Elias Oliveira

Automated multi-label text categorization with VG-RAM weightless neural networks

In automated multi-label text categorization, an automatic categorization system should output a label set, whose size is unknown a priori, for each document under analysis. Many machine learning techniques have been used for building such automatic text categorization systems. In this paper, we examine virtual generalizing random access memory weightless neural networks (VG-RAM WNN), an effective machine learning technique which offers simple implementation and fast training and test, as a tool for building automatic multi-label text categorization systems. We evaluated the performance of VG-RAM WNN on two real-world problems:, (i) categorization of free-text descriptions of economic activities and (ii) categorization of Web pages, and compared our results with that of the multi-label lazy learning approach (Multi-Label K-Nearest Neighbors, ML-KNN). Our experimental comparative analysis showed that, on average, VG-RAM WNN either outperforms ML-KNN or show similar categorization performance.

Face Recognition with VG-RAM Weightless Neural Networks

Virtual Generalizing Random Access Memory Weightless Neural Networks ( Vg-ram wnn ) are effective machine learning tools that offer simple implementation and fast training and test. We examined the performance of Vg-ram wnn on face recognition using a well known face database--the AR Face Database. We evaluated two Vg-ram wnn architectures configured with different numbers of neurons and synapses per neuron. Our experimental results show that, even when training with a single picture per person, Vg-ram wnn are robust to various facial expressions, occlusions and illumination conditions, showing better performance than many well known face recognition techniques.

Human automatic detection and tracking for outdoor video

Human automatic tracking is still a problem. This paper presents an approach to treat this problem, where neither the target color model nor the background color model are initially supposed to be known. Besides, the illumination conditions and background may change and, furthermore, the target can be occluded for a determined time. In such scenario, this work proposes an approach for human tracking in outdoor environment by identifying the target. A modified version of Mean Shift is employed, in which it is used HSV color space and a procedure to update its color model. In addition, a method to find the face of the person in video sequence is also presented, in order to identify people or to validate that the target is a human. For evaluation, four metrics were proposed and it has been carried out a series of experiments, and also the presented method was superior to other techniques evaluated.

An Evolving System Based on Probabilistic Neural Network

Although traditional techniques of machine learning have, in many cases, presented good results, they have been inefficient for data which are constantly expanding and changing over time. To address these problems, new learning techniques have been proposed in the literature. In this paper we propose a technique called ePNN presenting aspects of this recent paradigm of learning. We carried out a series of experiments that showed its efficiency over previous approaches.

Recommendation of programming activities by multi-label classification for a formative assessment of students

Computer programming ability is a type of knowledge that is considered to be quite complex because it demands many cognitive skills and extensive practice to be mastered. However, formative assessment is a strategy that can improve learning. For this reason, we developed a recommender system to aid in making choices on programming practices by recommending classes of activities. This system provides instructors with a means of semi-automatic assessment, with more individualised and accurate activities tailored to the needs of their learners. To achieve this goal, the system of recommendations analyses multidimensional profiles of new students and seeks the best match for them among profiles in the logs of previous recommendations, which were made manually. Based on these matched profiles, the system can now recommend to new learners classes of activities that are indicated by similar profiles that have already received recommendations. The recommendation of activities is thus treated by our system as a multi-label classification task in which each students profile is associated with one or more classes of programming activities. The results obtained under different evaluation metrics confirm that the chosen algorithm, the ML-kNN, correctly mimics human decisions on the recommendations of classes of activities most of the time. Furthermore, these metrics provide relevant information for instructors to perform better actions with regard to formative assessments.

An incremental neural network with a reduced architecture

This paper proposes a technique, called Evolving Probabilistic Neural Network (ePNN), that presents many interesting features, including incremental learning, evolving architecture, the capacity to learn continually throughout its existence and requiring that each training sample be used only once in the training phase without reprocessing. A series of experiments was performed on data sets in the public domain; the results indicate that ePNN is superior or equal to the other incremental neural networks evaluated in this paper. These results also demonstrate the advantage of the small ePNN architecture and show that its architecture is more stable than the other incremental neural networks evaluated. ePNN thus appears to be a promising alternative for a quick learning system and a fast classifier with a low computational cost.

Agglomeration and Elimination of Terms for Dimensionality Reduction

The vector space model is the usual representation of texts database for computational treatment. However, in such representation synonyms and/or related terms are treated as independent. Furthermore, there are some terms that do not add any information at all to the set of text documents, on the contrary they even might harm the performance of the information retrieval techniques. In an attempt to reduce this problem, some techniques have been proposed in the literature. In this work we present a method to tackle this problem. In order to validate our approach, we carried out a serie of experiments on four databases and we compare the achieved results with other well known techniques. The evaluation results is such that our method obtained in all cases a better or equal performance compared to the other literature techniques.

Multi-label incremental learning applied to web page categorization

Multi-label problems are challenging because each instance may be associated with an unknown number of categories, and the relationship among the categories is not always known. A large amount of data is necessary to infer the required information regarding the categories, but these data are normally available only in small batches and distributed over a period of time. In this work, multi-label problems are tackled using an incremental neural network known as the evolving Probabilistic Neural Network (ePNN). This neural network is capable of continuous learning while maintaining a reduced architecture, so that it can always receive training data when available with no drastic growth of its structure. We carried out a series of experiments on web page data sets and compared the performance of ePNN to that of other multi-label categorizers. On average, ePNN outperformed the other categorizers in four out of five metrics used for evaluation, and the structure of ePNN was less complex than that of the other algorithms evaluated.

Using a Probabilistic Neural Network for a Large Multi-label Problem

The automation of the categorization of economic activities from business descriptions in free text format is a huge challenge for the Brazilian governmental administration in the present day. When this problem is tackled by humans, the subjectivity on their classification brings another problem: different human classifiers can give different results when working on a set of the same business descriptions. This can cause a serious distortion on the information for the planning and taxation of the governmental administrations on the three levels: County, State and Federal. Furthermore, the number of possible categories considered is very large, more than 1000 in the Brazilian scenario. The large number of categories makes the problem even harder to be solved, as this is also a multi-labeled problem. In this work we compared the multi-label lazy learning technique, ML-kNN, to our probabilistic neural network approach. Our implementation overcome the ML-kNN algorithm in four metrics typically used in the literature for multi-label categorization problems.

Achieving a compromise between performance and complexity of structure

In incremental learning techniques, learning occurs continuously over time and does not cease once available data have been exhausted. Such techniques are useful in cases where problem data may be acquired in small quantities over time. This paper presents an incremental neural network called the evolving Probabilistic Neural Network. The main advantage of this technique lies in its adaptive architecture, which adjusts to data distributions. This method requires that each training sample be used only once throughout the training phase without being reprocessed. The technique is flexible and offers a simplified structure while maintaining performance levels comparable to those of other techniques. Experiments were conducted using publicly available benchmark data sets. These experiments show that overall, the proposed model achieves a quality of response that is comparable to those of the best techniques evaluated, and its structure size and classification time were as low as those of less complex techniques. These results indicate that the proposed model achieves a satisfactory balance between efficiency and efficacy.