Explaining Predictions

This page provides a summary sheet that includes the general goal, reference papers (both mine and external) for an overview of the topic, as well as the domains explored so far. We are also interested in extending the applications of these techniques beyond their traditional domains. If you have expertise in other areas (e.g., neuroscience, gaming, or audio/speech modeling), we would be happy to explore potential extensions into those fields.

Goal: The goal of this research area is to understand the motivations behind specific predictions of a deep neural network. This is the first and the most extensively studied area in eXplainable Artificial Intelligence. While the field has strong baselines for common domains (Vision and NLP), my research aims to explore the limitations of current methods in understudied and specific domains.

Domains: NLP, Robot Navigation.

Reference Papers:

  1. Memory-Based Explanations: [(La Rosa et al., 2020)]
  2. Perturbation-Based Explanations: [Seminal Paper]
  3. Concept-Based Explanations: [Seminal Paper]
  4. Gradient-Based Explanations: [Seminal Paper]

References

2020

  1. Conference
    Explainable Inference on Sequential Data via Memory-Tracking
    Biagio La Rosa, Roberto Capobianco, and Daniele Nardi
    In Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI-20, 2020
    Bib DOI Official PDF Code Blog