Biagio Mattia La Rosa bio photo

Biagio Mattia La Rosa

I am a PhD student on Computer Engineering @ sapienza. My focus is to study methodologies to improve the interpretability of Deep Learning techniques. Passionate also about neuroscience, psychology, games, and astronomy.

Twitter LinkedIn Github e-Mail

My Publications

Pre-prints

  • Detection Accuracy for Evaluating Compositional Explanations of Units
    Sayo M. Makinwa, Biagio La Rosa, Roberto Capobianco.
    [BibTeX] [PDF] 
@inproceedings{Makinwa,
  title     = {Detection Accuracy for Evaluating Compositional Explanations of Units},
  author    = {Makinwa, Sayo M. and La Rosa, Biagio and Capobianco, Roberto},
  year      = {2021},
  url       = {https://arxiv.org/abs/2109.07804},
  eprint={2109.07804},
  archivePrefix={arXiv},
  primaryClass={cs.LG}
}
  • Memory Wrap: a Data-Efficient and Interpretable Extension to Image Classification Models.
    Biagio La Rosa, Roberto Capobianco and Daniele Nardi.
    [BibTeX] [PDF] [Code] 
@inproceedings{LaRosa20,
  title     = {Memory Wrap: a Data-Efficient and Interpretable Extension to Image Classification Models},
  author    = {La Rosa, Biagio and Capobianco, Roberto and Nardi, Daniele},
  year      = {2021},
  url       = {https://arxiv.org/abs/2106.01440},
}

2020

  • “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)
    [BibTeX] [PDF] [Code] [Blog Post].
@inproceedings{LaRosa20,
  title     = {Explainable Inference on Sequential Data via Memory-Tracking},
  author    = {La Rosa, Biagio and Capobianco, Roberto and Nardi, Daniele},
  booktitle = {Proceedings of the Twenty-Ninth International Joint Conference on
               Artificial Intelligence, {IJCAI-20}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},             
  pages     = {2006--2013},
  year      = {2020},
  month     = {7},
  doi       = {10.24963/ijcai.2020/278},
  url       = {https://doi.org/10.24963/ijcai.2020/278},
}

Theses

Master

Short Abstract: The thesis presents a novel mechanism to get hints of explanation exploiting the capability of memory-based networks – Differential Neural Computers – to store data in memory and reusing it for inference. By tracking both the memory access at prediction time, and the information stored by the network at each step of the input sequence, it is possible to retrieve the most relevant input steps associated to each prediction. The mechanism is tested on two problems: a modified version of T-Maze and the Story Cloze Test task. The work studies also the influence of parameters and the adequacy of the extracted explanations.
Area: Explainable Artificial Intelligence
Supervisor: Roberto Capobianco

Bachelor

Short Abstract: The work presents a supersense tagger based on Wikipedia and BabelNet. SuperSense tagging is a Natural Language Processing task that consists in annotating each entity in a text according to a general semantic taxonomy. In our case BabelNet is used as taxonomy, extracting the most used words and collecting the nouns that can indicate a category. Exploiting the is-a relations of BabelNet, then Wikipedia is annotated with the extracted concepts and used as dataset for the training of an SVM classifier based on the framework It Makes Sense.
Area: Natural Language Processing
Supervisor: Roberto Navigli