nips nips2012 nips2012-31 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Hankz H. Zhuo, Qiang Yang, Subbarao Kambhampati
Abstract: Multi-Agent Plan Recognition (MAPR) aims to recognize dynamic team structures and team behaviors from the observed team traces (activity sequences) of a set of intelligent agents. Previous MAPR approaches required a library of team activity sequences (team plans) be given as input. However, collecting a library of team plans to ensure adequate coverage is often difficult and costly. In this paper, we relax this constraint, so that team plans are not required to be provided beforehand. We assume instead that a set of action models are available. Such models are often already created to describe domain physics; i.e., the preconditions and effects of effects actions. We propose a novel approach for recognizing multi-agent team plans based on such action models rather than libraries of team plans. We encode the resulting MAPR problem as a satisfiability problem and solve the problem using a state-of-the-art weighted MAX-SAT solver. Our approach also allows for incompleteness in the observed plan traces. Our empirical studies demonstrate that our algorithm is both effective and efficient in comparison to state-of-the-art MAPR methods based on plan libraries. 1
Reference: text
sentIndex sentText sentNum sentScore
1 edu Abstract Multi-Agent Plan Recognition (MAPR) aims to recognize dynamic team structures and team behaviors from the observed team traces (activity sequences) of a set of intelligent agents. [sent-7, score-2.395]
2 Previous MAPR approaches required a library of team activity sequences (team plans) be given as input. [sent-8, score-0.917]
3 However, collecting a library of team plans to ensure adequate coverage is often difficult and costly. [sent-9, score-1.324]
4 In this paper, we relax this constraint, so that team plans are not required to be provided beforehand. [sent-10, score-1.137]
5 We assume instead that a set of action models are available. [sent-11, score-0.121]
6 Such models are often already created to describe domain physics; i. [sent-12, score-0.019]
7 We propose a novel approach for recognizing multi-agent team plans based on such action models rather than libraries of team plans. [sent-15, score-2.047]
8 We encode the resulting MAPR problem as a satisfiability problem and solve the problem using a state-of-the-art weighted MAX-SAT solver. [sent-16, score-0.025]
9 Our approach also allows for incompleteness in the observed plan traces. [sent-17, score-0.243]
10 Our empirical studies demonstrate that our algorithm is both effective and efficient in comparison to state-of-the-art MAPR methods based on plan libraries. [sent-18, score-0.214]
11 1 Introduction Multi-Agent Plan Recognition (MAPR) seeks an explanation of observed team-action traces. [sent-19, score-0.029]
12 From the activity sequences of a set of agents, MAPR aims to identify the dynamic team structures and team behaviors of agents. [sent-20, score-1.572]
13 The MAPR problem has important applications in analyzing data from automated monitoring, situation awareness, intelligence surveillance and analysis [4]. [sent-21, score-0.022]
14 Many approaches have been proposed in the past to automatically recognize team plans given an observed team trace as input. [sent-22, score-2.046]
15 They solved MAPR problems using a first-cut approach, provided that a fully observed team trace and a library of full team plans were given as input. [sent-25, score-2.004]
16 To relax the full observability constraint, Zhuo and Li [19] proposed a MARS system to recognize team plans based on partially observed team traces and libraries of partial team plans. [sent-26, score-2.901]
17 1 Despite the success of these previous approaches, they all assume that a library of team plans has been collected beforehand and provided as input. [sent-27, score-1.256]
18 However, there are many applications where collecting and maintaining a library of team plans is difficult and costly. [sent-28, score-1.267]
19 For example, in military operations, it is difficult and expensive to collect team plans, since activities of team-mates may consume lots of resources such as ammunition and human labor. [sent-29, score-0.856]
20 Collecting a smaller library is not an option since it is infeasible to recognize team plans if they are not covered by the library. [sent-30, score-1.377]
21 It is thus useful to design approaches for solving the MAPR problem where we do not require libraries of team plans to be known. [sent-31, score-1.207]
22 In this paper, we advocate replacing the plan library with a compact action model of the domain. [sent-32, score-0.46]
23 In contrast to plan libraries, action models are easier to specify (in terms of preconditions and effects of each type of activity). [sent-33, score-0.469]
24 Moreover, in principle action models provide full coverage to recognize any team plans. [sent-34, score-0.983]
25 The specific algorithmic framework we develop is called DARE, which stands for Domain- model based multi-Agent REcognition, to recognize multi-agent plans. [sent-35, score-0.137]
26 DARE does not require plan libraries to be given as input. [sent-36, score-0.306]
27 Instead, DARE takes as input a team trace and a set of action models. [sent-37, score-0.871]
28 DARE also allows the observed traces to be incomplete, i. [sent-38, score-0.113]
29 , there can be To fill these gaps, DARE leverages all possible constraints both from the plan traces and from its knowledge of how a plan works in terms of its causal structure. [sent-40, score-0.538]
30 To do this, DARE first builds a set of hard constraints that encode the correctness property of the team plans, and a set of soft constraints that encode the optimal utility property of team plans based on the input team trace and action models. [sent-41, score-2.778]
31 After that, it solves all these constraints using a state-of-the-art weighted MAX-SAT solver, such as MaxSatz [10], and converts the solution to a set of team plans as output. [sent-42, score-1.159]
32 In the next section, we first introduce the related work including single agent plan recognition and multi-agent plan recognition, and then give our formulation of the MAPR problem. [sent-44, score-0.534]
33 2 Related work The plan recognition problem has been addressed by many researchers. [sent-47, score-0.281]
34 Kautz and Allen proposed an approach to recognize plans based on parsing observed actions as sequences of subactions and essentially model this knowledge as a context-free rule in an “action grammar” [9]. [sent-48, score-0.62]
35 presented approaches to probabilistic plan recognition problems [5, 7]. [sent-50, score-0.263]
36 Instead of using a library of plans, Ramrez and Geffner [12] proposed an approach to solving the plan recognition problem using slightly modified planning algorithms, assuming the action models were given as input. [sent-51, score-0.494]
37 Note that action models can be created by experts or learnt by previous systems, such as ARMS [18] and LAMP [20]. [sent-52, score-0.14]
38 Singla and Mooney proposed an approach to abductive reasoning using a first-order probabilistic logic to recognize plans [15]. [sent-53, score-0.577]
39 Amir and Gal addressed a plan recognition approach to recognizing student behaviors using virtual science laboratories [1]. [sent-54, score-0.369]
40 Ramirez and Geffner exploited off-the-shelf classical planners to recognize probabilistic plans [13]. [sent-55, score-0.583]
41 Despite the success of these systems, a limitation is that they all focus only on single agent plans. [sent-56, score-0.072]
42 For multi-agent plan recognition, Sukthankar and Sycara presented an approach that leveraged several types of agent resource dependencies and temporal ordering constraints in the plan library to prune the size of the plan library considered for each observation trace [16]. [sent-57, score-1.053]
43 Avrahami-Zilberbrand and Kaminka preferred a library of single agent plans to team plans, but identified dynamic teams based on the assumption that all agents in a team executing the same plan under the temporal constraints of that plan [2]. [sent-58, score-2.565]
44 The constraint on activities of the agents that can form a team can be severely limiting when team-mates can execute coordinated but different behaviors. [sent-59, score-0.977]
45 proposed a probabilistic model based on conditional random fields to automatically recognize the composition of teams and team activities in relation to a plan [11]. [sent-61, score-1.178]
46 In these systems, although coordinated activities can be recognized, they either assume there is a set of real-world GPS data available, or assume that team traces and team plans can be fully observed. [sent-62, score-2.043]
47 In this paper, we allow that: (1) agents can execute coordinated different activities in a team, (2) team traces can be partial, and (3) neither GPS data nor team plans are needed. [sent-63, score-2.176]
48 2 3 Problem Definition We first define a team trace. [sent-64, score-0.69]
49 , n } be a set of agents, and O = [otj ] be an observed team trace. [sent-68, score-0.719]
50 Let otj be the observed activity executed by agent j at time step t, where 0 < t T and 0 < j n. [sent-69, score-0.25]
51 A team trace O is partial, if some elements in O are empty (denoted by null), i. [sent-70, score-0.771]
52 In the STRIPS language [6], an action model is a tuple ha, Pre(a), Add(a), Del(a)i, where a is an action name with zero or more parameters, Pre(a) is a list of preconditions of a, Add(a) is a list of add effects, and Del(a) is a list of deleting effects. [sent-74, score-0.485]
53 An action name with zero of more parameters is called an activity. [sent-76, score-0.141]
54 An observed activity otj in a partial team trace O is either an instantiated action of A or noop or null, where noop is an empty activity that does nothing. [sent-77, score-1.299]
55 An initial state s0 is a set of propositions that describes a closed world state from which the team trace O starts to be observed. [sent-78, score-0.826]
56 In other words, activities at time step t = 0 can be applied in the initial state s0 . [sent-79, score-0.102]
57 When we say an activity can be applied in a state, we mean the activity’s preconditions are satisfied by the state. [sent-80, score-0.19]
58 A set of goals G, each of which is a set of propositions, describes the probable targets of the team trace. [sent-81, score-0.706]
59 We assume s0 and G can both be observed by sensing devices. [sent-82, score-0.029]
60 A team is composed of a subset of agents 0 = { j1 , j2 , . [sent-83, score-0.779]
wordName wordTfidf (topN-words)
[('team', 0.69), ('plans', 0.425), ('mapr', 0.279), ('plan', 0.214), ('dare', 0.203), ('recognize', 0.137), ('action', 0.121), ('library', 0.11), ('preconditions', 0.102), ('libraries', 0.092), ('agents', 0.089), ('activity', 0.088), ('activities', 0.087), ('traces', 0.084), ('otj', 0.076), ('coordinated', 0.067), ('trace', 0.06), ('agent', 0.057), ('geffner', 0.051), ('noop', 0.051), ('recognition', 0.049), ('kautz', 0.045), ('zhuo', 0.045), ('execute', 0.044), ('behaviors', 0.042), ('collecting', 0.042), ('del', 0.035), ('teams', 0.035), ('arizona', 0.035), ('coverage', 0.035), ('effects', 0.032), ('gps', 0.032), ('kong', 0.031), ('propositions', 0.03), ('hong', 0.029), ('pre', 0.029), ('sequences', 0.029), ('observed', 0.029), ('recognizing', 0.029), ('list', 0.028), ('constraints', 0.026), ('encode', 0.025), ('partial', 0.024), ('null', 0.024), ('atk', 0.022), ('adequate', 0.022), ('ammunition', 0.022), ('bui', 0.022), ('guangzhou', 0.022), ('lamp', 0.022), ('masato', 0.022), ('sadilek', 0.022), ('singla', 0.022), ('strips', 0.022), ('sukthankar', 0.022), ('surveillance', 0.022), ('relax', 0.022), ('empty', 0.021), ('ark', 0.021), ('consume', 0.021), ('gal', 0.021), ('planners', 0.021), ('name', 0.02), ('deleting', 0.019), ('jm', 0.019), ('tempe', 0.019), ('created', 0.019), ('allen', 0.018), ('noah', 0.018), ('observability', 0.018), ('aims', 0.018), ('add', 0.018), ('military', 0.018), ('awareness', 0.018), ('converts', 0.018), ('lots', 0.018), ('addressed', 0.018), ('laboratories', 0.017), ('amir', 0.017), ('leveraged', 0.017), ('qiang', 0.016), ('gaps', 0.016), ('organize', 0.016), ('beforehand', 0.016), ('grammar', 0.016), ('monitoring', 0.016), ('prune', 0.016), ('describes', 0.016), ('roles', 0.016), ('dynamic', 0.015), ('arms', 0.015), ('logic', 0.015), ('automatically', 0.015), ('success', 0.015), ('advocate', 0.015), ('covered', 0.015), ('resource', 0.015), ('state', 0.015), ('recognized', 0.015), ('ha', 0.015)]
simIndex simValue paperId paperTitle
same-paper 1 1.0000004 31 nips-2012-Action-Model Based Multi-agent Plan Recognition
Author: Hankz H. Zhuo, Qiang Yang, Subbarao Kambhampati
Abstract: Multi-Agent Plan Recognition (MAPR) aims to recognize dynamic team structures and team behaviors from the observed team traces (activity sequences) of a set of intelligent agents. Previous MAPR approaches required a library of team activity sequences (team plans) be given as input. However, collecting a library of team plans to ensure adequate coverage is often difficult and costly. In this paper, we relax this constraint, so that team plans are not required to be provided beforehand. We assume instead that a set of action models are available. Such models are often already created to describe domain physics; i.e., the preconditions and effects of effects actions. We propose a novel approach for recognizing multi-agent team plans based on such action models rather than libraries of team plans. We encode the resulting MAPR problem as a satisfiability problem and solve the problem using a state-of-the-art weighted MAX-SAT solver. Our approach also allows for incompleteness in the observed plan traces. Our empirical studies demonstrate that our algorithm is both effective and efficient in comparison to state-of-the-art MAPR methods based on plan libraries. 1
2 0.061499603 141 nips-2012-GenDeR: A Generic Diversified Ranking Algorithm
Author: Jingrui He, Hanghang Tong, Qiaozhu Mei, Boleslaw Szymanski
Abstract: Diversified ranking is a fundamental task in machine learning. It is broadly applicable in many real world problems, e.g., information retrieval, team assembling, product search, etc. In this paper, we consider a generic setting where we aim to diversify the top-k ranking list based on an arbitrary relevance function and an arbitrary similarity function among all the examples. We formulate it as an optimization problem and show that in general it is NP-hard. Then, we show that for a large volume of the parameter space, the proposed objective function enjoys the diminishing returns property, which enables us to design a scalable, greedy algorithm to find the (1 − 1/e) near-optimal solution. Experimental results on real data sets demonstrate the effectiveness of the proposed algorithm.
3 0.059623092 353 nips-2012-Transferring Expectations in Model-based Reinforcement Learning
Author: Trung Nguyen, Tomi Silander, Tze Y. Leong
Abstract: We study how to automatically select and adapt multiple abstractions or representations of the world to support model-based reinforcement learning. We address the challenges of transfer learning in heterogeneous environments with varying tasks. We present an efficient, online framework that, through a sequence of tasks, learns a set of relevant representations to be used in future tasks. Without predefined mapping strategies, we introduce a general approach to support transfer learning across different state spaces. We demonstrate the potential impact of our system through improved jumpstart and faster convergence to near optimum policy in two benchmark domains. 1
4 0.052967887 209 nips-2012-Max-Margin Structured Output Regression for Spatio-Temporal Action Localization
Author: Du Tran, Junsong Yuan
Abstract: Structured output learning has been successfully applied to object localization, where the mapping between an image and an object bounding box can be well captured. Its extension to action localization in videos, however, is much more challenging, because we need to predict the locations of the action patterns both spatially and temporally, i.e., identifying a sequence of bounding boxes that track the action in video. The problem becomes intractable due to the exponentially large size of the structured video space where actions could occur. We propose a novel structured learning approach for spatio-temporal action localization. The mapping between a video and a spatio-temporal action trajectory is learned. The intractable inference and learning problems are addressed by leveraging an efficient Max-Path search method, thus making it feasible to optimize the model over the whole structured space. Experiments on two challenging benchmark datasets show that our proposed method outperforms the state-of-the-art methods. 1
5 0.050230514 289 nips-2012-Recognizing Activities by Attribute Dynamics
Author: Weixin Li, Nuno Vasconcelos
Abstract: In this work, we consider the problem of modeling the dynamic structure of human activities in the attributes space. A video sequence is Ä?Ĺš rst represented in a semantic feature space, where each feature encodes the probability of occurrence of an activity attribute at a given time. A generative model, denoted the binary dynamic system (BDS), is proposed to learn both the distribution and dynamics of different activities in this space. The BDS is a non-linear dynamic system, which extends both the binary principal component analysis (PCA) and classical linear dynamic systems (LDS), by combining binary observation variables with a hidden Gauss-Markov state process. In this way, it integrates the representation power of semantic modeling with the ability of dynamic systems to capture the temporal structure of time-varying processes. An algorithm for learning BDS parameters, inspired by a popular LDS learning method from dynamic textures, is proposed. A similarity measure between BDSs, which generalizes the BinetCauchy kernel for LDS, is then introduced and used to design activity classiÄ?Ĺš ers. The proposed method is shown to outperform similar classiÄ?Ĺš ers derived from the kernel dynamic system (KDS) and state-of-the-art approaches for dynamics-based or attribute-based action recognition. 1
6 0.049182065 344 nips-2012-Timely Object Recognition
7 0.037332103 122 nips-2012-Exploration in Model-based Reinforcement Learning by Empirically Estimating Learning Progress
8 0.036542505 88 nips-2012-Cost-Sensitive Exploration in Bayesian Reinforcement Learning
9 0.031356052 208 nips-2012-Matrix reconstruction with the local max norm
10 0.030508783 173 nips-2012-Learned Prioritization for Trading Off Accuracy and Speed
11 0.029510545 3 nips-2012-A Bayesian Approach for Policy Learning from Trajectory Preference Queries
12 0.027918825 83 nips-2012-Controlled Recognition Bounds for Visual Learning and Exploration
13 0.026994804 160 nips-2012-Imitation Learning by Coaching
14 0.02688797 288 nips-2012-Rational inference of relative preferences
15 0.024622317 108 nips-2012-Efficient Bayes-Adaptive Reinforcement Learning using Sample-Based Search
16 0.024459507 238 nips-2012-Neurally Plausible Reinforcement Learning of Working Memory Tasks
17 0.024312923 183 nips-2012-Learning Partially Observable Models Using Temporally Abstract Decision Trees
18 0.023404753 316 nips-2012-Small-Variance Asymptotics for Exponential Family Dirichlet Process Mixture Models
19 0.023365853 77 nips-2012-Complex Inference in Neural Circuits with Probabilistic Population Codes and Topic Models
20 0.020738576 245 nips-2012-Nonparametric Bayesian Inverse Reinforcement Learning for Multiple Reward Functions
topicId topicWeight
[(0, 0.05), (1, -0.052), (2, -0.026), (3, 0.003), (4, -0.004), (5, -0.003), (6, 0.005), (7, -0.003), (8, 0.002), (9, 0.023), (10, -0.01), (11, 0.011), (12, 0.021), (13, -0.019), (14, 0.027), (15, 0.017), (16, 0.002), (17, 0.034), (18, 0.025), (19, -0.031), (20, 0.023), (21, -0.008), (22, -0.072), (23, 0.001), (24, 0.018), (25, -0.015), (26, 0.009), (27, 0.011), (28, 0.003), (29, 0.045), (30, 0.04), (31, 0.008), (32, 0.017), (33, 0.015), (34, -0.023), (35, 0.01), (36, 0.003), (37, -0.053), (38, 0.0), (39, -0.01), (40, 0.019), (41, -0.031), (42, 0.018), (43, 0.046), (44, -0.056), (45, -0.004), (46, 0.049), (47, 0.017), (48, -0.021), (49, 0.054)]
simIndex simValue paperId paperTitle
same-paper 1 0.95299858 31 nips-2012-Action-Model Based Multi-agent Plan Recognition
Author: Hankz H. Zhuo, Qiang Yang, Subbarao Kambhampati
Abstract: Multi-Agent Plan Recognition (MAPR) aims to recognize dynamic team structures and team behaviors from the observed team traces (activity sequences) of a set of intelligent agents. Previous MAPR approaches required a library of team activity sequences (team plans) be given as input. However, collecting a library of team plans to ensure adequate coverage is often difficult and costly. In this paper, we relax this constraint, so that team plans are not required to be provided beforehand. We assume instead that a set of action models are available. Such models are often already created to describe domain physics; i.e., the preconditions and effects of effects actions. We propose a novel approach for recognizing multi-agent team plans based on such action models rather than libraries of team plans. We encode the resulting MAPR problem as a satisfiability problem and solve the problem using a state-of-the-art weighted MAX-SAT solver. Our approach also allows for incompleteness in the observed plan traces. Our empirical studies demonstrate that our algorithm is both effective and efficient in comparison to state-of-the-art MAPR methods based on plan libraries. 1
2 0.59340167 353 nips-2012-Transferring Expectations in Model-based Reinforcement Learning
Author: Trung Nguyen, Tomi Silander, Tze Y. Leong
Abstract: We study how to automatically select and adapt multiple abstractions or representations of the world to support model-based reinforcement learning. We address the challenges of transfer learning in heterogeneous environments with varying tasks. We present an efficient, online framework that, through a sequence of tasks, learns a set of relevant representations to be used in future tasks. Without predefined mapping strategies, we introduce a general approach to support transfer learning across different state spaces. We demonstrate the potential impact of our system through improved jumpstart and faster convergence to near optimum policy in two benchmark domains. 1
3 0.54142731 209 nips-2012-Max-Margin Structured Output Regression for Spatio-Temporal Action Localization
Author: Du Tran, Junsong Yuan
Abstract: Structured output learning has been successfully applied to object localization, where the mapping between an image and an object bounding box can be well captured. Its extension to action localization in videos, however, is much more challenging, because we need to predict the locations of the action patterns both spatially and temporally, i.e., identifying a sequence of bounding boxes that track the action in video. The problem becomes intractable due to the exponentially large size of the structured video space where actions could occur. We propose a novel structured learning approach for spatio-temporal action localization. The mapping between a video and a spatio-temporal action trajectory is learned. The intractable inference and learning problems are addressed by leveraging an efficient Max-Path search method, thus making it feasible to optimize the model over the whole structured space. Experiments on two challenging benchmark datasets show that our proposed method outperforms the state-of-the-art methods. 1
4 0.4858681 289 nips-2012-Recognizing Activities by Attribute Dynamics
Author: Weixin Li, Nuno Vasconcelos
Abstract: In this work, we consider the problem of modeling the dynamic structure of human activities in the attributes space. A video sequence is Ä?Ĺš rst represented in a semantic feature space, where each feature encodes the probability of occurrence of an activity attribute at a given time. A generative model, denoted the binary dynamic system (BDS), is proposed to learn both the distribution and dynamics of different activities in this space. The BDS is a non-linear dynamic system, which extends both the binary principal component analysis (PCA) and classical linear dynamic systems (LDS), by combining binary observation variables with a hidden Gauss-Markov state process. In this way, it integrates the representation power of semantic modeling with the ability of dynamic systems to capture the temporal structure of time-varying processes. An algorithm for learning BDS parameters, inspired by a popular LDS learning method from dynamic textures, is proposed. A similarity measure between BDSs, which generalizes the BinetCauchy kernel for LDS, is then introduced and used to design activity classiÄ?Ĺš ers. The proposed method is shown to outperform similar classiÄ?Ĺš ers derived from the kernel dynamic system (KDS) and state-of-the-art approaches for dynamics-based or attribute-based action recognition. 1
5 0.48565701 88 nips-2012-Cost-Sensitive Exploration in Bayesian Reinforcement Learning
Author: Dongho Kim, Kee-eung Kim, Pascal Poupart
Abstract: In this paper, we consider Bayesian reinforcement learning (BRL) where actions incur costs in addition to rewards, and thus exploration has to be constrained in terms of the expected total cost while learning to maximize the expected longterm total reward. In order to formalize cost-sensitive exploration, we use the constrained Markov decision process (CMDP) as the model of the environment, in which we can naturally encode exploration requirements using the cost function. We extend BEETLE, a model-based BRL method, for learning in the environment with cost constraints. We demonstrate the cost-sensitive exploration behaviour in a number of simulated problems. 1
6 0.46042839 51 nips-2012-Bayesian Hierarchical Reinforcement Learning
7 0.44415653 183 nips-2012-Learning Partially Observable Models Using Temporally Abstract Decision Trees
8 0.44113195 198 nips-2012-Learning with Target Prior
9 0.42617288 344 nips-2012-Timely Object Recognition
10 0.41201735 153 nips-2012-How Prior Probability Influences Decision Making: A Unifying Probabilistic Model
11 0.40182659 273 nips-2012-Predicting Action Content On-Line and in Real Time before Action Onset – an Intracranial Human Study
12 0.39038229 350 nips-2012-Trajectory-Based Short-Sighted Probabilistic Planning
13 0.39034736 22 nips-2012-A latent factor model for highly multi-relational data
14 0.37600523 45 nips-2012-Approximating Equilibria in Sequential Auctions with Incomplete Information and Multi-Unit Demand
15 0.37596914 108 nips-2012-Efficient Bayes-Adaptive Reinforcement Learning using Sample-Based Search
16 0.37379241 122 nips-2012-Exploration in Model-based Reinforcement Learning by Empirically Estimating Learning Progress
17 0.37218249 311 nips-2012-Shifting Weights: Adapting Object Detectors from Image to Video
18 0.36838475 50 nips-2012-Bandit Algorithms boost Brain Computer Interfaces for motor-task selection of a brain-controlled button
19 0.3650701 238 nips-2012-Neurally Plausible Reinforcement Learning of Working Memory Tasks
20 0.35717848 173 nips-2012-Learned Prioritization for Trading Off Accuracy and Speed
topicId topicWeight
[(0, 0.026), (15, 0.416), (21, 0.025), (36, 0.022), (38, 0.057), (39, 0.012), (42, 0.048), (54, 0.048), (55, 0.015), (74, 0.042), (76, 0.098), (80, 0.051), (92, 0.018)]
simIndex simValue paperId paperTitle
1 0.97913337 308 nips-2012-Semi-Supervised Domain Adaptation with Non-Parametric Copulas
Author: David Lopez-paz, Jose M. Hernández-lobato, Bernhard Schölkopf
Abstract: A new framework based on the theory of copulas is proposed to address semisupervised domain adaptation problems. The presented method factorizes any multivariate density into a product of marginal distributions and bivariate copula functions. Therefore, changes in each of these factors can be detected and corrected to adapt a density model accross different learning domains. Importantly, we introduce a novel vine copula model, which allows for this factorization in a non-parametric manner. Experimental results on regression problems with real-world data illustrate the efficacy of the proposed approach when compared to state-of-the-art techniques. 1
same-paper 2 0.72182292 31 nips-2012-Action-Model Based Multi-agent Plan Recognition
Author: Hankz H. Zhuo, Qiang Yang, Subbarao Kambhampati
Abstract: Multi-Agent Plan Recognition (MAPR) aims to recognize dynamic team structures and team behaviors from the observed team traces (activity sequences) of a set of intelligent agents. Previous MAPR approaches required a library of team activity sequences (team plans) be given as input. However, collecting a library of team plans to ensure adequate coverage is often difficult and costly. In this paper, we relax this constraint, so that team plans are not required to be provided beforehand. We assume instead that a set of action models are available. Such models are often already created to describe domain physics; i.e., the preconditions and effects of effects actions. We propose a novel approach for recognizing multi-agent team plans based on such action models rather than libraries of team plans. We encode the resulting MAPR problem as a satisfiability problem and solve the problem using a state-of-the-art weighted MAX-SAT solver. Our approach also allows for incompleteness in the observed plan traces. Our empirical studies demonstrate that our algorithm is both effective and efficient in comparison to state-of-the-art MAPR methods based on plan libraries. 1
3 0.53111851 4 nips-2012-A Better Way to Pretrain Deep Boltzmann Machines
Author: Geoffrey E. Hinton, Ruslan Salakhutdinov
Abstract: We describe how the pretraining algorithm for Deep Boltzmann Machines (DBMs) is related to the pretraining algorithm for Deep Belief Networks and we show that under certain conditions, the pretraining procedure improves the variational lower bound of a two-hidden-layer DBM. Based on this analysis, we develop a different method of pretraining DBMs that distributes the modelling work more evenly over the hidden layers. Our results on the MNIST and NORB datasets demonstrate that the new pretraining algorithm allows us to learn better generative models. 1
4 0.43356571 149 nips-2012-Hierarchical Optimistic Region Selection driven by Curiosity
Author: Odalric-ambrym Maillard
Abstract: This paper aims to take a step forwards making the term “intrinsic motivation” from reinforcement learning theoretically well founded, focusing on curiositydriven learning. To that end, we consider the setting where, a fixed partition P of a continuous space X being given, and a process ν defined on X being unknown, we are asked to sequentially decide which cell of the partition to select as well as where to sample ν in that cell, in order to minimize a loss function that is inspired from previous work on curiosity-driven learning. The loss on each cell consists of one term measuring a simple worst case quadratic sampling error, and a penalty term proportional to the range of the variance in that cell. The corresponding problem formulation extends the setting known as active learning for multi-armed bandits to the case when each arm is a continuous region, and we show how an adaptation of recent algorithms for that problem and of hierarchical optimistic sampling algorithms for optimization can be used in order to solve this problem. The resulting procedure, called Hierarchical Optimistic Region SElection driven by Curiosity (HORSE.C) is provided together with a finite-time regret analysis. 1
5 0.33687004 229 nips-2012-Multimodal Learning with Deep Boltzmann Machines
Author: Nitish Srivastava, Ruslan Salakhutdinov
Abstract: A Deep Boltzmann Machine is described for learning a generative model of data that consists of multiple and diverse input modalities. The model can be used to extract a unified representation that fuses modalities together. We find that this representation is useful for classification and information retrieval tasks. The model works by learning a probability density over the space of multimodal inputs. It uses states of latent variables as representations of the input. The model can extract this representation even when some modalities are absent by sampling from the conditional distribution over them and filling them in. Our experimental results on bi-modal data consisting of images and text show that the Multimodal DBM can learn a good generative model of the joint space of image and text inputs that is useful for information retrieval from both unimodal and multimodal queries. We further demonstrate that this model significantly outperforms SVMs and LDA on discriminative tasks. Finally, we compare our model to other deep learning methods, including autoencoders and deep belief networks, and show that it achieves noticeable gains. 1
6 0.32983392 287 nips-2012-Random function priors for exchangeable arrays with applications to graphs and relational data
7 0.32696733 310 nips-2012-Semiparametric Principal Component Analysis
8 0.32674265 344 nips-2012-Timely Object Recognition
9 0.32487142 209 nips-2012-Max-Margin Structured Output Regression for Spatio-Temporal Action Localization
10 0.32432199 115 nips-2012-Efficient high dimensional maximum entropy modeling via symmetric partition functions
11 0.323524 289 nips-2012-Recognizing Activities by Attribute Dynamics
12 0.32104641 177 nips-2012-Learning Invariant Representations of Molecules for Atomization Energy Prediction
13 0.32101458 162 nips-2012-Inverse Reinforcement Learning through Structured Classification
14 0.32077569 38 nips-2012-Algorithms for Learning Markov Field Policies
15 0.3186411 245 nips-2012-Nonparametric Bayesian Inverse Reinforcement Learning for Multiple Reward Functions
16 0.31853974 348 nips-2012-Tractable Objectives for Robust Policy Optimization
17 0.3181704 242 nips-2012-Non-linear Metric Learning
18 0.31791312 219 nips-2012-Modelling Reciprocating Relationships with Hawkes Processes
19 0.31780037 188 nips-2012-Learning from Distributions via Support Measure Machines
20 0.31770912 173 nips-2012-Learned Prioritization for Trading Off Accuracy and Speed