This paper presents a specialised Bayesian model for analysing the covariance of data that are observed in the form of matrices, which is particularly suitable for images. Compared to existing general-purpose covariance learning techniques, we exploit the fact that the variables are organised as an array with two sets of ordered indexes, which induces innate relationship between the variables. Specifically, we adopt a factorised structure for the covariance matrix. The covariance of two variables is represented by the product of the covariance of the two corresponding rows and that of the two columns. The factors, i.e. the row-wise and column-wise covariance matrices are estimated by Bayesian inference with sparse priors. Empirical study has been conducted on image analysis. The model first learns correlations between the rows and columns in an image plane. Then the correlations between individual pixels can be inferred by their locations. This scheme utilises the structural information of an image, and benefits the analysis when the data are damaged or insufficient.

Thresholding a measure in conditional independence (CI) tests using a fixed value enables learning and removing edges as part of learning a Bayesian network structure. However, the learned structure is sensitive to the threshold that is commonly selected: 1) arbitrarily; 2) irrespective of characteristics of the domain; and 3) fixed for all CI tests. We analyze the impact on mutual information – a CI measure – of factors, such as sample size, degree of variable dependence, and variables’ cardinalities. Following, we suggest to adaptively threshold individual tests based on the factors. We show that adaptive thresholds better distinguish between pairs of dependent variables and pairs of independent variables and enable learning structures more accurately and quickly than when using fixed thresholds.

We investigate an interactive teaching scenario, where a human teaches a robot symbols which abstract the geometric properties of objects. There are multiple motivations for this scenario: First, state-of-the-art methods for relational reinforcement learning demonstrate that we can learn and employ strongly generalizing abstract models with great success for goal-directed object manipulation. However, these methods rely on given grounded action and state symbols and raise the classical question: Where do the symbols come from? Second, existing research on learning from human-robot interaction has focused mostly on the motion level (e.g., imitation learning). However, if the goal of teaching is to enable the robot to autonomously solve sequential manipulation tasks in a goal-directed manner, the human should have the possibility to teach the relevant abstractions to describe the task and let the robot eventually leverage powerful relational RL methods. In this paper we formalize human-robot teaching of grounded symbols as an active learning problem, where the robot actively generates pick-and-place geometric situations that maximize its information gain about the symbol to be learned. We demonstrate that the learned symbols can be used by a robot in a relational RL framework to learn probabilistic relational rules and use them to solve object manipulation tasks in a goal-directed manner.

In traditional topic models such as LDA, a word is generated by choosing a topic from a collection. However, existing topic models do not identify different types of topics in a document, such as topics that represent the content and topics that represent the sentiment. In this paper, our goal is to discover such different types of topics, if they exist. We represent our model as several parallel topic models (called topic factors), where each word is generated from topics from these factors jointly. Since the latent membership of the word is now a vector, the learning algorithms become challenging. We show that using a variational approximation still allows us to keep the algorithm tractable. Our experiments over several datasets show that our approach consistently outperforms many classic topic models while also discovering fewer, more meaningful, topics.

Hash function learning has been recently received more and more attentions in fast search for large scale data. However, existing popular learning based hashing methods are batch-based learning models and thus incur large scale computational problem for learning an optimal model on a large scale of labelled data and cannot handle data which comes sequentially. In this paper, we address the problem by developing an online hashing learning algorithm to get hashing model accommodate to each new pair of data. At the same time the new updated hash model is penalized by the last learned model in order to retain important information learned in previous rounds. We also derive a tight bound for the cumulative loss of our proposed online learning algorithm. The experimental results demonstrate superiority of the proposed online hashing model on searching both metric distance neighbors and semantical similar neighbors in the experiments.

In many real world scenarios, active learning methods are used to select the most informative points for labeling to reduce the expensive human action. One direction for active learning is selecting the most representative points, ie., selecting the points that other points can be approximated by linear combination of the selected points. However, these methods fails to consider the local geometrical information of the data space. In this paper, we propose a novel framework named Active Learning via Neighborhood Reconstruction (ALNR) by taking into account the locality information directly during the selection. Specifically, for the linear reconstruction of target point, the nearer neighbors should have a greater effect and the selected points distant from the target point should be penalized severely. We further develop an efficient two-stage iterative procedure to solve the final optimizationproblem. Our empirical study shows encouraging results of the proposed algorithms in comparison to other state-of-the-art active learning algorithms on both synthetic and real visual data sets.

Recent advances in Bayesian reinforcement learning (BRL) have shown that Bayes-optimality is theoretically achievable by modeling the environment's latent dynamics using Flat-Dirichlet-Multinomial (FDM) prior. In self-interested multi-agent environments, the transition dynamics are mainly controlled by the other agent's stochastic behavior for which FDM's independence and modeling assumptions do not hold. As a result, FDM does not allow the other agent's behavior to be generalized across different states nor specified using prior domain knowledge. To overcome these practical limitations of FDM, we propose a generalization of BRL to integrate the general class of parametric models and model priors, thus allowing practitioners' domain knowledge to be exploited to produce a fine-grained and compact representation of the other agent's behavior. Empirical evaluation shows that our approach outperforms existing multi-agent reinforcement learning algorithms.

Supervised feature selection determines feature relevance by evaluating feature's correlation with the classes. Joint minimization of a classifier's loss function and an l 2,1 -norm regularization has been shown to be effective for feature selection. However, the appropriate feature subset learned from different classifiers' loss function may be different. Less effort has been made on improving the performance of feature selection by the ensemble of different classifiers' criteria and take advantages of them. Furthermore, for the cases when only a few labeled data per class are available, over-fitting would be a potential problem and the performance of each classifier is restrained. In this paper, we add a joint l 2,1 -norm on multiple feature selection matrices to ensemble different classifiers' loss function into a joint optimization framework. This added co-regularization term has twofold role in enhancing the effect of regularization for each criterion and uncovering common irrelevant features. The problem of over-fitting can be alleviated and thus the performance of feature selection is improved. Extensive experiment on different data types demonstrates the effectiveness of our algorithm.

Multi-label classification is a critical problem in many areas of data analysis such as image labeling and text categorization. In this paper we propose a probabilistic multi-label classification model based on novel sparse feature learning. By employing an individual sparsity inducing ℓ1-norm and a group sparsity inducing ℓ2,1-norm, the proposed model has the capacity of capturing both label interdependencies and common predictive model structures. We formulate this sparse norm regularized learning problem as a non-smooth convex optimization problem, and develop a fast proximal gradient algorithm to solve it for an optimal solution. Our empirical study demonstrates the efficacy of the proposed method on a set of multi-label tasks given a limited number of labeled training instances.

We introduce MiningZinc, a general framework for constraint-based pattern mining, one of the most popular tasks in data mining. MiningZinc consists of two key components: a language component and a toolchain component. The language allows for high-level and natu- ral modeling of mining problems, such that MiningZinc models closely resemble definitions found in the data mining literature. It is inspired by the Zinc family of languages and systems and supports user-defined constraints and optimization criteria. The toolchain allows for finding solutions to the models. It ensures the solver independence of the language and supports both standard constraint solvers and specialized data mining systems. Au- tomatic model transformations enable the efficient use of different solvers and systems. The combination of both components allows one to rapidly model constraint-based mining problems and execute these with a wide variety of methods. We demonstrate this experimentally for a number of well-known solvers and data mining tasks.