Pablo Hernandez-Leal

Multi-label classification with Bayesian network-based chain classifiers

We introduce a novel method for chaining Bayesian classifiers.We provide a detailed analysis of the proposed method.We perform an extensive empirical evaluation.We show very competitive results against related approaches. In multi-label 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 labels (label power-set methods) or by building independent classifiers for each class (binary relevance methods). The first approach suffers from high computationally complexity, while the second approach ignores possible dependencies among classes. Chain classifiers have been recently proposed to address these problems, where each classifier in the chain learns and predicts the label of one class given the attributes and all the predictions of the previous classifiers in the chain. In this paper we introduce a method for chaining Bayesian classifiers that combines the strengths of classifier chains and Bayesian networks for multi-label classification. A Bayesian network is induced from data to: (i) represent the probabilistic dependency relationships between classes, (ii) constrain the number of class variables used in the chain classifier by considering conditional independence conditions, and (iii) reduce the number of possible chain orders. The effects in the Bayesian chain classifier performance of considering different chain orders, training strategies, number of class variables added in the base classifiers, and different base classifiers, are experimentally assessed. In particular, it is shown that a random chain order considering the constraints imposed by a Bayesian network with a simple tree-based structure can have very competitive results in terms of predictive performance and time complexity against related state-of-the-art approaches.

A framework for learning and planning against switching strategies in repeated games

Intelligent agents, human or artificial, often change their behaviour as they interact with other agents. For an agent to optimise its performance when interacting with such agents, it must be capable of detecting and adapting according to such changes. This work presents an approach on how to effectively deal with non-stationary switching opponents in a repeated game context. Our main contribution is a framework for online learning and planning against opponents that switch strategies. We present how two opponent modelling techniques work within the framework and prove the usefulness of the approach experimentally in the iterated prisoners dilemma, when the opponent is modelled as an agent that switches between different strategies (e.g. TFT, Pavlov and Bully). The results of both models were compared against each other and against a state-of-the-art non-stationary reinforcement learning technique. Results reflect that our approach obtains competitive results without needing an offline training phase, as opposed to the state-of-the-art techniques.

InstanceRank based on borders for instance selection

Instance selection algorithms are used for reducing the number of training instances. However, most of them suffer from long runtimes which results in the incapability to be used with large datasets. In this work, we introduce an Instance Ranking per class using Borders (instances near to instances belonging to different classes), using this ranking we propose an instance selection algorithm (IRB). We evaluated the proposed algorithm using k-NN with small and large datasets, comparing it against state of the art instance selection algorithms. In our experiments, for large datasets IRB has the best compromise between time and accuracy. We also tested our algorithm using SVM, LWLR and C4.5 classifiers, in all cases the selection computed by our algorithm obtained the best accuracies in average.

Stress modelling and prediction in presence of scarce data

OBJECTIVE Stress at work is a significant occupational health concern. Recent studies have used various sensing modalities to model stress behaviour based on non-obtrusive data obtained from smartphones. However, when the data for a subject is scarce it becomes a challenge to obtain a good model. METHODS We propose an approach based on a combination of techniques: semi-supervised learning, ensemble methods and transfer learning to build a model of a subject with scarce data. Our approach is based on the comparison of decision trees to select the closest subject for knowledge transfer. RESULTS We present a real-life, unconstrained study carried out with 30 employees within two organisations. The results show that using information (instances or model) from similar subjects can improve the accuracy of the subjects with scarce data. However, using transfer learning from dissimilar subjects can have a detrimental effect on the accuracy. Our proposed ensemble approach increased the accuracy by ≈10% to 71.58% compared to not using any transfer learning technique. CONCLUSIONS In contrast to high precision but highly obtrusive sensors, using smartphone sensors for measuring daily behaviours allowed us to quantify behaviour changes, relevant to occupational stress. Furthermore, we have shown that use of transfer learning to select data from close models is a useful approach to improve accuracy in presence of scarce data.

Learning temporal nodes Bayesian networks

Abstract Temporal Nodes Bayesian Networks (TNBNs) are an alternative to Dynamic Bayesian Networks for temporal reasoning with much simpler and efficient models in some domains. TNBNs are composed of temporal nodes, temporal intervals, and probabilistic dependencies. However, methods for learning this type of models from data have not yet been developed. In this paper, we propose a learning algorithm to obtain the structure and temporal intervals for TNBNs from data. The method consists of three phases: (i) obtain an initial approximation of the intervals, (ii) obtain a structure using a standard algorithm and (iii) refine the intervals for each temporal node based on a clustering algorithm. We evaluated the method with synthetic data from three different TNBNs of different sizes. Our method obtains the best score using a combined measure of interval quality and prediction accuracy, and a competitive structural quality with lower running times, compared to other related algorithms. We also present a real world application of the algorithm with data obtained from a combined cycle power plant in order to diagnose temporal faults.

Using a Priori Information for Fast Learning Against Non-stationary Opponents

For an agent to be successful in interacting against many different and unknown types of opponents it should excel at learning fast a model of the opponent and adapt online to non-stationary (changing) strategies. Recent works have tackled this problem by continuously learning models of the opponent while checking for switches in the opponent strategy. However, these approaches fail to use a priori information which can be useful for a faster detection of the opponent model. Moreover, if an opponent uses only a finite set of strategies, then maintaining a list of those strategies would also provide benefits for future interactions, in case of opponents who return to previous strategies (such as periodic opponents). Our contribution is twofold, first, we propose an algorithm that can use a priori information, in the form of a set of models, in order to promote a faster detection of the opponent model. The second is an algorithm that while learning new models keeps a record of them in case the opponent reuses one of those. Our approach outperforms the state of the art algorithms in the field (in terms of model quality and cumulative rewards) in the domain of the iterated prisoner’s dilemma against a non-stationary opponent that switches among different strategies.

Discovering human immunodeficiency virus mutational pathways using temporal Bayesian networks

OBJECTIVE The human immunodeficiency virus (HIV) is one of the fastest evolving organisms in the planet. Its remarkable variation capability makes HIV able to escape from multiple evolutionary forces naturally or artificially acting on it, through the development and selection of adaptive mutations. Although most drug resistance mutations have been well identified, the dynamics and temporal patterns of appearance of these mutations can still be further explored. The use of models to predict mutational pathways as well as temporal patterns of appearance of adaptive mutations could greatly benefit clinical management of individuals under antiretroviral therapy. METHODS AND MATERIAL We apply a temporal nodes Bayesian network (TNBN) model to data extracted from the Stanford HIV drug resistance database in order to explore the probabilistic relationships between drug resistance mutations and antiretroviral drugs unveiling possible mutational pathways and establishing their probabilistic-temporal sequence of appearance. RESULTS In a first experiment, we compared the TNBN approach with other models such as static Bayesian networks, dynamic Bayesian networks and association rules. TNBN achieved a 64.2% sparser structure over the static network. In a second experiment, the TNBN model was applied to a dataset associating antiretroviral drugs with mutations developed under different antiretroviral regimes. The learned models captured previously described mutational pathways and associations between antiretroviral drugs and drug resistance mutations. Predictive accuracy reached 90.5%. CONCLUSION Our results suggest possible applications of TNBN for studying drug-mutation and mutation-mutation networks in the context of antiretroviral therapy, with direct impact on the clinical management of patients under antiretroviral therapy. This opens new horizons for predicting HIV mutational pathways in immune selection with relevance for antiretroviral drug development and therapy plan.

Using Intermediate Models and Knowledge Learning to Improve Stress Prediction

Motor activity in physical and psychological stress exposure has been studied almost exclusively with self-assessment questionnaires and from reports that derive from human observer, such as verbal rating and simple descriptive scales. However, these methods are limited in objectively quantifying typical behaviour of stress. We propose to use accelerometer data from smartphones to objectively quantify stress levels. Used data was collected in real-world setting, from 29 employees in two different organisations over 5 weeks. To improve classification performance we propose to use intermediate models. These intermediate models represent the mood state of a person which is used to build the final stress prediction model. In particular, we obtained an accuracy of 78.2 % to classify stress levels.

Efficiently detecting switches against non-stationary opponents

Interactions in multiagent systems are generally more complicated than single agent ones. Game theory provides solutions on how to act in multiagent scenarios; however, it assumes that all agents will act rationally. Moreover, some works also assume the opponent will use a stationary strategy. These assumptions usually do not hold in real world scenarios where agents have limited capacities and may deviate from a perfect rational response. Our goal is still to act optimally in these cases by learning the appropriate response and without any prior policies on how to act. Thus, we focus on the problem when another agent in the environment uses different stationary strategies over time. This will turn the problem into learning in a non-stationary environment, posing a problem for most learning algorithms. This paper introduces DriftER, an algorithm that (1) learns a model of the opponent, (2) uses that to obtain an optimal policy and then (3) determines when it must re-learn due to an opponent strategy change. We provide theoretical results showing that DriftER guarantees to detect switches with high probability. Also, we provide empirical results showing that our approach outperforms state of the art algorithms, in normal form games such as prisoner’s dilemma and then in a more realistic scenario, the Power TAC simulator.

Stress Modelling Using Transfer Learning in Presence of Scarce Data

Stress at work is a significant occupational health concern nowadays. Thus, researchers are looking to find comprehensive approaches for improving wellness interventions relevant to stress. Recent studies have been conducted for inferring stress in labour settings; they model stress behaviour based on non-obtrusive data obtained from smartphones. However, if the data for a subject is scarce, a good model cannot be obtained. We propose an approach based on transfer learning for building a model of a subject with scarce data. It is based on the comparison of decision trees to select the closest subject for knowledge transfer. We present an study carried out on 30 employees within two organisations. The results show that the in the case of identifying a “similar” subject, the classification accuracy is improved via transfer learning.