Colloquium Computer Science - Prof. J. Vitria I Marca University of Barcelona

Title: Explainable Data Science

Abstract:

Machine learning methods, especially with the rise of neural networks (NNs), are nowadays used widely in scientific, industrial and commercial applications. Usually, these models are trained with regard to high accuracy, but recently there is also a high demand for understanding the way a specific model operates and the underlying reasons for the produced decisions. In this talk, we will analyze different reasons that drive the demand for interpretability and explanations. We will also review some model-agnostic approaches to explainability and how this problem is approached when dealing with deep learning models.