log n/n). If σ → 0 as n → ∞, and as long as nσ d / log n → ∞, then after appropriate normalization, the regularizer converges to a density weighted gradient penalty term [7, 8]: d lim d+2 In (y) n→∞ Cσ (σ) d (y) d+2 I σ→0 Cσ = lim ∇y(x) 2 p(x)2 dx = J(y) = (5) Ω where C = Rd z 2 G( z )dz, and assuming 0 < C < ∞ (which is the case for both the Gaussian and the step ﬁlters). This energy functional J(f ) therefore encodes the notion of “smoothness” with respect to p(x) that is the basis of the SSL formulation (1) with the graph constructions speciﬁed by (3). To understand the behavior and appropriateness of (1) we must understand this functional and the associated limit problem: y (x) = arg min J(y) ˆ subject to y(xi ) = yi , i = 1, . . . , l (6) y p When σ = o( d 1/n) then all non-diagonal weights Wi,j vanish (points no longer have any “close by” p neighbors). We are not aware of an analysis covering the regime where σ decays roughly as d 1/n, but would be surprised if a qualitatively different meaningful limit is reached. 1 2 3 Graph Laplacian Regularization in R1 We begin by considering the solution of (6) for one dimensional data, i.e. d = 1 and x ∈ R. We ﬁrst consider the situation where the support of p(x) is a continuous interval Ω = [a, b] ⊂ R (a and/or b may be inﬁnite). Without loss of generality, we assume the labeled data is sorted in increasing order a x1 < x2 < · · · < xl b. Applying the theory of variational calculus, the solution y (x) ˆ satisﬁes inside each interval (xi , xi+1 ) the Euler-Lagrange equation d dy p2 (x) = 0. dx dx Performing two integrations and enforcing the constraints at the labeled points yields y(x) = yi + x 1/p2 (t)dt xi (yi+1 xi+1 1/p2 (t)dt xi − yi ) for xi x xi+1 (7) with y(x) = x1 for a x x1 and y(x) = xl for xl x b. If the support of p(x) is a union of disjoint intervals, the above analysis and the form of the solution applies in each interval separately. The solution (7) seems reasonable and desirable from the point of view of the “smoothness” assumptions: when p(x) is uniform, the solution interpolates linearly between labeled data points, whereas across low-density regions, where p(x) is close to zero, y(x) can change abruptly. Furthermore, the regularizer J(y) can be interpreted as a Reproducing Kernel Hilbert Space (RKHS) squared semi-norm, giving us additional insight into this choice of regularizer: b 1 Theorem 1. Let p(x) be a smooth density on Ω = [a, b] ⊂ R such that Ap = 4 a 1/p2 (t)dt < ∞. 2 Then, J(f ) can be written as a squared semi-norm J(f ) = f Kp induced by the kernel x′ ′ Kp (x, x ) = Ap − 1 2 x with a null-space of all constant functions. That is, f the RKHS induced by Kp . 1 p2 (t) dt Kp . (8) is the norm of the projection of f onto If p(x) is supported on several disjoint intervals, Ω = ∪i [ai , bi ], then J(f ) can be written as a squared semi-norm induced by the kernel 1 bi dt 4 ai p2 (t) ′ Kp (x, x ) = − 1 2 x′ dt x p2 (t) if x, x′ ∈ [ai , bi ] (9) if x ∈ [ai , bi ], x′ ∈ [aj , bj ], i = j 0 with a null-space spanned by indicator functions 1[ai ,bi ] (x) on the connected components of Ω. Proof. For any f (x) = i αi Kp (x, xi ) in the RKHS induced by Kp : df dx J(f ) = 2 p2 (x)dx = αi αj Jij (10) i,j where Jij = d d Kp (x, xi ) Kp (x, xj )p2 (x)dx dx dx When xi and xj are in different connected components of Ω, the gradients of Kp (·, xi ) and Kp (·, xj ) are never non-zero together and Jij = 0 = Kp (xi , xj ). When they are in the same connected component [a, b], and assuming w.l.o.g. a xi xj b: Jij = = xi 1 4 1 4 a b a 1 dt + p2 (t) 1 1 dt − p2 (t) 2 xj xi xj xi −1 dt + p2 (t) xj 1 dt p2 (t) 1 dt = Kp (xi , xj ). p2 (t) Substituting Jij = Kp (xi , xj ) into (10) yields J(f ) = 3 b αi αj Kp (xi , xj ) = f (11) Kp . Combining Theorem 1 with the Representer Theorem [13] establishes that the solution of (6) (or of any variant where the hard constraints are replaced by a data term) is of the form: l y(x) = αj Kp (x, xj ) + βi 1[ai ,bi ] (x), j=1 i where i ranges over the connected components [ai , bi ] of Ω, and we have: l J(y) = αi αj Kp (xi , xj ). (12) i,j=1 Viewing the regularizer as y 2 p suggests understanding (6), and so also its empirical approximaK tion (1), by interpreting Kp (x, x′ ) as a density-based “similarity measure” between x and x′ . This similarity measure indeed seems sensible: for a uniform density it is simply linearly decreasing as a function of the distance. When the density is non-uniform, two points are relatively similar only if they are connected by a region in which 1/p2 (x) is low, i.e. the density is high, but are much less “similar”, i.e. related to each other, when connected by a low-density region. Furthermore, there is no dependence between points in disjoint components separated by zero density regions. 4 Graph Laplacian Regularization in Higher Dimensions The analysis of the previous section seems promising, at it shows that in one dimension, the SSL method (1) is well posed and converges to a sensible limit. Regretfully, in higher dimensions this is not the case anymore. In the following theorem we show that the inﬁmum of the limit problem (6) is zero and can be obtained by a sequence of functions which are certainly not a sensible extrapolation of the labeled points. Theorem 2. Let p(x) be a smooth density over Rd , d 2, bounded from above by some constant pmax , and let (x1 , y1 ), . . . , (xl , yl ) be any (non-repeating) set of labeled examples. There exist continuous functions yǫ (x), for any ǫ > 0, all satisfying the constraints yǫ (xj ) = yj , j = 1, . . . , l, such ǫ→0 ǫ→0 that J(yǫ ) −→ 0 but yǫ (x) −→ 0 for all x = xj , j = 1, . . . , l. Proof. We present a detailed proof for the case of l = 2 labeled points. The generalization of the proof to more labeled points is straightforward. Furthermore, without loss of generality, we assume the ﬁrst labeled point is at x0 = 0 with y(x0 ) = 0 and the second labeled point is at x1 with x1 = 1 and y(x1 ) = 1. In addition, we assume that the ball B1 (0) of radius one centered around the origin is contained in Ω = {x ∈ Rd | p(x) > 0}. We ﬁrst consider the case d > 2. Here, for any ǫ > 0, consider the function x ǫ yǫ (x) = min ,1 which indeed satisﬁes the two constraints yǫ (xi ) = yi , i = 0, 1. Then, J(yǫ ) = Bǫ (0) p2 (x) dx ǫ2 pmax ǫ2 dx = p2 Vd ǫd−2 max (13) Bǫ (0) where Vd is the volume of a unit ball in Rd . Hence, the sequence of functions yǫ (x) satisfy the constraints, but for d > 2, inf ǫ J(yǫ ) = 0. For d = 2, a more extreme example is necessary: consider the functions 2 x yǫ (x) = log +ǫ ǫ log 1+ǫ ǫ for x 1 and yǫ (x) = 1 for x > 1. These functions satisfy the two constraints yǫ (xi ) = yi , i = 0, 1 and: J(yǫ ) = 4 h “ ”i 1+ǫ 2 log ǫ 4πp2 max h “ ”i 1+ǫ 2 log ǫ x B1 (0) log ( x 1+ǫ ǫ 2 2 +ǫ)2 p2 (x)dx 4p2 h “ max ”i2 1+ǫ log ǫ 4πp2 max ǫ→0 = −→ 0. log 1+ǫ ǫ 4 1 0 r2 (r 2 +ǫ)2 2πrdr The implication of Theorem 2 is that regardless of the values at the labeled points, as u → ∞, the solution of (1) is not well posed. Asymptotically, the solution has the form of an almost everywhere constant function, with highly localized spikes near the labeled points, and so no learning is performed. In particular, an interpretation in terms of a density-based kernel Kp , as in the onedimensional case, is not possible. Our analysis also carries over to a formulation where a loss-based data term replaces the hard label constraints, as in l 1 y = arg min ˆ (y(xj ) − yj )2 + γIn (y) y(x) l j=1 In the limit of inﬁnite unlabeled data, functions of the form yǫ (x) above have a zero data penalty term (since they exactly match the labels) and also drive the regularization term J(y) to zero. Hence, it is possible to drive the entire objective functional (the data term plus the regularization term) to zero with functions that do not generalize at all to unlabeled points. 4.1 Numerical Example We illustrate the phenomenon detailed by Theorem 2 with a simple example. Consider a density p(x) in R2 , which is a mixture of two unit variance spherical Gaussians, one per class, centered at the origin and at (4, 0). We sample a total of n = 3000 points, and label two points from each of the two components (four total). We then construct a similarity matrix using a Gaussian ﬁlter with σ = 0.4. Figure 1 depicts the predictor y (x) obtained from (1). In fact, two different predictors are shown, ˆ obtained by different numerical methods for solving (1). Both methods are based on the observation that the solution y (x) of (1) satisﬁes: ˆ n y (xi ) = ˆ n Wij y (xj ) / ˆ j=1 Wij on all unlabeled points i = l + 1, . . . , l + u. (14) j=1 Combined with the constraints of (1), we obtain a system of linear equations that can be solved by Gaussian elimination (here invoked through MATLAB’s backslash operator). This is the method used in the top panels of Figure 1. Alternatively, (14) can be viewed as an update equation for y (xi ), ˆ which can be solved via the power method, or label propagation [2, 6]: start with zero labels on the unlabeled points and iterate (14), while keeping the known labels on x1 , . . . , xl . This is the method used in the bottom panels of Figure 1. As predicted, y (x) is almost constant for almost all unlabeled points. Although all values are very ˆ close to zero, thresholding at the “right” threshold does actually produce sensible results in terms of the true -1/+1 labels. However, beyond being inappropriate for regression, a very ﬂat predictor is still problematic even from a classiﬁcation perspective. First, it is not possible to obtain a meaningful conﬁdence measure for particular labels. Second, especially if the size of each class is not known apriori, setting the threshold between the positive and negative classes is problematic. In our example, setting the threshold to zero yields a generalization error of 45%. The differences between the two numerical methods for solving (1) also point out to another problem with the ill-posedness of the limit problem: the solution is numerically very un-stable. A more quantitative evaluation, that also validates that the effect in Figure 1 is not a result of choosing a “wrong” bandwidth σ, is given in Figure 2. We again simulated data from a mixture of two Gaussians, one Gaussian per class, this time in 20 dimensions, with one labeled point per class, and an increasing number of unlabeled points. In Figure 2 we plot the squared error, and the classiﬁcation error of the resulting predictor y (x). We plot the classiﬁcation error both when a threshold ˆ of zero is used (i.e. the class is determined by sign(ˆ(x))) and with the ideal threshold minimizing y the test error. For each unlabeled sample size, we choose the bandwidth σ yielding the best test performance (this is a “cheating” approach which provides a lower bound on the error of the best method for selecting the bandwidth). As the number of unlabeled examples increases the squared error approaches 1, indicating a ﬂat predictor. Using a threshold of zero leads to an increase in the classiﬁcation error, possibly due to numerical instability. Interestingly, although the predictors become very ﬂat, the classiﬁcation error using the ideal threshold actually improves slightly. Note that 5 DIRECT INVERSION SQUARED ERROR SIGN ERROR: 45% OPTIMAL BANDWIDTH 1 0.9 1 5 0 4 2 0.85 y(x) > 0 y(x) < 0 6 0.95 10 0 0 −1 10 0 200 400 600 800 0−1 ERROR (THRESHOLD=0) 0.32 −5 10 0 5 −10 0 −10 −5 −5 0 5 10 10 1 0 0 200 400 600 800 OPTIMAL BANDWIDTH 0.5 0 0 200 400 600 800 0−1 ERROR (IDEAL THRESHOLD) 0.19 5 200 400 600 800 OPTIMAL BANDWIDTH 1 0.28 SIGN ERR: 17.1 0.3 0.26 POWER METHOD 0 1.5 8 0 0.18 −1 10 6 0.17 4 −5 10 0 5 −10 0 −5 −10 −5 0 5 10 Figure 1: Left plots: Minimizer of Eq. (1). Right plots: the resulting classiﬁcation according to sign(y). The four labeled points are shown by green squares. Top: minimization via Gaussian elimination (MATLAB backslash). Bottom: minimization via label propagation with 1000 iterations - the solution has not yet converged, despite small residuals of the order of 2 · 10−4 . 0.16 0 200 400 600 800 2 0 200 400 600 800 Figure 2: Squared error (top), classiﬁcation error with a threshold of zero (center) and minimal classiﬁcation error using ideal threhold (bottom), of the minimizer of (1) as a function of number of unlabeled points. For each error measure and sample size, the bandwidth minimizing the test error was used, and is plotted. ideal classiﬁcation performance is achieved with a signiﬁcantly larger bandwidth than the bandwidth minimizing the squared loss, i.e. when the predictor is even ﬂatter. 4.2 Probabilistic Interpretation, Exit and Hitting Times As mentioned above, the Laplacian regularization method (1) has a probabilistic interpretation in terms of a random walk on the weighted graph. Let x(t) denote a random walk on the graph with transition matrix M = D−1 W where D is a diagonal matrix with Dii = j Wij . Then, for the binary classiﬁcation case with yi = ±1 we have [15]: y (xi ) = 2 Pr x(t) hits a point labeled +1 before hitting a point labeled -1 x(0) = xi − 1 ˆ We present an interpretation of our analysis in terms of the limiting properties of this random walk. Consider, for simplicity, the case where the two classes are separated by a low density region. Then, the random walk has two intrinsic quantities of interest. The ﬁrst is the mean exit time from one cluster to the other, and the other is the mean hitting time to the labeled points in that cluster. As the number of unlabeled points increases and σ → 0, the random walk converges to a diffusion process [12]. While the mean exit time then converges to a ﬁnite value corresponding to its diffusion analogue, the hitting time to a labeled point increases to inﬁnity (as these become absorbing boundaries of measure zero). With more and more unlabeled data the random walk will fully mix, forgetting where it started, before it hits any label. Thus, the probability of hitting +1 before −1 will become uniform across the entire graph, independent of the starting location xi , yielding a ﬂat predictor. 5 Keeping σ Finite At this point, a reader may ask whether the problems found in higher dimensions are due to taking the limit σ → 0. One possible objection is that there is an intrinsic characteristic scale for the data σ0 where (with high probability) all points at a distance xi − xj < σ0 have the same label. If this is the case, then it may not necessarily make sense to take values of σ < σ0 in constructing W . However, keeping σ ﬁnite while taking the number of unlabeled points to inﬁnity does not resolve the problem. On the contrary, even the one-dimensional case becomes ill-posed in this case. To see this, consider a function y(x) which is zero everywhere except at the labeled points, where y(xj ) = yj . With a ﬁnite number of labeled points of measure zero, I (σ) (y) = 0 in any dimension 6 50 points 500 points 3500 points 1 1 0.5 0.5 0.5 0 0 0 −0.5 y 1 −0.5 −0.5 −1 −2 0 2 4 6 −1 −2 0 2 4 6 −1 −2 0 2 4 6 x Figure 3: Minimizer of (1) for a 1-d problem with a ﬁxed σ = 0.4, two labeled points and an increasing number of unlabeled points. and for any ﬁxed σ > 0. While this limiting function is discontinuous, it is also possible to construct ǫ→0 a sequence of continuous functions yǫ that all satisfy the constraints and for which I (σ) (yǫ ) −→ 0. This behavior is illustrated in Figure 3. We generated data from a mixture of two 1-D Gaussians centered at the origin and at x = 4, with one Gaussian labeled −1 and the other +1. We used two labeled points at the centers of the Gaussians and an increasing number of randomly drawn unlabeled points. As predicted, with a ﬁxed σ, although the solution is reasonable when the number of unlabeled points is small, it becomes ﬂatter, with sharp spikes on the labeled points, as u → ∞. 6 Fourier-Eigenvector Based Methods Before we conclude, we discuss a different approach for SSL, also based on the Graph Laplacian, suggested by Belkin and Niyogi [3]. Instead of using the Laplacian as a regularizer, constraining candidate predictors y(x) non-parametrically to those with small In (y) values, here the predictors are constrained to the low-dimensional space spanned by the ﬁrst few eigenvectors of the Laplacian: The similarity matrix W is computed as before, and the Graph Laplacian matrix L = D − W is considered (recall D is a diagonal matrix with Dii = j Wij ). Only predictors p j=1 aj ej y (x) = ˆ (15) spanned by the ﬁrst p eigenvectors e1 , . . . , ep of L (with smallest eigenvalues) are considered. The coefﬁcients aj are chosen by minimizing a loss function on the labeled data, e.g. the squared loss: (ˆ1 , . . . , ap ) = arg min a ˆ l j=1 (yj − y (xj ))2 . ˆ (16) Unlike the Laplacian Regularization method (1), the Laplacian Eigenvector method (15)–(16) is well posed in the limit u → ∞. This follows directly from the convergence of the eigenvectors of the graph Laplacian to the eigenfunctions of the corresponding Laplace-Beltrami operator [10, 4]. Eigenvector based methods were shown empirically to provide competitive generalization performance on a variety of simulated and real world problems. Belkin and Niyogi [3] motivate the approach by arguing that ‘the eigenfunctions of the Laplace-Beltrami operator provide a natural basis for functions on the manifold and the desired classiﬁcation function can be expressed in such a basis’. In our view, the success of the method is actually not due to data lying on a low-dimensional manifold, but rather due to the low density separation assumption, which states that different class labels form high-density clusters separated by low density regions. Indeed, under this assumption and with sufﬁcient separation between the clusters, the eigenfunctions of the graph Laplace-Beltrami operator are approximately piecewise constant in each of the clusters, as in spectral clustering [12, 11], providing a basis for a labeling that is constant within clusters but variable across clusters. In other settings, such as data uniformly distributed on a manifold but without any signiﬁcant cluster structure, the success of eigenvector based methods critically depends on how well can the unknown classiﬁcation function be approximated by a truncated expansion with relatively few eigenvectors. We illustrate this issue with the following three-dimensional example: Let p(x) denote the uniform density in the box [0, 1] × [0, 0.8] × [0, 0.6], where the box lengths are different to prevent eigenvalue multiplicity. Consider learning three different functions, y1 (x) = 1x1 >0.5 , y2 (x) = 1x1 >x2 /0.8 and y3 (x) = 1x2 /0.8>x3 /0.6 . Even though all three functions are relatively simple, all having a linear separating boundary between the classes on the manifold, as shown in the experiment described in Figure 4, the Eigenvector based method (15)–(16) gives markedly different generalization performances on the three targets. This happens both when the number of eigenvectors p is set to p = l/5 as suggested by Belkin and Niyogi, as well as for the optimal (oracle) value of p selected on the test set (i.e. a “cheating” choice representing an upper bound on the generalization error of this method). 7 Prediction Error (%) p = #labeled points/5 40 optimal p 20 labeled points 40 Approx. Error 50 20 20 0 20 20 40 60 # labeled points 0 10 20 40 60 # labeled points 0 0 5 10 15 # eigenvectors 0 0 5 10 15 # eigenvectors Figure 4: Left three panels: Generalization Performance of the Eigenvector Method (15)–(16) for the three different functions described in the text. All panels use n = 3000 points. Prediction counts the number of sign agreements with the true labels. Rightmost panel: best ﬁt when many (all 3000) points are used, representing the best we can hope for with a few leading eigenvectors. The reason for this behavior is that y2 (x) and even more so y3 (x) cannot be as easily approximated by the very few leading eigenfunctions—even though they seem “simple” and “smooth”, they are signiﬁcantly more complicated than y1 (x) in terms of measure of simplicity implied by the Eigenvector Method. Since the density is uniform, the graph Laplacian converges to the standard Laplacian and its eigenfunctions have the form ψi,j,k (x) = cos(iπx1 ) cos(jπx2 /0.8) cos(kπx3 /0.6), making it hard to represent simple decision boundaries which are not axis-aligned. 7 Discussion Our results show that a popular SSL method, the Laplacian Regularization method (1), is not wellbehaved in the limit of inﬁnite unlabeled data, despite its empirical success in various SSL tasks. The empirical success might be due to two reasons. First, it is possible that with a large enough number of labeled points relative to the number of unlabeled points, the method is well behaved. This regime, where the number of both labeled and unlabeled points grow while l/u is ﬁxed, has recently been analyzed by Wasserman and Lafferty [9]. However, we do not ﬁnd this regime particularly satisfying as we would expect that having more unlabeled data available should improve performance, rather than require more labeled points or make the problem ill-posed. It also places the user in a delicate situation of choosing the “just right” number of unlabeled points without any theoretical guidance. Second, in our experiments we noticed that although the predictor y (x) becomes extremely ﬂat, in ˆ binary tasks, it is still typically possible to ﬁnd a threshold leading to a good classiﬁcation performance. We do not know of any theoretical explanation for such behavior, nor how to characterize it. Obtaining such an explanation would be very interesting, and in a sense crucial to the theoretical foundation of the Laplacian Regularization method. On a very practical level, such a theoretical understanding might allow us to correct the method so as to avoid the numerical instability associated with ﬂat predictors, and perhaps also make it appropriate for regression. The reason that the Laplacian regularizer (1) is ill-posed in the limit is that the ﬁrst order gradient is not a sufﬁcient penalty in high dimensions. This fact is well known in spline theory, where the Sobolev Embedding Theorem [1] indicates one must control at least d+1 derivatives in Rd . In the 2 context of Laplacian regularization, this can be done using the iterated Laplacian: replacing the d+1 graph Laplacian matrix L = D − W , where D is the diagonal degree matrix, with L 2 (matrix to d+1 the 2 power). In the inﬁnite unlabeled data limit, this corresponds to regularizing all order- d+1 2 (mixed) partial derivatives. In the typical case of a low-dimensional manifold in a high dimensional ambient space, the order of iteration should correspond to the intrinsic, rather then ambient, dimensionality, which poses a practical problem of estimating this usually unknown dimensionality. We are not aware of much practical work using the iterated Laplacian, nor a good understanding of its appropriateness for SSL. A different approach leading to a well-posed solution is to include also an ambient regularization term [5]. However, the properties of the solution and in particular its relation to various assumptions about the “smoothness” of y(x) relative to p(x) remain unclear. Acknowledgments The authors would like to thank the anonymous referees for valuable suggestions. The research of BN was supported by the Israel Science Foundation (grant 432/06). 8 References [1] R.A. Adams, Sobolev Spaces, Academic Press (New York), 1975. [2] A. Azran, The rendevous algorithm: multiclass semi-supervised learning with Markov Random Walks, ICML, 2007. [3] M. Belkin, P. Niyogi, Using manifold structure for partially labelled classiﬁcation, NIPS, vol. 15, 2003. [4] M. Belkin and P. Niyogi, Convergence of Laplacian Eigenmaps, NIPS 19, 2007. [5] M. Belkin, P. Niyogi and S. Sindhwani, Manifold Regularization: A Geometric Framework for Learning from Labeled and Unlabeled Examples, JMLR, 7:2399-2434, 2006. [6] Y. Bengio, O. Delalleau, N. Le Roux, label propagation and quadratic criterion, in Semi-Supervised Learning, Chapelle, Scholkopf and Zien, editors, MIT Press, 2006. [7] O. Bosquet, O. Chapelle, M. Hein, Measure Based Regularization, NIPS, vol. 16, 2004. [8] M. Hein, Uniform convergence of adaptive graph-based regularization, COLT, 2006. [9] J. Lafferty, L. Wasserman, Statistical Analysis of Semi-Supervised Regression, NIPS, vol. 20, 2008. [10] U. von Luxburg, M. Belkin and O. Bousquet, Consistency of spectral clustering, Annals of Statistics, vol. 36(2), 2008. [11] M. Meila, J. Shi. A random walks view of spectral segmentation, AI and Statistics, 2001. [12] B. Nadler, S. Lafon, I.G. Kevrekidis, R.R. Coifman, Diffusion maps, spectral clustering and eigenfunctions of Fokker-Planck operators, NIPS, vol. 18, 2006. [13] B. Sch¨ lkopf, A. Smola, Learning with Kernels, MIT Press, 2002. o [14] D. Zhou, O. Bousquet, T. Navin Lal, J. Weston, B. Sch¨ lkopf, Learning with local and global consistency, o NIPS, vol. 16, 2004. [15] X. Zhu, Z. Ghahramani, J. Lafferty, Semi-Supervised Learning using Gaussian ﬁelds and harmonic functions, ICML, 2003. 9

Reference: text

Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 edu Abstract We study the behavior of the popular Laplacian Regularization method for SemiSupervised Learning at the regime of a ﬁxed number of labeled points but a large number of unlabeled points. [sent-9, score-0.76]

2 We show that in Rd , d 2, the method is actually not well-posed, and as the number of unlabeled points increases the solution degenerates to a noninformative function. [sent-10, score-0.476]

3 1 Introduction and Setup In this paper we consider the limit behavior of two popular semi-supervised learning (SSL) methods based on the graph Laplacian: the regularization approach [15] and the spectral approach [3]. [sent-12, score-0.41]

4 We consider the limit when the number of labeled points is ﬁxed and the number of unlabeled points goes to inﬁnity. [sent-13, score-0.933]

5 This is a natural limit for SSL as the basic SSL scenario is one in which unlabeled data is virtually inﬁnite. [sent-14, score-0.428]

6 We can also think of this limit as “perfect” SSL, having full knowledge of the marginal density p(x). [sent-15, score-0.248]

7 We consider the following setup: Let p(x) be an unknown smooth density on a compact domain Ω ⊂ Rd with a smooth boundary. [sent-21, score-0.2]

8 The standard (transductive) semisupervised learning problem is formulated as follows: Given l labeled points, (x1 , y1 ), . [sent-24, score-0.256]

9 , (xl , yl ), with yi = y(xi ), and u unlabeled points xl+1 , . [sent-27, score-0.521]

10 from p(x), the goal is to construct an estimate of y(xl+i ) for any unlabeled point xl+i , utilizing both the labeled and the unlabeled points. [sent-33, score-0.809]

11 We denote the total number of points by n = l + u. [sent-34, score-0.14]

12 , l y where 1 n2 In (y) = Wi,j (y(xi ) − y(xj ))2 (1) (2) i,j is a Laplacian regularization term enforcing “smoothness” with respect to the n×n similarity matrix W . [sent-41, score-0.185]

13 These interpretations, however, refer to a ﬁxed graph, over a ﬁnite set of points with given similarities. [sent-45, score-0.14]

14 In contrast, our focus here is on the more typical scenario where the points xi ∈ Rd are a random sample from a density p(x), and W is constructed based on this sample. [sent-46, score-0.379]

15 We would like to understand the behavior of the method in terms of the density p(x), particularly in the limit where the number of unlabeled points grows. [sent-47, score-0.755]

16 For simplicity, we focus here on the formulation (1) where the solution is required to satisfy the constraints at the labeled points exactly. [sent-52, score-0.462]

17 In practice, the hard labeling constraints are often replaced with a softer loss-based data term, which is balanced against the smoothness term In (y), e. [sent-53, score-0.168]

18 Limit of the Laplacian Regularization Term As the number of unlabeled examples grows the regularization term (2) converges to its expectation, where the summation is replaced by integration w. [sent-57, score-0.478]

19 the density p(x): lim In (y) = I (σ) (y) = n→∞ G Ω Ω x−x′ σ (y(x) − y(x′ ))2 p(x)p(x′ )dxdx′ . [sent-60, score-0.148]

20 Typically, one would also drive the bandwidth σ to zero as n → ∞. [sent-62, score-0.223]

21 Second, from a theoretical perspective, this leads to a clear and well deﬁned limit of the smoothness regularization term In (y), at least when σ → 0 slowly enough1 , namely when σ = ω( d log n/n). [sent-65, score-0.363]

22 To understand the behavior and appropriateness of (1) we must understand this functional and the associated limit problem: y (x) = arg min J(y) ˆ subject to y(xi ) = yi , i = 1, . [sent-68, score-0.349]

23 We are not aware of an analysis covering the regime where σ decays roughly as d 1/n, but would be surprised if a qualitatively different meaningful limit is reached. [sent-72, score-0.2]

24 Without loss of generality, we assume the labeled data is sorted in increasing order a x1 < x2 < · · · < xl b. [sent-77, score-0.402]

25 dx dx Performing two integrations and enforcing the constraints at the labeled points yields y(x) = yi + x 1/p2 (t)dt xi (yi+1 xi+1 1/p2 (t)dt xi − yi ) for xi x xi+1 (7) with y(x) = x1 for a x x1 and y(x) = xl for xl x b. [sent-79, score-1.513]

26 The solution (7) seems reasonable and desirable from the point of view of the “smoothness” assumptions: when p(x) is uniform, the solution interpolates linearly between labeled data points, whereas across low-density regions, where p(x) is close to zero, y(x) can change abruptly. [sent-81, score-0.313]

27 Furthermore, the regularizer J(y) can be interpreted as a Reproducing Kernel Hilbert Space (RKHS) squared semi-norm, giving us additional insight into this choice of regularizer: b 1 Theorem 1. [sent-82, score-0.148]

28 Let p(x) be a smooth density on Ω = [a, b] ⊂ R such that Ap = 4 a 1/p2 (t)dt < ∞. [sent-83, score-0.156]

29 a xi xj b: Jij = = xi 1 4 1 4 a b a 1 dt + p2 (t) 1 1 dt − p2 (t) 2 xj xi xj xi −1 dt + p2 (t) xj 1 dt p2 (t) 1 dt = Kp (xi , xj ). [sent-94, score-2.013]

30 p2 (t) Substituting Jij = Kp (xi , xj ) into (10) yields J(f ) = 3 b αi αj Kp (xi , xj ) = f (11) Kp . [sent-95, score-0.308]

31 This similarity measure indeed seems sensible: for a uniform density it is simply linearly decreasing as a function of the distance. [sent-98, score-0.15]

32 When the density is non-uniform, two points are relatively similar only if they are connected by a region in which 1/p2 (x) is low, i. [sent-99, score-0.298]

33 Furthermore, there is no dependence between points in disjoint components separated by zero density regions. [sent-104, score-0.358]

34 4 Graph Laplacian Regularization in Higher Dimensions The analysis of the previous section seems promising, at it shows that in one dimension, the SSL method (1) is well posed and converges to a sensible limit. [sent-105, score-0.144]

35 In the following theorem we show that the inﬁmum of the limit problem (6) is zero and can be obtained by a sequence of functions which are certainly not a sensible extrapolation of the labeled points. [sent-107, score-0.506]

36 Let p(x) be a smooth density over Rd , d 2, bounded from above by some constant pmax , and let (x1 , y1 ), . [sent-109, score-0.21]

37 , (xl , yl ) be any (non-repeating) set of labeled examples. [sent-112, score-0.26]

38 , l, such ǫ→0 ǫ→0 that J(yǫ ) −→ 0 but yǫ (x) −→ 0 for all x = xj , j = 1, . [sent-116, score-0.154]

39 We present a detailed proof for the case of l = 2 labeled points. [sent-121, score-0.225]

40 The generalization of the proof to more labeled points is straightforward. [sent-122, score-0.398]

41 Furthermore, without loss of generality, we assume the ﬁrst labeled point is at x0 = 0 with y(x0 ) = 0 and the second labeled point is at x1 with x1 = 1 and y(x1 ) = 1. [sent-123, score-0.45]

42 Then, J(yǫ ) = Bǫ (0) p2 (x) dx ǫ2 pmax ǫ2 dx = p2 Vd ǫd−2 max (13) Bǫ (0) where Vd is the volume of a unit ball in Rd . [sent-127, score-0.306]

43 log 1+ǫ ǫ 4 1 0 r2 (r 2 +ǫ)2 2πrdr The implication of Theorem 2 is that regardless of the values at the labeled points, as u → ∞, the solution of (1) is not well posed. [sent-131, score-0.269]

44 Asymptotically, the solution has the form of an almost everywhere constant function, with highly localized spikes near the labeled points, and so no learning is performed. [sent-132, score-0.269]

45 Hence, it is possible to drive the entire objective functional (the data term plus the regularization term) to zero with functions that do not generalize at all to unlabeled points. [sent-135, score-0.555]

46 Consider a density p(x) in R2 , which is a mixture of two unit variance spherical Gaussians, one per class, centered at the origin and at (4, 0). [sent-138, score-0.148]

47 We sample a total of n = 3000 points, and label two points from each of the two components (four total). [sent-139, score-0.14]

48 Both methods are based on the observation that the solution y (x) of (1) satisﬁes: ˆ n y (xi ) = ˆ n Wij y (xj ) / ˆ j=1 Wij on all unlabeled points i = l + 1, . [sent-144, score-0.476]

49 (14) j=1 Combined with the constraints of (1), we obtain a system of linear equations that can be solved by Gaussian elimination (here invoked through MATLAB’s backslash operator). [sent-148, score-0.146]

50 Alternatively, (14) can be viewed as an update equation for y (xi ), ˆ which can be solved via the power method, or label propagation [2, 6]: start with zero labels on the unlabeled points and iterate (14), while keeping the known labels on x1 , . [sent-150, score-0.51]

51 As predicted, y (x) is almost constant for almost all unlabeled points. [sent-155, score-0.292]

52 Although all values are very ˆ close to zero, thresholding at the “right” threshold does actually produce sensible results in terms of the true -1/+1 labels. [sent-156, score-0.15]

53 In our example, setting the threshold to zero yields a generalization error of 45%. [sent-160, score-0.201]

54 The differences between the two numerical methods for solving (1) also point out to another problem with the ill-posedness of the limit problem: the solution is numerically very un-stable. [sent-161, score-0.217]

55 We again simulated data from a mixture of two Gaussians, one Gaussian per class, this time in 20 dimensions, with one labeled point per class, and an increasing number of unlabeled points. [sent-163, score-0.517]

56 In Figure 2 we plot the squared error, and the classiﬁcation error of the resulting predictor y (x). [sent-164, score-0.19]

57 We plot the classiﬁcation error both when a threshold ˆ of zero is used (i. [sent-165, score-0.168]

58 For each unlabeled sample size, we choose the bandwidth σ yielding the best test performance (this is a “cheating” approach which provides a lower bound on the error of the best method for selecting the bandwidth). [sent-168, score-0.48]

59 As the number of unlabeled examples increases the squared error approaches 1, indicating a ﬂat predictor. [sent-169, score-0.415]

60 Using a threshold of zero leads to an increase in the classiﬁcation error, possibly due to numerical instability. [sent-170, score-0.155]

61 Interestingly, although the predictors become very ﬂat, the classiﬁcation error using the ideal threshold actually improves slightly. [sent-171, score-0.255]

62 The four labeled points are shown by green squares. [sent-188, score-0.365]

63 16 0 200 400 600 800 2 0 200 400 600 800 Figure 2: Squared error (top), classiﬁcation error with a threshold of zero (center) and minimal classiﬁcation error using ideal threhold (bottom), of the minimizer of (1) as a function of number of unlabeled points. [sent-192, score-0.639]

64 For each error measure and sample size, the bandwidth minimizing the test error was used, and is plotted. [sent-193, score-0.238]

65 ideal classiﬁcation performance is achieved with a signiﬁcantly larger bandwidth than the bandwidth minimizing the squared loss, i. [sent-194, score-0.395]

66 2 Probabilistic Interpretation, Exit and Hitting Times As mentioned above, the Laplacian regularization method (1) has a probabilistic interpretation in terms of a random walk on the weighted graph. [sent-198, score-0.197]

67 Let x(t) denote a random walk on the graph with transition matrix M = D−1 W where D is a diagonal matrix with Dii = j Wij . [sent-199, score-0.169]

68 Then, for the binary classiﬁcation case with yi = ±1 we have [15]: y (xi ) = 2 Pr x(t) hits a point labeled +1 before hitting a point labeled -1 x(0) = xi − 1 ˆ We present an interpretation of our analysis in terms of the limiting properties of this random walk. [sent-200, score-0.788]

69 Consider, for simplicity, the case where the two classes are separated by a low density region. [sent-201, score-0.145]

70 The ﬁrst is the mean exit time from one cluster to the other, and the other is the mean hitting time to the labeled points in that cluster. [sent-203, score-0.539]

71 As the number of unlabeled points increases and σ → 0, the random walk converges to a diffusion process [12]. [sent-204, score-0.601]

72 While the mean exit time then converges to a ﬁnite value corresponding to its diffusion analogue, the hitting time to a labeled point increases to inﬁnity (as these become absorbing boundaries of measure zero). [sent-205, score-0.483]

73 With more and more unlabeled data the random walk will fully mix, forgetting where it started, before it hits any label. [sent-206, score-0.425]

74 Thus, the probability of hitting +1 before −1 will become uniform across the entire graph, independent of the starting location xi , yielding a ﬂat predictor. [sent-207, score-0.236]

75 One possible objection is that there is an intrinsic characteristic scale for the data σ0 where (with high probability) all points at a distance xi − xj < σ0 have the same label. [sent-209, score-0.455]

76 However, keeping σ ﬁnite while taking the number of unlabeled points to inﬁnity does not resolve the problem. [sent-211, score-0.469]

77 To see this, consider a function y(x) which is zero everywhere except at the labeled points, where y(xj ) = yj . [sent-213, score-0.313]

78 With a ﬁnite number of labeled points of measure zero, I (σ) (y) = 0 in any dimension 6 50 points 500 points 3500 points 1 1 0. [sent-214, score-0.785]

79 4, two labeled points and an increasing number of unlabeled points. [sent-221, score-0.657]

80 We generated data from a mixture of two 1-D Gaussians centered at the origin and at x = 4, with one Gaussian labeled −1 and the other +1. [sent-225, score-0.261]

81 We used two labeled points at the centers of the Gaussians and an increasing number of randomly drawn unlabeled points. [sent-226, score-0.657]

82 As predicted, with a ﬁxed σ, although the solution is reasonable when the number of unlabeled points is small, it becomes ﬂatter, with sharp spikes on the labeled points, as u → ∞. [sent-227, score-0.701]

83 Only predictors p j=1 aj ej y (x) = ˆ (15) spanned by the ﬁrst p eigenvectors e1 , . [sent-230, score-0.242]

84 The coefﬁcients aj are chosen by minimizing a loss function on the labeled data, e. [sent-234, score-0.266]

85 ˆ (16) Unlike the Laplacian Regularization method (1), the Laplacian Eigenvector method (15)–(16) is well posed in the limit u → ∞. [sent-240, score-0.168]

86 This follows directly from the convergence of the eigenvectors of the graph Laplacian to the eigenfunctions of the corresponding Laplace-Beltrami operator [10, 4]. [sent-241, score-0.297]

87 Belkin and Niyogi [3] motivate the approach by arguing that ‘the eigenfunctions of the Laplace-Beltrami operator provide a natural basis for functions on the manifold and the desired classiﬁcation function can be expressed in such a basis’. [sent-243, score-0.237]

88 In our view, the success of the method is actually not due to data lying on a low-dimensional manifold, but rather due to the low density separation assumption, which states that different class labels form high-density clusters separated by low density regions. [sent-244, score-0.323]

89 7 Prediction Error (%) p = #labeled points/5 40 optimal p 20 labeled points 40 Approx. [sent-259, score-0.365]

90 Since the density is uniform, the graph Laplacian converges to the standard Laplacian and its eigenfunctions have the form ψi,j,k (x) = cos(iπx1 ) cos(jπx2 /0. [sent-265, score-0.337]

91 7 Discussion Our results show that a popular SSL method, the Laplacian Regularization method (1), is not wellbehaved in the limit of inﬁnite unlabeled data, despite its empirical success in various SSL tasks. [sent-268, score-0.461]

92 First, it is possible that with a large enough number of labeled points relative to the number of unlabeled points, the method is well behaved. [sent-270, score-0.657]

93 This regime, where the number of both labeled and unlabeled points grow while l/u is ﬁxed, has recently been analyzed by Wasserman and Lafferty [9]. [sent-271, score-0.657]

94 However, we do not ﬁnd this regime particularly satisfying as we would expect that having more unlabeled data available should improve performance, rather than require more labeled points or make the problem ill-posed. [sent-272, score-0.721]

95 It also places the user in a delicate situation of choosing the “just right” number of unlabeled points without any theoretical guidance. [sent-273, score-0.432]

96 Second, in our experiments we noticed that although the predictor y (x) becomes extremely ﬂat, in ˆ binary tasks, it is still typically possible to ﬁnd a threshold leading to a good classiﬁcation performance. [sent-274, score-0.144]

97 The reason that the Laplacian regularizer (1) is ill-posed in the limit is that the ﬁrst order gradient is not a sufﬁcient penalty in high dimensions. [sent-278, score-0.211]

98 In the inﬁnite unlabeled data limit, this corresponds to regularizing all order- d+1 2 (mixed) partial derivatives. [sent-281, score-0.292]

99 A different approach leading to a well-posed solution is to include also an ambient regularization term [5]. [sent-284, score-0.244]

100 Coifman, Diffusion maps, spectral clustering and eigenfunctions of Fokker-Planck operators, NIPS, vol. [sent-334, score-0.141]

similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('kp', 0.408), ('unlabeled', 0.292), ('laplacian', 0.285), ('ssl', 0.232), ('labeled', 0.225), ('xl', 0.177), ('xj', 0.154), ('dt', 0.147), ('points', 0.14), ('bandwidth', 0.138), ('limit', 0.136), ('xi', 0.127), ('dx', 0.126), ('density', 0.112), ('regularization', 0.112), ('hitting', 0.109), ('belkin', 0.103), ('eigenfunctions', 0.102), ('niyogi', 0.099), ('eigenvector', 0.095), ('jij', 0.092), ('walk', 0.085), ('graph', 0.084), ('predictors', 0.082), ('smoothness', 0.08), ('threshold', 0.077), ('eigenvectors', 0.077), ('regularizer', 0.075), ('sensible', 0.073), ('squared', 0.073), ('manifold', 0.07), ('ai', 0.067), ('predictor', 0.067), ('exit', 0.065), ('regime', 0.064), ('panels', 0.064), ('rd', 0.06), ('bi', 0.058), ('chicago', 0.058), ('backslash', 0.054), ('pmax', 0.054), ('yi', 0.054), ('ambient', 0.053), ('walks', 0.053), ('constraints', 0.053), ('wij', 0.052), ('error', 0.05), ('ap', 0.048), ('cheating', 0.048), ('appropriateness', 0.048), ('nadler', 0.048), ('hits', 0.048), ('yj', 0.047), ('ideal', 0.046), ('connected', 0.046), ('rkhs', 0.045), ('diffusion', 0.045), ('sign', 0.045), ('smooth', 0.044), ('drive', 0.044), ('solution', 0.044), ('gaussians', 0.043), ('classi', 0.042), ('spanned', 0.042), ('aj', 0.041), ('hein', 0.041), ('zero', 0.041), ('converges', 0.039), ('spectral', 0.039), ('behavior', 0.039), ('israel', 0.039), ('elimination', 0.039), ('sobolev', 0.039), ('similarity', 0.038), ('cos', 0.038), ('keeping', 0.037), ('lafferty', 0.037), ('wasserman', 0.037), ('vd', 0.037), ('dii', 0.037), ('numerical', 0.037), ('origin', 0.036), ('lim', 0.036), ('lter', 0.036), ('understand', 0.036), ('yl', 0.035), ('term', 0.035), ('intrinsic', 0.034), ('interpretations', 0.034), ('iterated', 0.034), ('operator', 0.034), ('success', 0.033), ('minimizer', 0.033), ('separated', 0.033), ('clusters', 0.033), ('generalization', 0.033), ('disjoint', 0.032), ('posed', 0.032), ('functions', 0.031), ('semisupervised', 0.031)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 1.0000011 229 nips-2009-Statistical Analysis of Semi-Supervised Learning: The Limit of Infinite Unlabelled Data

Author: Boaz Nadler, Nathan Srebro, Xueyuan Zhou

2 0.277821 213 nips-2009-Semi-supervised Learning using Sparse Eigenfunction Bases

Author: Kaushik Sinha, Mikhail Belkin

Abstract: We present a new framework for semi-supervised learning with sparse eigenfunction bases of kernel matrices. It turns out that when the data has clustered, that is, when the high density regions are sufﬁciently separated by low density valleys, each high density area corresponds to a unique representative eigenvector. Linear combination of such eigenvectors (or, more precisely, of their Nystrom extensions) provide good candidates for good classiﬁcation functions when the cluster assumption holds. By ﬁrst choosing an appropriate basis of these eigenvectors from unlabeled data and then using labeled data with Lasso to select a classiﬁer in the span of these eigenvectors, we obtain a classiﬁer, which has a very sparse representation in this basis. Importantly, the sparsity corresponds naturally to the cluster assumption. Experimental results on a number of real-world data-sets show that our method is competitive with the state of the art semi-supervised learning algorithms and outperforms the natural base-line algorithm (Lasso in the Kernel PCA basis). 1

3 0.27609092 212 nips-2009-Semi-Supervised Learning in Gigantic Image Collections

Author: Rob Fergus, Yair Weiss, Antonio Torralba

Abstract: With the advent of the Internet it is now possible to collect hundreds of millions of images. These images come with varying degrees of label information. “Clean labels” can be manually obtained on a small fraction, “noisy labels” may be extracted automatically from surrounding text, while for most images there are no labels at all. Semi-supervised learning is a principled framework for combining these different label sources. However, it scales polynomially with the number of images, making it impractical for use on gigantic collections with hundreds of millions of images and thousands of classes. In this paper we show how to utilize recent results in machine learning to obtain highly efﬁcient approximations for semi-supervised learning that are linear in the number of images. Speciﬁcally, we use the convergence of the eigenvectors of the normalized graph Laplacian to eigenfunctions of weighted Laplace-Beltrami operators. Our algorithm enables us to apply semi-supervised learning to a database of 80 million images gathered from the Internet. 1

4 0.23671515 82 nips-2009-Entropic Graph Regularization in Non-Parametric Semi-Supervised Classification

Author: Amarnag Subramanya, Jeff A. Bilmes

Abstract: We prove certain theoretical properties of a graph-regularized transductive learning objective that is based on minimizing a Kullback-Leibler divergence based loss. These include showing that the iterative alternating minimization procedure used to minimize the objective converges to the correct solution and deriving a test for convergence. We also propose a graph node ordering algorithm that is cache cognizant and leads to a linear speedup in parallel computations. This ensures that the algorithm scales to large data sets. By making use of empirical evaluation on the TIMIT and Switchboard I corpora, we show this approach is able to outperform other state-of-the-art SSL approaches. In one instance, we solve a problem on a 120 million node graph. 1

5 0.2330019 214 nips-2009-Semi-supervised Regression using Hessian energy with an application to semi-supervised dimensionality reduction

Author: Kwang I. Kim, Florian Steinke, Matthias Hein

Abstract: Semi-supervised regression based on the graph Laplacian suffers from the fact that the solution is biased towards a constant and the lack of extrapolating power. Based on these observations, we propose to use the second-order Hessian energy for semi-supervised regression which overcomes both these problems. If the data lies on or close to a low-dimensional submanifold in feature space, the Hessian energy prefers functions whose values vary linearly with respect to geodesic distance. We ﬁrst derive the Hessian energy for smooth manifolds and continue to give a stable estimation procedure for the common case where only samples of the underlying manifold are given. The preference of ‘’linear” functions on manifolds renders the Hessian energy particularly suited for the task of semi-supervised dimensionality reduction, where the goal is to ﬁnd a user-deﬁned embedding function given some labeled points which varies smoothly (and ideally linearly) along the manifold. The experimental results suggest superior performance of our method compared with semi-supervised regression using Laplacian regularization or standard supervised regression techniques applied to this task. 1

6 0.18942456 128 nips-2009-Learning Non-Linear Combinations of Kernels

7 0.16786052 26 nips-2009-Adaptive Regularization for Transductive Support Vector Machine

8 0.15101838 120 nips-2009-Kernels and learning curves for Gaussian process regression on random graphs

9 0.15028398 122 nips-2009-Label Selection on Graphs

10 0.13397507 15 nips-2009-A Rate Distortion Approach for Semi-Supervised Conditional Random Fields

11 0.13334791 38 nips-2009-Augmenting Feature-driven fMRI Analyses: Semi-supervised learning and resting state activity

12 0.11450823 146 nips-2009-Manifold Regularization for SIR with Rate Root-n Convergence

13 0.10534361 169 nips-2009-Nonlinear Learning using Local Coordinate Coding

14 0.098861314 142 nips-2009-Locality-sensitive binary codes from shift-invariant kernels

15 0.095177829 137 nips-2009-Learning transport operators for image manifolds

16 0.090215161 240 nips-2009-Sufficient Conditions for Agnostic Active Learnable

17 0.085843481 35 nips-2009-Approximating MAP by Compensating for Structural Relaxations

18 0.083777629 20 nips-2009-A unified framework for high-dimensional analysis of $M$-estimators with decomposable regularizers

19 0.082309611 102 nips-2009-Graph-based Consensus Maximization among Multiple Supervised and Unsupervised Models

20 0.081919044 118 nips-2009-Kernel Choice and Classifiability for RKHS Embeddings of Probability Distributions

similar papers computed by lsi model

lsi for this paper:

topicId topicWeight

[(0, -0.278), (1, 0.142), (2, -0.1), (3, 0.21), (4, -0.173), (5, -0.019), (6, -0.012), (7, 0.101), (8, -0.126), (9, 0.131), (10, -0.185), (11, 0.251), (12, -0.03), (13, -0.294), (14, -0.087), (15, -0.032), (16, -0.122), (17, 0.045), (18, -0.052), (19, 0.059), (20, -0.061), (21, -0.002), (22, -0.002), (23, -0.063), (24, 0.023), (25, -0.032), (26, 0.004), (27, -0.007), (28, -0.032), (29, 0.007), (30, -0.035), (31, 0.02), (32, 0.018), (33, -0.066), (34, -0.065), (35, -0.007), (36, -0.047), (37, -0.012), (38, 0.042), (39, 0.019), (40, 0.086), (41, 0.049), (42, 0.027), (43, 0.007), (44, 0.004), (45, 0.005), (46, -0.025), (47, 0.027), (48, -0.054), (49, -0.021)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.9652884 229 nips-2009-Statistical Analysis of Semi-Supervised Learning: The Limit of Infinite Unlabelled Data

Author: Boaz Nadler, Nathan Srebro, Xueyuan Zhou

2 0.85031569 26 nips-2009-Adaptive Regularization for Transductive Support Vector Machine

Author: Zenglin Xu, Rong Jin, Jianke Zhu, Irwin King, Michael Lyu, Zhirong Yang

Abstract: We discuss the framework of Transductive Support Vector Machine (TSVM) from the perspective of the regularization strength induced by the unlabeled data. In this framework, SVM and TSVM can be regarded as a learning machine without regularization and one with full regularization from the unlabeled data, respectively. Therefore, to supplement this framework of the regularization strength, it is necessary to introduce data-dependant partial regularization. To this end, we reformulate TSVM into a form with controllable regularization strength, which includes SVM and TSVM as special cases. Furthermore, we introduce a method of adaptive regularization that is data dependant and is based on the smoothness assumption. Experiments on a set of benchmark data sets indicate the promising results of the proposed work compared with state-of-the-art TSVM algorithms. 1

3 0.81842715 213 nips-2009-Semi-supervised Learning using Sparse Eigenfunction Bases

Author: Kaushik Sinha, Mikhail Belkin

4 0.77091259 212 nips-2009-Semi-Supervised Learning in Gigantic Image Collections

Author: Rob Fergus, Yair Weiss, Antonio Torralba

5 0.73054242 214 nips-2009-Semi-supervised Regression using Hessian energy with an application to semi-supervised dimensionality reduction

Author: Kwang I. Kim, Florian Steinke, Matthias Hein

6 0.67074734 82 nips-2009-Entropic Graph Regularization in Non-Parametric Semi-Supervised Classification

7 0.63426811 146 nips-2009-Manifold Regularization for SIR with Rate Root-n Convergence

8 0.54951215 15 nips-2009-A Rate Distortion Approach for Semi-Supervised Conditional Random Fields

9 0.50439006 38 nips-2009-Augmenting Feature-driven fMRI Analyses: Semi-supervised learning and resting state activity

10 0.47497192 122 nips-2009-Label Selection on Graphs

11 0.4700937 103 nips-2009-Graph Zeta Function in the Bethe Free Energy and Loopy Belief Propagation

12 0.44903171 160 nips-2009-Multiple Incremental Decremental Learning of Support Vector Machines

13 0.43843493 120 nips-2009-Kernels and learning curves for Gaussian process regression on random graphs

14 0.4297252 106 nips-2009-Heavy-Tailed Symmetric Stochastic Neighbor Embedding

15 0.42691144 130 nips-2009-Learning from Multiple Partially Observed Views - an Application to Multilingual Text Categorization

16 0.41440955 102 nips-2009-Graph-based Consensus Maximization among Multiple Supervised and Unsupervised Models

17 0.4087159 169 nips-2009-Nonlinear Learning using Local Coordinate Coding

18 0.40569243 34 nips-2009-Anomaly Detection with Score functions based on Nearest Neighbor Graphs

19 0.40242368 91 nips-2009-Fast, smooth and adaptive regression in metric spaces

20 0.39270425 72 nips-2009-Distribution Matching for Transduction

similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(7, 0.022), (24, 0.095), (25, 0.125), (35, 0.034), (36, 0.119), (39, 0.045), (50, 0.159), (58, 0.085), (61, 0.019), (71, 0.038), (81, 0.019), (86, 0.122), (91, 0.049)]

similar papers list:

simIndex simValue paperId paperTitle

1 0.90852624 182 nips-2009-Optimal Scoring for Unsupervised Learning

Author: Zhihua Zhang, Guang Dai

Abstract: We are often interested in casting classiﬁcation and clustering problems as a regression framework, because it is feasible to achieve some statistical properties in this framework by imposing some penalty criteria. In this paper we illustrate optimal scoring, which was originally proposed for performing the Fisher linear discriminant analysis by regression, in the application of unsupervised learning. In particular, we devise a novel clustering algorithm that we call optimal discriminant clustering. We associate our algorithm with the existing unsupervised learning algorithms such as spectral clustering, discriminative clustering and sparse principal component analysis. Experimental results on a collection of benchmark datasets validate the effectiveness of the optimal discriminant clustering algorithm.

same-paper 2 0.90478569 229 nips-2009-Statistical Analysis of Semi-Supervised Learning: The Limit of Infinite Unlabelled Data

Author: Boaz Nadler, Nathan Srebro, Xueyuan Zhou

3 0.87959731 228 nips-2009-Speeding up Magnetic Resonance Image Acquisition by Bayesian Multi-Slice Adaptive Compressed Sensing

Author: Matthias Seeger

Abstract: We show how to sequentially optimize magnetic resonance imaging measurement designs over stacks of neighbouring image slices, by performing convex variational inference on a large scale non-Gaussian linear dynamical system, tracking dominating directions of posterior covariance without imposing any factorization constraints. Our approach can be scaled up to high-resolution images by reductions to numerical mathematics primitives and parallelization on several levels. In a ﬁrst study, designs are found that improve signiﬁcantly on others chosen independently for each slice or drawn at random. 1

4 0.82705653 169 nips-2009-Nonlinear Learning using Local Coordinate Coding

Author: Kai Yu, Tong Zhang, Yihong Gong

Abstract: This paper introduces a new method for semi-supervised learning on high dimensional nonlinear manifolds, which includes a phase of unsupervised basis learning and a phase of supervised function learning. The learned bases provide a set of anchor points to form a local coordinate system, such that each data point x on the manifold can be locally approximated by a linear combination of its nearby anchor points, and the linear weights become its local coordinate coding. We show that a high dimensional nonlinear function can be approximated by a global linear function with respect to this coding scheme, and the approximation quality is ensured by the locality of such coding. The method turns a difﬁcult nonlinear learning problem into a simple global linear learning problem, which overcomes some drawbacks of traditional local learning methods. 1

5 0.81851912 213 nips-2009-Semi-supervised Learning using Sparse Eigenfunction Bases

Author: Kaushik Sinha, Mikhail Belkin

6 0.81689298 122 nips-2009-Label Selection on Graphs

7 0.8125537 212 nips-2009-Semi-Supervised Learning in Gigantic Image Collections

8 0.80649447 37 nips-2009-Asymptotically Optimal Regularization in Smooth Parametric Models

9 0.80541623 214 nips-2009-Semi-supervised Regression using Hessian energy with an application to semi-supervised dimensionality reduction

10 0.80265909 17 nips-2009-A Sparse Non-Parametric Approach for Single Channel Separation of Known Sounds

11 0.80204201 166 nips-2009-Noisy Generalized Binary Search

12 0.80042827 112 nips-2009-Human Rademacher Complexity

13 0.79965055 104 nips-2009-Group Sparse Coding

14 0.79960012 225 nips-2009-Sparsistent Learning of Varying-coefficient Models with Structural Changes

15 0.79918528 26 nips-2009-Adaptive Regularization for Transductive Support Vector Machine

16 0.7990433 71 nips-2009-Distribution-Calibrated Hierarchical Classification

17 0.7988472 142 nips-2009-Locality-sensitive binary codes from shift-invariant kernels

18 0.7962746 118 nips-2009-Kernel Choice and Classifiability for RKHS Embeddings of Probability Distributions

19 0.79593045 137 nips-2009-Learning transport operators for image manifolds

20 0.79589653 139 nips-2009-Linear-time Algorithms for Pairwise Statistical Problems