Stress assessment is a complex issue and numerous studies have examined factors that influence stress in working environments. Research studies have shown that monitoring individuals' behaviour parameters during daily life can also help assess stress levels. In this study, we examine assessment of work-related stress using features derived from sensors in smartphones. In particular, we use information from physical activity levels, location, social-interactions, social-activity, and application usage during working days. Our study included 30 employees chosen from two different private companies, monitored over a period of 8 weeks in real work environments. The findings suggest that information from phone sensors shows important correlation with employees perceived stress level. Second, we used machine learning methods to classify perceived stress levels based on the analysis of information provided by smartphones. We used decision trees obtaining 67.57 percent accuracy and 71.73 percent after applying a semi-supervised method. Our results show that stress levels can be monitored in unobtrusive manner, through analysis of smartphone data.

Nearest-neighbor (NN) methods are highly effective and widely used pattern classification techniques. There are, however, some issues that hinder their application for large scale and noisy data sets; including, its high storage requirements, its sensitivity to noisy instances, and the fact that test cases must be compared to all of the training instances. Prototype (PG) and feature generation (FG) techniques aim at alleviating these issues to some extent; where, traditionally, both techniques have been implemented separately. This paper introduces a genetic programming approach to tackle the simultaneous generation of prototypes and features to be used for classification with a NN classifier. The proposed method learns to combine instances and attributes to produce a set of prototypes and a new feature space for each class of the classification problem via genetic programming. An heterogeneous representation is proposed together with ad-hoc genetic operators. The proposed approach overcomes some limitations of NN without degradation in its classification performance. Experimental results are reported and compared with several other techniques. The empirical assessment provides evidence of the effectiveness of the proposed approach in terms of classification accuracy and instance/feature reduction.

Transfer learning focuses on developing methods to reuse information gathered from a source task in order to improve the learning performance in a related task. In this work, we present a novel approach to transfer knowledge between tasks in a reinforcement learning (RL) framework with continuous states and actions, where the transition and policy functions are approximated by Gaussian processes. The novelty in the proposed approach lies in the idea of transferring information about the hyper-parameters of the state transition function from the source task, which represents qualitative knowledge about the type of transition function that the target task might have, constraining the search space and accelerating the learning process. We performed experiments on relevant tasks for RL, which show a clear improvement in the overall performance when compared to state-of-the-art reinforcement learning and transfer learning algorithms for continuous state and action spaces.

In multidimensional classification the goal is to assign an instance to a set of different classes. This task is normally addressed either by defining a compound class variable with all the possible combinations of classes (label power-set methods, LPMs) or by building independent classifiers for each class (binary-relevance methods, BRMs). However, LPMs do not scale well and BRMs ignore the dependency relations between classes. We introduce a method for chaining binary Bayesian classifiers that combines the strengths of classifier chains and Bayesian networks for multidimensional classification. The method consists of two phases. In the first phase, a Bayesian network (BN)that represents the dependency relations between the class variables is learned from data. In the second phase, several chain classifiers are built, such that the order of the class variables in the chain is consistent with the class BN. At the end we combine the results of the different generated orders. Our method considers the dependencies between class variables and takes advantage of the conditional independence relations to build simplified models. We perform experiments with a chain of naïve Bayes classifiers on different benchmark multidimensional datasets and show that our approach out performs other state-of-the-art methods.

Markov decision processes (MDPs) have become a standard method for planning under uncertainty, however they usually assume a sequential process, so a single action is executed at each time step. In some applications, as in robotics, it is required to execute several actions concurrently. For this we propose a framework based on a functional decomposition of the problem into several sub-problems, each represented as a subMDP. Each subMDP is solved independently and their policies are combined to obtain a global solution, such that the actions of each subMDP can be executed concurrently. As we combine the local policies, conflicts between them can arise. We define two kinds of conflicts, resource and behavior conflicts, and propose solutions for both. Resource conflicts are solved off-line via a two-phase process which guarantees a near-optimal global policy. Behavior conflicts are solved on-line based on a set of restrictions specified by the user. If there are no restrictions, all the actions are executed concurrently; otherwise, an arbiter selects the action(s) with higher expected utility. We present experimental results in two cases: (i) a simulated robot navigation problem, with resource conflicts, and (ii) a simulated robot in a message delivery task, with behavior conflicts.

The ability to recognize people is a key element for improving human-robot interaction in service robots. There are many approaches for face recognition; however, these assume unrealistic conditions for a service robot, like having an image with a centered face under controlled illumination. We have developed a novel face recognition system so that a mobile robot can learn new faces and recognize them in real-time in realistic indoor environments. It is able to learn online a new face based on a single frame, which is later used to recognize the person even under different environmental conditions. We employ a preprocessing step to reduce the effect of different illumination conditions, and then identify 3 regions in the face: left eye, right eye and nose-mouth. SIFT features are extracted from each region and stored in a feature vector, which is used for recognition. The matching strategy is able to discard unknown faces and the recognition process uses a Bayesian approach over several frames to improve accuracy. Experimental results in natural environment and with Yale's face database show that that our approach is able to learn different faces from a single image and recognize them on average in three seconds, with very competitive results.

Programming a robot to perform tasks in dynamic environments is a complex process. Teleo-reactive programs (TRPs) (S. Benson and N.J. Nilsson, 1995) have proved to be an effective framework to continuously perform a set of actions to achieve particular goals and react in the presence of unexpected events, however, their definition is a difficult and time-consuming process. In this paper, it is shown how a robot can learn TRPs from human guided traces. A user guides a robot to perform a task and the robot learns how to perform such task in similar dynamic environments.

No abstract available.

Reinforcement learning deals with learning optimal or near optimal policies while inter-acting with the environment. Application domains with many continuous variables are dicult to solve with existing reinforcement learning methods due to the large search space. In this paper, we use a relational rep-resentation to dene powerful abstractions that allow us to incorporate domain knowl-edge and re-use previously learned policies in other similar problems. We also describe how to learn useful actions from human traces us-ing a behavioural cloning approach combined with an exploration phase. Since several con-icting actions may be induced for the same abstract state, reinforcement learning is used to learn an optimal policy over this reduced space. It is shown experimentally how a com-bination of behavioural cloning and reinforce-ment learning using a relational representa-tion is powerful enough to learn how to \ny an aircraft through dierent points in space and dierent turbulence conditions. 1.

A method for exploring and navigating autonomously in indoor environments is described. This method merges a local strategy, similar to wall following to keep the robot close to obstacles, within a global search frame, based on a dynamic programming algorithm. We introduce the concept of travel space as a way to map costs to grid cells based on distances to obstacles. This hybrid approach takes advantages of local strategies that consider perceptual limitations of sensors without losing the completeness of a global search. This exploration and navigation method is tested using a simulated and a real mobile robot with promising results.

This paper describes the application and performance evaluation of a new algorithm for multiple objective optimization problems (MOOP) based on reinforcement learning. The new algorithm, called MDQL, considers a family of agents for each objective function involved in a MOOP. Each agent proposes a solution for its corresponding objective function. Agents leave traces while they construct solutions considering traces made by other agents. The solutions proposed by the agents are evaluated using a non-domination criterion and solutions in the final Pareto set for each iteration are rewarded. A mechanism for the application of MDQL in continuous spaces which considers a fixed set of possible actions for the states (the number of actions depends on the dimensionality of the MOOP), is also proposed. Each action represents a path direction and its magnitude is changed dynamically depending on the evaluation of the state that the agent reached. Constraint handling, based on reinforcement comparison, considers reference values for constraints, penalizing agents violating any of them proportionally to the violation committed. MDQL performance was measured with "error ratio" and "spacing" metrics on four test bed problems suggested in the literature, showing competitive results with state-of-the-art algorithms.