Given several tasks, multi-task learning (MTL) learns multiple tasks jointly by exploring the interdependence between them. The basic assumption in MTL is that those tasks are indeed related. Existing MTL methods model the task relatedness/interdependence in two different ways, either common parameter-sharing or common feature-sharing across tasks. In this paper, we propose a novel multi-task learning method to jointly learn shared parameters and shared feature representation. Our objective is to learn a set of common features with which the tasks are related as closely as possible, therefore common parameters shared across tasks can be optimally learned. We present a detailed deviation of our multi-task learning method and propose an alternating algorithm to solve the non-convex optimization problem. We further present a theoretical bound which directly demonstrates that the proposed multi-task learning method can successfully model the relatedness via joint common parameter- and common feature-learning. Extensive experiments are conducted on several real world multi-task learning datasets. All results demonstrate the effectiveness of our multi-task model and feature joint learning method.

This paper revisits the problem of learning a k-CNF Boolean function from examples, for fixed k, in the context of online learning under the logarithmic loss. We give a Bayesian interpretation to one of Valiant’s classic PAC learning algorithms, which we then build upon to derive three efficient, online, probabilistic, supervised learning algorithms for predicting the output of an unknown k-CNF Boolean function. We analyze the loss of our methods, and show that the cumulative log-loss can be upper bounded by a polynomial function of the size of each example.

Recommender systems have become essential tools in many application areas as they help alleviate information overload by tailoring their recommendations to users' personal preferences. Users' interests in items, however, may change over time depending on their current situation. Without considering the current circumstances of a user, recommendations may match the general preferences of the user, but they may have small utility for the user in his/her current situation.We focus on designing systems that interact with the user over a number of iterations and at each step receive feedback from the user in the form of a reward or utility value for the recommended items. The goal of the system is to maximize the sum of obtained utilities over each interaction session. We use a multi-armed bandit strategy to model this online learning problem and we propose techniques for detecting changes in user preferences. The recommendations are then generated based on the most recent preferences of a user. Our evaluation results indicate that our method can improve the existing bandit algorithms by considering the sudden variations in the user's feedback behavior.

Cross-validation (CV) is one of the main tools for performance estimation and parameter tuning in machine learning. The general recipe for computing CV estimate is to run a learning algorithm separately for each CV fold, a computationally expensive process. In this paper, we propose a new approach to reduce the computational burden of CV-based performance estimation. As opposed to all previous attempts, which are specific to a particular learning model or problem domain, we propose a general method applicable to a large class of incremental learning algorithms, which are uniquely fitted to big data problems. In particular, our method applies to a wide range of supervised and unsupervised learning tasks with different performance criteria, as long as the base learning algorithm is incremental. We show that the running time of the algorithm scales logarithmically, rather than linearly, in the number of CV folds. Furthermore, the algorithm has favorable properties for parallel and distributed implementation. Experiments with state-of-the-art incremental learning algorithms confirm the practicality of the proposed method.

Research shows that for novices learning from worked examples is superior to unsupported problem solving. Additionally, several studies have shown that learning from examples results in faster learning in comparison to supported problem solving in Intelligent Tutoring Systems. In a previous study, we have shown that alternating worked examples and problem solving was superior to using just one type of learning tasks. In this paper we present a study that compares learning from a fixed sequence of alternating worked examples and tutored problem solving to a strategy that adaptively decides how much assistance to provide to the student. The adaptive strategy determines the type of task (a worked example, a faded example or a problem to solve) based on how much assistance the student needed in the previous problem. In faded examples, the student needed to complete one or two steps. The results show that students in the adaptive condition learned significantly more than their peers who were presented with a fixed sequence of worked examples and problems.

We introduce kLog, a novel language for kernel-based learning on expressive logical and relational representations. kLog allows users to specify logical and relational learning problems declaratively. It builds on simple but powerful concepts: learning from interpretations, entity/relationship data modeling, and logic programming. Access by the kernel to the rich representation is mediated by a technique we call graphicalization: the relational representation is first transformed into a graph — in particular, a grounded entity/relationship diagram. Subsequently, a choice of graph kernel defines the feature space. The kLog framework can be applied to tackle the same range of tasks that has made statistical relational learning so popular, including classification, regression, multitask learning, and collective classification. An empirical evaluation shows that kLog can be either more accurate, or much faster at the same level of accuracy, than Tilde and Alchemy.

In this extended abstract we introduce the Arcade Learning Environment (ALE): both a challenge problem and a platform and methodology for evaluating the development of general, domain-independent AI technology. ALE provides an interface to hundreds of Atari 2600 game environments, each one different, interesting, and designed to be a challenge for human players. ALE presents significant research challenges for reinforcement learning, model learning, model-based planning, imitation learning, transfer learning, and intrinsic motivation. Most importantly, it provides a rigorous testbed for evaluating and comparing approaches to these problems. We illustrate the promise of ALE by presenting a benchmark set of domain-independent agents designed using well-established AI techniques for both reinforcement learning and planning. In doing so, we also propose an evaluation methodology made possible by ALE, reporting empirical results on over 55 different games. We conclude with a brief update on the latest ALE developments. All of the software, including the benchmark agents, is publicly available.

But growing up, in conflict with her father We present a graphic novel project aiming at illustrating current research results and issues regarding the creative process and its relation with artificial intelligence. The main character, Max Order, is an artist who symbolizes the difficulty of coming up with new, creative ideas, giving up imitation of others and finding one's own style.

We propose a technique for decomposing and compressing the dataset in the parameter learning problem in Markov random fields. Our technique applies to incomplete datasets and exploits variables that are always observed in the given dataset. We show that our technique allows exact computation of the gradient and the likelihood, and can lead to orders-of-magnitude savings in learning time.

Multi-label classification is a challenging and appealing supervised learning problem where a subset of labels, rather than a single label seen in traditional classification problems, is assigned to a single test instance. Classifier chains based methods are a promising strategy to tackle multi-label classification problems as they model label correlations at acceptable complexity. However, these methods are difficult to approximate the underlying dependency in the label space, and suffer from the problems of poorly ordered chain and error propagation. In this paper, we propose a novel polytree-augmented classifier chains method to remedy these problems. A polytree is used to model reasonable conditional dependence between labels over attributes, under which the directional relationship between labels within causal basins could be appropriately determined. In addition, based on the max-sum algorithm, exact inference would be performed on polytrees at reasonable cost, preventing from error propagation. The experiments performed on both artificial and benchmark multi-label data sets demonstrated that the proposed method is competitive with the state-of-the-art multi-label classification methods.