For large, real-world inductive learning problems, the number of training examples often must be limited due to the costs associated with procuring, preparing, and storing the training examples and/or the computational costs associated with learning from them. In such circumstances, one question of practical importance is: if only n training examples can be selected, in what proportion should the classes be represented? In this article we help to answer this question by analyzing, for a fixed training-set size, the relationship between the class distribution of the training data and the performance of classification trees induced from these data. We study twenty-six data sets and, for each, determine the best class distribution for learning. The naturally occurring class distribution is shown to generally perform well when classifier performance is evaluated using undifferentiated error rate (0/1 loss). However, when the area under the ROC curve is used to evaluate classifier performance, a balanced distribution is shown to perform well. Since neither of these choices for class distribution always generates the best-performing classifier, we introduce a "budget-sensitive" progressive sampling algorithm for selecting training examples based on the class associated with each example. An empirical analysis of this algorithm shows that the class distribution of the resulting training set yields classifiers with good (nearly-optimal) classification performance.

We introduce new online and batch algorithms that are robust to data with missing features, a situation that arises in many practical applications. In the online setup, we allow for the comparison hypothesis to change as a function of the subset of features that is observed on any given round, extending the standard setting where the comparison hypothesis is fixed throughout. In the batch setup, we present a convex relation of a non-convex problem to jointly estimate an imputation function, used to fill in the values of missing features, along with the classification hypothesis. We prove regret bounds in the online setting and Rademacher complexity bounds for the batch i.i.d. setting. The algorithms are tested on several UCI datasets, showing superior performance over baselines.

This paper discusses a system that accelerates reinforcement learning by using transfer from related tasks. Without such transfer, even if two tasks are very similar at some abstract level, an extensive re-learning effort is required. The system achieves much of its power by transferring parts of previously learned solutions rather than a single complete solution. The system exploits strong features in the multi-dimensional function produced by reinforcement learning in solving a particular task. These features are stable and easy to recognize early in the learning process. They generate a partitioning of the state space and thus the function. The partition is represented as a graph. This is used to index and compose functions stored in a case base to form a close approximation to the solution of the new task. Experiments demonstrate that function composition often produces more than an order of magnitude increase in learning rate compared to a basic reinforcement learning algorithm.

Imitation can be viewed as a means of enhancing learning in multiagent environments. It augments an agent's ability to learn useful behaviors by making intelligent use of the knowledge implicit in behaviors demonstrated by cooperative teachers or other more experienced agents. We propose and study a formal model of implicit imitation that can accelerate reinforcement learning dramatically in certain cases. Roughly, by observing a mentor, a reinforcement-learning agent can extract information about its own capabilities in, and the relative value of, unvisited parts of the state space. We study two specific instantiations of this model, one in which the learning agent and the mentor have identical abilities, and one designed to deal with agents and mentors with different action sets. We illustrate the benefits of implicit imitation by integrating it with prioritized sweeping, and demonstrating improved performance and convergence through observation of single and multiple mentors. Though we make some stringent assumptions regarding observability and possible interactions, we briefly comment on extensions of the model that relax these restricitions.

The recursive least-squares (RLS) algorithm is one of the most well-known algorithms used in adaptive filtering, system identification and adaptive control. Its popularity is mainly due to its fast convergence speed, which is considered to be optimal in practice. In this paper, RLS methods are used to solve reinforcement learning problems, where two new reinforcement learning algorithms using linear value function approximators are proposed and analyzed. The two algorithms are called RLS-TD(lambda) and Fast-AHC (Fast Adaptive Heuristic Critic), respectively. RLS-TD(lambda) can be viewed as the extension of RLS-TD(0) from lambda=0 to general lambda within interval [0,1], so it is a multi-step temporal-difference (TD) learning algorithm using RLS methods. The convergence with probability one and the limit of convergence of RLS-TD(lambda) are proved for ergodic Markov chains. Compared to the existing LS-TD(lambda) algorithm, RLS-TD(lambda) has advantages in computation and is more suitable for online learning. The effectiveness of RLS-TD(lambda) is analyzed and verified by learning prediction experiments of Markov chains with a wide range of parameter settings. The Fast-AHC algorithm is derived by applying the proposed RLS-TD(lambda) algorithm in the critic network of the adaptive heuristic critic method. Unlike conventional AHC algorithm, Fast-AHC makes use of RLS methods to improve the learning-prediction efficiency in the critic. Learning control experiments of the cart-pole balancing and the acrobot swing-up problems are conducted to compare the data efficiency of Fast-AHC with conventional AHC. From the experimental results, it is shown that the data efficiency of learning control can also be improved by using RLS methods in the learning-prediction process of the critic. The performance of Fast-AHC is also compared with that of the AHC method using LS-TD(lambda). Furthermore, it is demonstrated in the experiments that different initial values of the variance matrix in RLS-TD(lambda) are required to get better performance not only in learning prediction but also in learning control. The experimental results are analyzed based on the existing theoretical work on the transient phase of forgetting factor RLS methods.

Such maps specify the topology of important landmarks and situations (states), and routes or transitions (arcs) between them. They are concerned less with the physical location of landmarks, and more with topological relationships between situations. Typically, they are less complex and support much more ecient planning than metric maps. Topological maps are built on lowerlevel abstractions that allow the robot to move along arcs (perhaps by wall-or road-following), to recognize properties of locations, and to distinguish signicant locations as states; they are exible in allowing a more general notion of state, possibly including information about the non-geometrical aspects of the robot's situation. There are two typical strategies for deriving topological maps: one is to learn the topological map directly; the other is to rst learn a geometric map, then to derive a topological model from it through some process of analysis. A nice example of the second approach is provided by Thrun and B--ucken (1996a, 1996b; Thrun, 1999), who use occupancy-grid techniques to build the initial map. This strategy is appropriate when the primary cues for decomposition and abstraction of the map are geometric. However, in many cases, the nodes of a topological map are dened in terms of other sensory data (e.g., labels on a door or whether or not the robot is holding a bagel). Learning a geometric map rst also relies on the odometric abilities of a robot; if they are weak and the space is large, it is very dicult to derive a consistent map.

A major problem in machine learning is that of inductive bias: how to choose a learner's hypothesis space so that it is large enough to contain a solution to the problem being learnt, yet small enough to ensure reliable generalization from reasonably-sized training sets. Typically such bias is supplied by hand through the skill and insights of experts. In this paper a model for automatically learning bias is investigated. The central assumption of the model is that the learner is embedded within an environment of related learning tasks. Within such an environment the learner can sample from multiple tasks, and hence it can search for a hypothesis space that contains good solutions to many of the problems in the environment. Under certain restrictions on the set of all hypothesis spaces available to the learner, we show that a hypothesis space that performs well on a sufficiently large number of training tasks will also perform well when learning novel tasks in the same environment. Explicit bounds are also derived demonstrating that learning multiple tasks within an environment of related tasks can potentially give much better generalization than learning a single task.

There are two distinct approaches to solving reinforcement learning problems, namely, searching in value function space and searching in policy space. Temporal difference methods and evolutionary algorithms are well-known examples of these approaches. Kaelbling, Littman and Moore recently provided an informative survey of temporal difference methods. This article focuses on the application of evolutionary algorithms to the reinforcement learning problem, emphasizing alternative policy representations, credit assignment methods, and problem-specific genetic operators. Strengths and weaknesses of the evolutionary approach to reinforcement learning are presented, along with a survey of representative applications.

In many real-world learning tasks, it is expensive to acquire a sufficient number of labeled examples for training. This paper investigates methods for reducing annotation cost by `sample selection'. In this approach, during training the learning program examines many unlabeled examples and selects for labeling only those that are most informative at each stage. This avoids redundantly labeling examples that contribute little new information. Our work follows on previous research on Query By Committee, extending the committee-based paradigm to the context of probabilistic classification. We describe a family of empirical methods for committee-based sample selection in probabilistic classification models, which evaluate the informativeness of an example by measuring the degree of disagreement between several model variants. These variants (the committee) are drawn randomly from a probability distribution conditioned by the training set labeled so far. The method was applied to the real-world natural language processing task of stochastic part-of-speech tagging. We find that all variants of the method achieve a significant reduction in annotation cost, although their computational efficiency differs. In particular, the simplest variant, a two member committee with no parameters to tune, gives excellent results. We also show that sample selection yields a significant reduction in the size of the model used by the tagger.

The goal of this approach is to improve classication accuracies produced by learning algorithms by improving the quality of the training data. Our approach uses a set of learning algorithms to create classiers that serve as noise lters for the training data. We evaluate single algorithm, majority vote and consensus lters on ve datasets that are prone to labeling errors. Our experiments illustrate that ltering signicantly improves classication accuracy for noise levels up to 30%. An analytical and empirical evaluation of the precision of our approach shows that consensus lters are conservative at throwing away good data at the expense of retaining bad data and that majority lters are better at detecting bad data at the expense of throwing away good data. This suggests that for situations in which there is a paucity of data, consensus lters are preferable, whereas majority vote lters are preferable for situations with an abundance of data. 1. Introducti The maximum accuracy achievable depends on the quality of the data and on the appropriateness of the chosen learning algorithm for the data. The work described here focuses on improving the quality of training data by identifying and eliminating mislabeled instances prior to applying the chosen learning algorithm, thereby increasing classication accuracy. Labeling error can occur for several reasons including subjectivity, data-entry error, or inadequacy of the information used to label each object. Subjectivity may arise when observations need to be ranked in some way such as disease severity or when the information used to label an object is dierent from the information to which the learning algorithm will have access. For example, when labeling pixels in image data, the analyst typically uses visual input rather than the numeric values of the feature vector corresponding to the observation. Domains in which experts disagree are natural places for subjective labeling errors (Smyth, 1996). A third cause of labeling error arises when the information used to label each observation is inadequate. For example, in the medical domain it may not be possible to perform the tests necessary to guarantee that a diagnosis is 100% accurate. For domains in which labeling errors occur, an automated method of eliminating or correcting mislabeled observations will improve the predictive accuracy of the classier formed from the training data. In this article we address the problem of identifying training instances that are mislabeled.