Interpretable Deep Neural Networks | Biagio (Mattia) La Rosa

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 direction is to develop self-interpretable deep neural networks capable of providing explanations alongside their predictions. This involves either generalizing existing designs to novel domains and tasks or creating new ones from scratch. Thus far, our work has focused on leveraging memory mechanisms, prototypes, and interpretable axes as foundational elements for these designs.

Domains: NLP, Vision, Chemistry, Robot Navigation.

Reference Papers:

Memory: [(La Rosa et al., 2022)]
Protoypes-based [(Ragno et al., 2022)] [Seminal Paper]
Interpretable Axes [(Proietti et al., 2023)] [Seminal Paper]

References

2023

Journal

Explainable AI in drug discovery: self-interpretable graph neural network for molecular property prediction using concept whitening

Michela Proietti, Alessio Ragno, Biagio La Rosa, and 2 more authors

Machine Learning, 2023

Bib DOI Official PDF

@article{Proietti2023,
  author = {Proietti, Michela and Ragno, Alessio and {La Rosa}, Biagio and Ragno, Rino and Capobianco, Roberto},
  journal = {Machine Learning},
  title = {Explainable {AI} in drug discovery: self-interpretable graph neural network for molecular property prediction using concept whitening},
  year = {2023},
  doi = {10.1007/s10994-023-06369-y},
  publisher = {Springer Science and Business Media {LLC}},
}

2022

Journal

A self-interpretable module for deep image classification on small data

Biagio La Rosa, Roberto Capobianco, and Daniele Nardi

Applied Intelligence, 2022

Bib DOI Official PDF Code

@article{LaRosa2022,
  author = {La Rosa, Biagio and Capobianco, Roberto and Nardi, Daniele},
  journal = {Applied Intelligence},
  title = {A self-interpretable module for deep image classification on small data},
  year = {2022},
  issn = {1573-7497},
  number = {8},
  pages = {9115--9147},
  volume = {53},
  doi = {10.1007/s10489-022-03886-6},
  publisher = {Springer Science and Business Media LLC},
}

Journal

Prototype-based Interpretable Graph Neural Networks

Alessio Ragno, Biagio La Rosa, and Roberto Capobianco

IEEE Transactions on Artificial Intelligence, 2022

Bib DOI Official PDF Code

@article{Ragno2022,
  author = {Ragno, Alessio and {La Rosa}, Biagio and Capobianco, Roberto},
  journal = {IEEE Transactions on Artificial Intelligence},
  title = {Prototype-based Interpretable Graph Neural Networks},
  year = {2022},
  volume = {},
  number = {},
  pages = {1-11},
  doi = {10.1109/TAI.2022.3222618},
}