Interpretable Neural Networks based classifiers for categorical inputs

Because of the pervasive usage of Neural Networks in human sensitive applications, their interpretability is becoming an increasingly important topic in machine learning. In this work we introduce a simple way to interpret the output function of a neural network classifier that take as input categorical variables. By exploiting a mapping between a neural network classifier and a physical energy model, we show that in these cases each layer of the network, and the logits layer in particular, can be expanded as a sum of terms that account for the contribution to the classification of each input pattern. For instance, at the first order, the expansion considers just the linear relation between input features and output while at the second order pairwise dependencies between input features are also accounted for. The analysis of the contributions of each pattern, after an appropriate gauge transformation, is presented in two cases where the effectiveness of the method can be appreciated.

Entropic selection of concepts unveils hidden topics in documents corpora

The organization and evolution of science has recently become itself an object of scientific quantitative investigation, thanks to the wealth of information that can be extracted from scientific documents, such as citations between papers and co-authorship between researchers. However, only few studies have focused on the concepts that characterize full documents and that can be extracted and analyzed, revealing the deeper organization of scientific knowledge. Unfortunately, several concepts can be so common across documents that they hinder the emergence of the underlying topical structure of the document corpus, because they give rise to a large amount of spurious and trivial relations among documents. To identify and remove common concepts, we introduce a method to gauge their relevance according to an objective information-theoretic measure related to the statistics of their occurrence across the document corpus. After progressively removing concepts that, according to this metric, can be considered as generic, we find that the topic organization displays a correspondingly more refined structure.