Showing 1 to 37 of 37 matching Articles
Results per page:
Export (CSV)
By
RamírezCorona, Mallinali; Sucar, L. Enrique; Morales, Eduardo F.
1 Citations
Hierarchical Multilabel Classification (HMC) is the task of assigning a set of classes to a single instance with the peculiarity that the classes are ordered in a predefined structure. We propose a novel HMC method for tree and Directed Acyclic Graphs (DAG) hierarchies. Using the combined predictions of locals classifiers and a weighting scheme according to the level in the hierarchy, we select the “best” single path for tree hierarchies, and multiple paths for DAG hierarchies. We developed a method that returns paths from the root down to a leaf node (Mandatory Leaf Node Prediction or MLNP) and an extension for Non Mandatory Leaf Node Prediction (NMLNP). For NMLNP we compared several pruning approaches varying the pruning direction, pruning time and pruning condition. Additionally, we propose a new evaluation metric for hierarchical classifiers, that avoids the bias of current measures which favor conservative approaches when using NMLNP. The proposed approach was experimentally evaluated with 10 tree and 8 DAG hierarchical datasets in the domain of protein function prediction. We concluded that our method works better for deep, DAG hierarchies and in general NMLNP improves MLNP.
more …
By
Montero, José Antonio; Sucar, L. Enrique
1 Citations
Most gesture recognition systems are based only on hand motion information, and are designed mainly for communicative gestures. However, many activities of everyday life involve interaction with surrounding objects. We propose a new approach for the recognition of manipulative gestures that interact with objects in the environment. The method uses nonintrusive visionbased techniques. The hands of a person are detected and tracked using an adaptive skin color segmentation process, so the system can operate in a wide range of lighting conditions. Gesture recognition is based on hidden Markov models, combining motion and contextual information, where the context refers to the relation of the position of the hand with other objects. The approach was implemented and evaluated on two different domains: video conference and assistance, obtaining gesture recognition rates from 94 % to 99.47 %. The system is very efficient so it is adequate for use in realtime applications.
more …
By
HernandezLeal, Pablo; Escalante, Hugo Jair; Sucar, L. Enrique
1 Citations
Video surveillance is an important problem that has been studied for several years. Nowadays, in the context of smart cities, intelligent video surveillance is an important topic which has several subproblems which need to be solved and then integrated. For example, on one side there are several algorithms for detection, recognition and tracking of objects and people. On the other side, it is necessary to recognize not only objects and persons but complex behaviors (fights, thefts, attacks). To solve these challenges, the use of ontologies has been proposed as a tool to reduce this gap between low and high level information. In this work, we present the foundations of an ontology to be used in an intelligent video surveillance system.
more …
By
HernandezLeal, Pablo; Sucar, L. Enrique; Gonzalez, Jesus A.; Morales, Eduardo F.; Ibarguengoytia, Pablo H.
Show all (5)
1 Citations
Diagnosis in industrial domains is a complex problem because it includes uncertainty management and temporal reasoning. Dynamic Bayesian Networks (DBN) can deal with this type of problem, however they usually lead to complex models. Temporal Nodes Bayesian Networks (TNBNs) are an alternative to DBNs for temporal reasoning that result in much simpler and efficient models in certain domains. However, methods for learning this type of models from data have not been developed. In this paper we propose a learning algorithm to obtain the structure and temporal intervals for TNBNs from data. The method has three phases: (i) obtain an initial interval approximation, (ii) learn the network structure based on the intervals, and (iii) refine the intervals for each temporal node. The number of possible sets of intervals is obtained for each temporal node based on a clustering algorithm and the set of intervals that maximizes the prediction accuracy is selected. We applied this method to learn a TNBN for diagnosis and prediction in a combined cycle power plant. The proposed algorithm obtains a simple model with high predictive accuracy.
more …
By
Escalante, H. Jair; Montes, Manuel; Sucar, L. Enrique
2 Citations
Accuracy of current automatic image labeling methods is under the requirements of annotationbased image retrieval systems. The performance of most of these labeling methods is poor if we just consider the most relevant label for a given region. However, if we look within the set of the top− k candidate labels for a given region, accuracy of most of these systems is improved. In this paper we take advantage of this fact and propose a method (NBI) based on word cooccurrences that uses the naïve Bayes formulation for improving automatic image annotation methods. Our approach utilizes cooccurrence information of the candidate labels for a region with those candidate labels for the other surrounding regions, within the same image, for selecting the correct label. Cooccurrence information is obtained from an external collection of manually annotated images: the IAPRTC12 benchmark. Experimental results using a k −nearest neighbors method as our annotation system, give evidence of significant improvements after applying the NBI method. NBI is efficient since the cooccurrence information was obtained offline. Furthermore, our method can be applied to any other annotation system that ranks labels by their relevance.
more …
By
Gomez, G.; Sucar, L. Enrique; Gillies, Duncan F.
We propose a novel visual navigation clue based on summarising the local 3D information into a single structure, called pq histogram. This structure is obtained by discretising the local orientation or [p, q] of an image, and building a two dimensional histogram. This histogram gives a global view of the 3D shape of the world, therefore it could be used for navigation. The potential application of the pqhistogram is illustrated in two domains. Firstly, semiautomatic navigation of an endoscope inside the human colon. Secondly, mobile robot navigating in certain environments, such as corridors and mines. In both cases, the method was tested with real images with very good results.
more …
By
Romero, Leonardo; Morales, Eduardo F.; Sucar, L. Enrique
To learn a map of an environment a mobile robot has to explore its workspace. This paper introduces a new exploration approach that minimizes movements of the robot to reach the nearest unexplored region of the environment. In contrast to other methods, this approach takes into account rotations of the robot as well as the distance traveled by the robot, to compute an optimal movement policy to reach the nearest unexplored region. The robot acquires a kind of inertial mass that decreases the number of movements that changes the orientation of the robot, and hence reduces odometric errors. This approach is tested using a mobile robot simulator with very good results.
more …
By
Chang, Leonardo; AriasEstrada, Miguel; HernándezPalancar, José; Sucar, L. Enrique
Show all (4)
Shape information have proven to be useful in many computer vision applications. In this work, a selfcontaining shape descriptor for open and closed contours is proposed. Also, a partial shape matching method robust to partial occlusion and noise in the contour is proposed. Both the shape descriptor and the matching method are invariant to rotation and translation. Experiments were carried out in the Shapes99 and Shapes216 datasets, where contour segments of different lengths were removed to obtain partial occlusion as high as 70%. For the highest occlusion levels the proposed method outperformed other popular shape description methods, with up to 50% higher bull’s eye score.
more …
By
Ruiz, Elías; Sucar, L. Enrique
A novel proposal for a compositional model for object recognition is presented. The proposed method is based on visual grammars and Bayesian networks. An object is modeled as a hierarchy of features and spatial relationships. The grammar is learned automatically from examples. This representation is automatically transformed into a Bayesian network. Thus, recognition is based on probabilistic inference in the Bayesian network representation. Preliminary results in recognition of natural objects are presented. The main contribution of this work is a general methodology for building object recognition systems which combines the expressivity of a grammar with the robustness of probabilistic inference.
more …
By
IzquierdoCordova, Ramon; Morales, Eduardo F.; Sucar, L. Enrique; MurrietaCid, Rafael
Show all (4)
We consider the problem of exploring a known structured environment to find an object with a mobile robot. We proposed a novel heuristicbased strategy for reducing the traveled distance by first obtaining an exploration order of the rooms in the environment and then, searching for the object in each room by positioning the robot through a set of viewpoints. For the exploration order we proposed a heuristic based on the distance from the robot to the room, the probability of finding the object therein and the room area; integrated in a
$$O(n^2)$$
complexity greedy algorithm that selects the next room. The experimental results show an advantage of the proposed heuristic over other methods in terms of expected traveled distance, except for full search which has a complexity of O(n!). For the exploration within each room, we integrate the localization of horizontal flat surfaces with the generation of poses. With the set of poses, a similar heuristic establishes the exploration order that guides the robot path inside the room. The evaluation of the set of poses shows an average coverage of the flat surfaces of more than 90% when it is configured with an overlap of 40%. Experiments were performed with a real robot using three objects in a sixroom environment. The success rate for the robot finding the object is 86.6%.
more …
By
Ruiz, Elias; Melendez, Augusto; Sucar, L. Enrique
A novel approach to create a general vision system is presented. The proposed method is based on a visual grammar representation which is transformed to a Bayesian network which is used for object recognition. We use a symbolrelational grammar for a hierarchical description of objects, incorporating spatial relations. The structure of a Bayesian network is obtained automatically from the grammar, and its parameters are learned from examples. The method is illustrated with two examples for face recognition.
more …
By
HernándezGracidas, Carlos; Sucar, L. Enrique
7 Citations
Contentbased image retrieval (CBIR) is currently limited because of the lack of representational power of the lowlevel image features, which fail to properly represent the actual contents of an image, and consequently poor results are achieved with the use of this sole information. Spatial relations represent a class of highlevel image features which can improve image annotation. We apply spatial relations to automatic image annotation, a task which is usually a first step towards CBIR. We follow a probabilistic approach to represent different types of spatial relations to improve the automatic annotations which are obtained based on lowlevel features. Different configurations and subsets of the computed spatial relations were used to perform experiments on a database of landscape images. Results show a noticeable improvement of almost 9% compared to the base results obtained using the kNearest Neighbor classifier.
more …
By
Rodríguez, Andrés F.; Vadera, Sunil; Sucar, L. Enrique
An exemplarbased model with foundations in Bayesian networks is described. The proposed model utilises two Bayesian networks: one for indexing of categories, and another for identifying exemplars within categories. Learning is incrementally conducted each time a new case is classified. The representation structure dynamically changes each time a new case is classified and a prototypicality function is used as a basis for selecting suitable exemplars. The results of evaluating the model on three datasets are presented.
more …
By
Oves García, Reinier; Valentin, Luis; MartínezCarranza, José; Sucar, L. Enrique
Show all (4)
This paper presents a fast algorithm for camera selection in a robotic multicamera localization system. The scenario we study is that where a robot is navigating in an indoor environment using a fourcamera vision system to localize itself inside the world. In this context, when something occludes the current camera used for localization, the system has to switch to one of the other three available cameras to remain localized. In this context, the question that arises is that of “what camera should be selected?”. We address this by proposing an approach that aims at selecting the next best view to carry on the localization. For that, the number of static features at each direction is estimated using the optical flow. In order to validate our approach, experiments in a real scenario with a mobile robot system are presented.
more …
By
HernandezLeal, Pablo; Cote, Enrique Munoz; Sucar, L. Enrique
1 Citations
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 nonstationary (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 nonstationary opponent that switches among different strategies.
more …
By
Oves García, Reinier; Valentin, Luis; Pérez Risquet, Carlos; Sucar, L. Enrique
Show all (4)
Background subtraction is an important task in video processing and many algorithms are developed for solving this task. The vast majority uses the static behavior of the scene or texture information for separating foreground and background. In this paper we present a novel approach based on the integration of the unsteady vector field embedded in the video. Our method does not learn from the background and neither uses static behavior or texture for detecting the background. This solution is based on motion extraction from the scene by planecurve intersection. The set of blobs generated by the algorithm are equipped with local motion information which can be used for further image analysis tasks. The proposed approach has been evaluated with a standard benchmark with competitive results against state of the art methods.
more …
By
Maxhuni, Alban; HernandezLeal, Pablo; Morales, Eduardo F.; Sucar, L. Enrique; Osmani, Venet; MuńozMeléndez, Angelica; Mayora, Oscar
Show all (7)
1 Citations
Motor activity in physical and psychological stress exposure has been studied almost exclusively with selfassessment 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 realworld 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.
more …
By
Noguez, Julieta; Sucar, L. Enrique; Espinosa, Enrique
2 Citations
We have developed a novel student model based on probabilistic relational models (PRMs). This model combines the advantages of Bayesian networks and objectoriented systems. It facilitates knowledge acquisition and makes it easier to apply the model for different domains. The model is oriented towards virtual laboratories, in which a student interacts by doing experiments in a simulated or remote environment. It represents the students’ knowledge at different levels of granularity, combining the performance and exploration behavior in several experiments, to decide the best way to guide the student in the next experiments. Based on this model, we have developed tutors for virtual laboratories in different domains. An evaluation of with a group of students, show a significant improvement in learning when a tutor based on the PRM model is incorporated to a virtual robotics lab.
more …
By
Hernández, Yasmín; Sucar, L. Enrique; ArroyoFigueroa, Gustavo
4 Citations
We are developing an affective behavior model for intelligent tutoring systems. This model provides students with a suitable response considering the knowledge and affective state of the student. The affective behavior model comprises two components: an affective student model and an affective tutor model. The affective student model is based on the OCC cognitive model of emotions. The main contribution of this work is the affective tutor model, which is built with base on the expertise of a group of teachers. For this end, we conducted a study to ask teachers how they deal with affective aspects when they are teaching; and we used the results of the study to refine our model. The affective behavior model relies on dynamic Bayesian networks and dynamic decision networks. In this paper, we present the affective behavior model built based on the teachers´ responses, its application to a robotics tutor and an evaluation with a Wizard of Oz study.
more …
By
Reyes, Alberto; Sucar, L. Enrique; Morales, Eduardo F.; Ibargüengoytia, Pablo H.
Show all (4)
2 Citations
Markov decision processes (MDPs) have developed as a standard for representing uncertainty in decisiontheoretic planning. However, MDPs require an explicit representation of the state space and the probabilistic transition model which, in continuous or hybrid continuousdiscrete domains, are not always easy to define. Even when this representation is available, the size of the state space and the number of state variables to consider in the transition function may be such that the resulting MDP cannot be solved using traditional techniques. In this paper a rewardbased abstraction for solving hybrid MDPs is presented. In the proposed method, we gather information about the rewards and the dynamics of the system by exploring the environment. This information is used to build a decision tree (C4.5) representing a small set of abstract states with equivalent rewards, and then is used to learn a probabilistic transition function using a Bayesian networks learning algorithm (K2). The system output is a problem specification ready for its solution with traditional dynamic programming algorithms. We have tested our abstract MDP model approximation in realworld problem domains. We present the results in terms of the models learned and their solutions for different configurations showing that our approach produces fast solutions with satisfying policies.
more …
By
Vásquez, Juan Irving; Sucar, L. Enrique
3 Citations
To acquire a 3D model of an object it is necessary to plan a set of locations, called views, where a range sensor will be placed. The problem is solved in greedy manner, by selecting iteratively nextbestviews. When a mobile robot is used, we have to take into account positioning errors, given that they can affect the quality and efficiency of the plan. We propose a method to plan “safe views” which are successful even when there is positioning error. The method is based on a reevaluation of the candidate views according to their neighbors, so view points which are safer against positioning error are preferred. The method was tested in simulation with objects of different complexities. Experimental results show that the proposed method achieves similar results as the ideal case without error, reducing the number of views required against the standard approach that does not consider positioning error.
more …
By
Alonso, Víctor E.; EnríquezCaldera, Rogerio; Sucar, L. Enrique
This paper presents a novel hybrid twodirectional, twodimensional Principal Component Analysis based correlation filter for face recognition. This hybrid (2D)
$$^2$$
PCAcorrelation filter is capable of simultaneously dealing with several uncontrolled factors that are present in video surveillance cameras making it difficult to properly recognize faces. Such factors are addressed by linking (2D)
$$^2$$
PCA in the Fourier domain with correlation filters (CFs) to speed up the process of videobased face recognition. The former method helps to extract and to represent more efficiently the facial features using the original image matrices, while the later method is used to simultaneously handle illumination variations, expression, partial occlusions and spatial shifts. An exploration of the capabilities of this novel method is performed using the YaleB, AR, and YouTube face databases, showing an improvement in face recognition despite using a subspace of smaller dimensionality.
more …
By
Noguez, Julieta; Sucar, L. Enrique
10 Citations
Open learning environments often involve simulation where learners can experiment with different aspects and parameters of a given phenomenon to observe the effects of these changes. These are desirable in virtual laboratories. However, an important limitation of open learning environments is the effectiveness for learning, because it strongly depends on the learner ability to explore adequately. We have developed a semiopen learning environment for a virtual robotics laboratory based on simulation, to learn through free exploration, but with specific performance criteria that guide the learning process. We proposed a generic architecture for this environment, in which the key element is an intelligent tutoring system coupled to a virtual laboratory. The tutor module combines the performance and exploration behaviour of a student in several experiments, to decide the best way to guide his/her. We present an evaluation with an initial group of 20 students. The results show how this semiopen leraning environment can help to accelerate and improve the learning process.
more …
By
Ibargüengoytia, Pablo H.; Sucar, L. Enrique; Morales, Eduardo
2 Citations
Diagnosis, in artificial intelligence, has traditionally utilized heuristic rules which in many domains are difficult to acquire. An alternative approach, modelbased diagnosis, utilizes a model of the system and compares its predicted behavior against the actual behavior of the system for diagnosis. This paper presents a novel technique based on probabilistic models. Therefore, it is natural to include uncertainty in the model and in the measurements for diagnosis. This characteristic makes the proposed approach suitable for applications where reliable measurements are unlikely to occur or where a deterministic analytical model is difficult to obtain. The proposed approach can detect single or multiple faults through a vector of probabilities which reflects the degree of belief in the state of all the components of the system. A comparison against GDE, a classical approach for multiple fault diagnosis, is given.
more …
By
Chang, Leonardo; HernándezPalancar, José; Sucar, L. Enrique; AriasEstrada, Miguel
Show all (4)
15 Citations
The use of local features in images has become very popular due to its promising results. They have shown significant benefits in a variety of applications such as object recognition, image retrieval, robot navigation, panorama stitching, and others. SIFT is one of the local features methods that have shown better results. Among its main disadvantages is its high computational cost. In order to speedup this algorithm, this work proposes the design and implementation of an efficient hardware architecture based on FPGAs for SIFT interest point detection In order to take full advantage of the parallelism in this algorithm and to minimize the device area occupied by its implementation in hardware, part of the algorithm was reformulated. The main contribution of the hardware architecture proposed in this paper and the main difference with the rest of the architectures reported in the literature is that as the number of octaves to be processed is increased, the amount of occupied device area remains almost constant. The evaluations and experiments to the architecture support this contribution, as well as accuracy, repeatability, and distinctiveness of the results. Experiments also showed device area occupation and time constraints of the hardware implementation. The architecture presented in this paper is able to detect interest points in an image of 320 × 240 in 11 ms, which represents a speedup of 250 × with respect to a software implementation.
more …
By
Rodriguez, Andrés F.; Vadera, Sunil; Sucar, L. Enrique
3 Citations
An exemplarbased model with foundations in Bayesian networks is described. The proposed model utilises two Bayesian networks: one for indexing of categories, and another for identifying exemplars within categories. Learning is incrementally conducted each time a new case is classified. The representation structure dynamically changes each time a new case is classified and a coverage function is used as a basis for selecting suitable exemplars. Finally, a simple example is given to illustrate the concepts in the model.
more …
By
Luis, Roger; Sucar, L. Enrique; Morales, Eduardo F.
In several domains it is common to have data from different, but closely related problems. For instance, in manufacturing many products follow the same industrial process but with different conditions; or in industrial diagnosis, where there is equipment with similar specifications. In these cases, it is common to have plenty of data for some scenarios but very little for other. In order to learn accurate models for rare cases, it is desirable to use data and knowledge from similar cases; a technique known as “transfer learning”. In this paper, we propose a transfer learning method for Bayesian networks, that considers both, structure and parameter learning. For structure learning, we use conditional independence tests, by combining measures from the target domain with those obtained from one or more auxiliary domains, using a weighted sum of the conditional independence measures. For parameter learning, we compared two techniques for probability aggregation that combine probabilities estimated from the target domain with those obtained from the auxiliary data. To validate our approach, we used three Bayesian networks models that are commonly used for evaluating learning techniques, and generated variants of each model by changing the structure as well as the parameters. We then learned one of the variants with a small data set and combined it with information from the other variants. The experimental results show a significant improvement in terms of structure and parameters when we transfer knowledge from similar problems.
more …
By
ArroyoFigueroa, G.; Sucar, L. Enrique
A methodology for online diagnosis and prediction of power plant disturbances has been developed, implemented, and tested. The approach is sufficiently comprehensive to enable a wide variety of disturbances to be analyzed correctly and efficiently. The analysis is based on a novel knowledge representation, called Temporal Nodes Bayesian Networks (TNBN), a type of probabilistic network that include temporal information. A TNBN has a set of temporal nodes that represent state changes. Each temporal node is defined by a event and a time interval associated to its occurrence. The method has been implemented and integrated with a power plant training simulator. Disturbance models for the feedwater and superheater systems have been developed and implemented as the knowledge database for the disturbance analysis system.
more …
By
Vallejo, Antonio G., Jr.; NolazcoFlores, Juan A.; MoralesMenéndez, Rubén; Sucar, L. Enrique; Rodríguez, Ciro A.
Show all (5)
7 Citations
In this work we propose to monitor the cutting toolwear condition in a CNCmachining center by using continuous Hidden Markov Models (HMM). A database was built with the vibration signals obtained during the machining process. The workpiece used in the milling process was aluminum 6061. Cutting tests were performed on a Huron milling machine equipped with a Sinumerik 840D open CNC. We trained/tested the HMM under 18 different operating conditions. We identified three key transitions in the signals. First, the cutting tool touches the workpiece. Second, a stable waveform is observed when the tool is in contact with the workpiece. Third, the tool finishes the milling process. Considering these transitions, we use a fivestate HMM for modeling the process. The HMMs are created by preprocessing the waveforms, followed by training step using BaumWelch algorithm. In the recognition process, the signal waveform is also preprocessed, then the trained HMM are used for decoding. Early experimental results validate our proposal in exploiting speech recognition frameworks in monitoring machining centers. The classifier was capable of detecting the cutting tool condition within large variations of spindle speed and feed rate, and accuracy of 84.19%.
more …
By
Chang, Leonardo; AriasEstrada, Miguel; HernándezPalancar, José; Sucar, L. Enrique
Show all (4)
Shape information is an important cue for many computer vision applications. In this work we propose an invariant shape feature extraction, description and matching method for binary images, named LISF. The proposed method extracts local features from the contour to describe shape and these features are later matched globally. Combining local features with global matching allows us to a obtaining a tradeoff between discriminative power and robustness to noise and occlusion in the contour. The proposed extraction, description and matching methods are invariant to rotation, translation, and scale and present certain robustness to partial occlusion. The conducted experiments in the Shapes99, Shapes216, and MPEG7 datasets support the mentioned contributions, where different artifacts were artificially added to obtain partial occlusion as high as 60 %. For the highest occlusion levels LISF outperformed other popular shape description methods, with about 20 % higher bull’s eye score and 25 % higher accuracy in classification. Also, in this paper, we present a massively parallel implementation in CUDA of the two most timeconsuming stages of LISF, i.e., the feature extraction and feature matching steps; which achieves speedups of up to 32x and 34x, respectively.
more …
By
HernandezLeal, Pablo; Maxhuni, Alban; Sucar, L. Enrique; Osmani, Venet; Morales, Eduardo F.; Mayora, Oscar
Show all (6)
3 Citations
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 nonobtrusive 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.
more …
By
Oña García, Ana Li; Sucar, L. Enrique; Morales, Eduardo F.
Fault diagnosis in complex systems is important due to the impact it may have for reducing breakage costs or for avoiding production losses in industrial systems. Several approaches have been proposed for fault diagnosis, some of which are based on Bayesian Networks. Bayesian Networks are an adequate formalism for representing and reasoning under uncertainty conditions, however, they do not scale well for complex systems. For overcoming this limitation, researchers have proposed Multiply Sectioned Bayesian Networks. These are an extension of the Bayesian Networks for representing large domains, while ensuring the network inference in an efficient way. In this work we propose a distributed method for fault diagnosis in complex systems using Multiply Sectioned Bayesian Networks. The method was tested in the detection of multiple faults in combinational logic circuits showing comparable results with the literature in terms of accuracy, but with a significant reduction in the runtime.
more …
By
PalaciosAlonso, Miguel A.; Brizuela, Carlos A.; Sucar, L. Enrique
14 Citations
Many problems such as voice recognition, speech recognition and many other tasks have been tackled with Hidden Markov Models (HMMs). These problems can also be dealt with an extension of the Naive Bayesian Classifier (NBC) known as Dynamic NBC (DNBC). From a dynamic Bayesian network (DBN) perspective, in a DNBC at each time there is a NBC. NBCs work well in data sets with independent attributes. However, they perform poorly when the attributes are dependent or when there are one or more irrelevant attributes which are dependent of some relevant ones. Therefore, to increase this classifier accuracy, we need a method to design network structures that can capture the dependencies and get rid of irrelevant attributes. Furthermore, when we deal with dynamical processes there are temporal relations that should be considered in the network design. In order to learn automatically these models from data and increase the classifier accuracy we propose an evolutionary optimization algorithm to solve this design problem. We introduce a new encoding scheme and new genetic operators which are natural extensions of previously proposed encoding and operators for grouping problems. The design methodology is applied to solve the recognition problem for nine hand gestures. Experimental results show that the evolved network has higher average classification accuracy than the basic DNBC and a HMM.
more …
By
Luis, Roger; Sucar, L. Enrique; Morales, Eduardo F.
25 Citations
In several domains it is common to have data from different, but closely related problems. For instance, in manufacturing, many products follow the same industrial process but with different conditions; or in industrial diagnosis, where there is equipment with similar specifications. In these cases it is common to have plenty of data for some scenarios but very little for others. In order to learn accurate models for rare cases, it is desirable to use data and knowledge from similar cases; a technique known as transfer learning. In this paper we propose an inductive transfer learning method for Bayesian networks, that considers both structure and parameter learning. For structure learning we use conditional independence tests, by combining measures from the target task with those obtained from one or more auxiliary tasks, using a novel weighted sum of the conditional independence measures. For parameter learning, we propose two variants of the linear pool for probability aggregation, combining the probability estimates from the target task with those from the auxiliary tasks. To validate our approach, we used three Bayesian networks models that are commonly used for evaluating learning techniques, and generated variants of each model by changing the structure as well as the parameters. We then learned one of the variants with a small dataset and combined it with information from the other variants. The experimental results show a significant improvement in terms of structure and parameters when we transfer knowledge from similar tasks. We also evaluated the method with realworld data from a manufacturing process considering several products, obtaining an improvement in terms of loglikelihood between the data and the model when we do transfer learning from related products.
more …
By
Fiedler, Lindsey J.; Sucar, L. Enrique; Morales, Eduardo F.
1 Citations
Traditional machine learning algorithms depend heavily on the assumption that there is sufficient data to learn a reliable model. This is not always the case, and in situations where data is limited, transfer learning can be applied to compensate for the lack of information by learning from several sources. In this work, we present a novel methodology for inducing a Temporal Nodes Bayesian Network (TNBN) when training data is scarce by applying a transfer learning strategy. A TNBN is a probabilistic graphical model that offers a compact representation for dynamic domains by defining multiple time intervals in which events can occur. Learning a TNBN poses additional challenges to learning traditional Bayesian networks due to the incorporation of time intervals. Our proposal incorporates novel approaches to transfer knowledge from several TNBNs to learn the structure, parameters and intervals of a target TNBN. To evaluate our algorithm, we performed experiments with a synthetic network, where we created auxiliary models by altering the structure, parameters and temporal intervals of the original model. Results show that the proposed algorithm is capable of retrieving a reliable model even when few records are available for the target domain. We also performed experiments with a realworld data set belonging to the medical domain of HIV, where we were able to learn some documented mutational pathways and their temporal relations by applying transfer learning.
more …
By
HernandezLeal, Pablo; Zhan, Yusen; Taylor, Matthew E.; Sucar, L. Enrique; Munoz de Cote, Enrique
Show all (5)
1 Citations
The success or failure of any learning algorithm is partially due to the exploration strategy it exerts. However, most exploration strategies assume that the environment is stationary and nonstrategic. In this work we shed light on how to design exploration strategies in nonstationary and adversarial environments. Our proposed adversarial drift exploration (DE) is able to efficiently explore the state space while keeping track of regions of the environment that have changed. This proposed exploration is general enough to be applied in single agent nonstationary environments as well as in multiagent settings where the opponent changes its strategy in time. We use a two agent strategic interaction setting to test this new type of exploration, where the opponent switches between different behavioral patterns to emulate a nondeterministic, stochastic and adversarial environment. The agent’s objective is to learn a model of the opponent’s strategy to act optimally. Our contribution is twofold. First, we present DE as a strategy for switch detection. Second, we propose a new algorithm called Rmax# for learning and planning against nonstationary opponent. To handle such opponents, Rmax# reasons and acts in terms of two objectives: (1) to maximize utilities in the short term while learning and (2) eventually explore opponent behavioral changes. We provide theoretical results showing that Rmax# is guaranteed to detect the opponent’s switch and learn a new model in terms of finite sample complexity. Rmax# makes efficient use of exploration experiences, which results in rapid adaptation and efficient DE, to deal with the nonstationary nature of the opponent. We show experimentally how using DE outperforms the state of the art algorithms that were explicitly designed for modeling opponents (in terms average rewards) in two complimentary domains.
more …
By
HernandezLeal, Pablo; Zhan, Yusen; Taylor, Matthew E.; Sucar, L. Enrique; Munoz de Cote, Enrique
Show all (5)
3 Citations
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 nonstationary 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 relearn 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.
more …
