Loading Events

« All Events

  • This event has passed.

DiSIA – ARTES4.0 Webinar

March 27, 2023 @ 16:00 - 17:00

The Department of Statistics, Computer Science, Applications DiSIA as part of the activities of the UNIFI Macronode in the ARTES4.0 Competence Center organizes the following webinar scheduled for March 27, 2023 at 4 pm.

Webinar DISIA#3: Explainable Artificial Intelligences;
Title: Counterfactual Explanations of (some) Machine Learning Models
Speaker: Prof. FABRIZIO SILVESTRI,
Language: Italian
For those interested in participating, please  register at the link:  https://lp.artes4.it/it/ webinar-27-marzo-2023
Webinar contents:
Start 16.00: Introduction
16.05 – Talk
16.45- Discussion, questions with participants
17.00 – Conclusion
The third webinar promoted by DISIA focuses on Machine Learning (ML), in the context of Explicable Artificial Intelligence understood as a set of methodologies that allow a human to understand and trust the results produced by an ML algorithm. Such methodologies are used to describe the AI ​​model in terms of expected impact and potential bias in the decision-making processes in which an ML algorithm is integrated. These aspects are crucial to ensure that decision-making processes are impartial, transparent and accurate.
Abstract : In this presentation we will review some of the most recent contributions in the field of Explainable Artificial Intelligence (XAI) and, in particular, those of counterfactual (CF) explanations. We will start by reviewing our first contribution to the field and show some recent papers on CF explanations of graph neural network models. Guided by the interest of generalizing the CF explainability of ML models as much as possible, we propose a solution based on reinforcement learning that treats the model as a black box that we cannot access directly.
Brief Biography of Fabrizio Silvestri: Fabrizio Silvestri is a full professor at the DIAG of the Sapienza University of Rome. His research interests concern artificial intelligence and in particular machine learning applied to web search and natural language processing problems. He is the author of more than 150 articles in international journals and conference proceedings. He holds nine industrial patents. At Facebook AI, Fabrizio Silvestri led research groups to develop artificial intelligence techniques to combat malicious actors who use the Facebook platform for malicious purposes (hate speech, disinformation, terrorism, etc.). Fabrizio Silvestri holds a PhD in computer science from the University of Pisa.

Details

Date:
March 27, 2023
Time:
16:00 - 17:00
Event Category:

Organizers

DISIA
ARTES4.0

Venue

Online