This page organizes papers related to artificial intelligence published around the world. This page is summarized using Google Gemini and is operated on a non-profit basis. The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.
Massimiliano Ciranni, Luca Molinaro, Carlo Alberto Barbano, Attilio Fiandrotti, Vittorio Murino, Vito Paolo Pastore, Enzo Tartaglione
Outline
This paper introduces "Say My Name" (SaMyNa), a text-based tool for identifying bias in deep learning models. SaMyNa semantically analyzes the biases learned by the model, enhancing its explainability for non-experts. This provides semantic information about biased features that conventional bias removal methods cannot. SaMyNa can be applied both during training and post-validation, separating task-related information from biases for use in model diagnostics.
Takeaways, Limitations
•
Takeaways:
◦
Presenting the first tool to semantically identify bias in deep learning models.
◦
Analyze the model's bias in an explanatory manner so that even non-experts can understand it.
◦
Flexibility applicable to both training and post-validation.
◦
Demonstrated effective performance in bias detection and removal.
◦
Presenting a wide range of applicability for model diagnostics.
•
Limitations:
◦
Lack of description of specific bias removal methodology.
◦
Further research is needed to determine applicability in real-world environments beyond benchmark evaluations.
◦
Further validation of the tool's accuracy and effectiveness is needed.
