An optimal linear separator for the Sonar Signals Classification task

The problem of classifying sonar signals from rocks and mines first studied by Gorman and Sejnowski has become a benchmark against which many learning algorithms have been tested. We show that both the training set and the test set of this benchmark are linearly separable, although with different hyperplanes. Moreover, the complete set of learning and test patterns together, is also linearly separable. We give the weights that separate these sets, which may be used to compare results found by other algorithms.

Automatic Summarization System coupled with a Question-Answering System (QAAS)

To select the most relevant sentences of a document, it uses an optimal decision algorithm that combines several metrics. The metrics processes, weighting and extract pertinence sentences by statistical and informational algorithms. This technique might improve a Question-Answering system, whose function is to provide an exact answer to a question in natural language. In this paper, we present the results obtained by coupling the Cortex summarizer with a Question-Answering system (QAAS). Two configurations have been evaluated. In the first one, a low compression level is selected and the summarization system is only used as a noise filter. In the second configuration, the system actually functions as a summarizer, with a very high level of compression. Our results on French corpus demonstrate that the coupling of Automatic Summarization system with a Question-Answering system is promising. Then the system has been adapted to generate a customized summary depending on the specific question. Tests on a french multi-document corpus have been realized, and the personalized QAAS system obtains the best performances.

Combining Supervised and Unsupervised Learning for GIS Classification

This paper presents a new hybrid learning algorithm for unsupervised classification tasks. We combined Fuzzy c-means learning algorithm and a supervised version of Minimerror to develop a hybrid incremental strategy allowing unsupervised classifications. We applied this new approach to a real-world database in order to know if the information contained in unlabeled features of a Geographic Information System (GIS), allows to well classify it. Finally, we compared our results to a classical supervised classification obtained by a multilayer perceptron.

Statistical Automatic Summarization in Organic Chemistry

We present an oriented numerical summarizer algorithm, applied to producing automatic summaries of scientific documents in Organic Chemistry. We present its implementation named Yachs (Yet Another Chemistry Summarizer) that combines a specific document pre-processing with a sentence scoring method relying on the statistical properties of documents. We show that Yachs achieves the best results among several other summarizers on a corpus of Organic Chemistry articles.

Adaptive Learning with Binary Neurons

A efficient incremental learning algorithm for classification tasks, called NetLines, well adapted for both binary and real-valued input patterns is presented. It generates small compact feedforward neural networks with one hidden layer of binary units and binary output units. A convergence theorem ensures that solutions with a finite number of hidden units exist for both binary and real-valued input patterns. An implementation for problems with more than two classes, valid for any binary classifier, is proposed. The generalization error and the size of the resulting networks are compared to the best published results on well-known classification benchmarks. Early stopping is shown to decrease overfitting, without improving the generalization performance.