Towards AI Security: Formal Verification of Robust Mean Estimation [Webinar]
At this webinar we unpack how robust mean estimation algorithm can improve the performance of AI classification in the scenarios of adversarial attacks.
29th November 2 - 3 PM CET
Speaker
Ieva Daukantas is a PhD at IT University of Copenhagen, Denmark, with interest in the areas of information security, AI and natural language processing. Before joining ITU, she gained 7 years of experience in the industry and the degree of MSc in software development (ITU, Denmark).
About the webinar
Trimming datasets is used as a cleaning or processing technique in many AI systems. It serves the purpose of improving the robustness of AI algorithms, so that they become less vulnerable to data related attacks. Theory states that the outliers in a data set occur with low probability, and so it follows that they can be removed without causing loss of precision in the classification result.
The talk is based on the published paper “Trimming Data Sets: a Verified Algorithm for Robust Mean Estimation” (Ieva Daukantas, Alessandro Bruni, Carsten Schürmann), where we introduce a mechanized proof of robustness of the trimmed mean algorithm. This algorithm has high applicability as a statistical technique and can be used in many complex applications of deep learning. We combine theoretical and practical approaches: the Coq proof assistant is used to formalize the robustness of the trimmed mean algorithm and Python Naïve Bayes experiments to illustrate the applicability in AI systems.
Agenda:
• The Context of the Problem
• Theory and Formal verification of Trimmed Mean Algorithm
• Empirical Examples
The talk is based on the published paper “Trimming Data Sets: a Verified Algorithm for Robust Mean Estimation” (Ieva Daukantas, Alessandro Bruni, Carsten Schürmann), where we introduce a mechanized proof of robustness of the trimmed mean algorithm. This algorithm has high applicability as a statistical technique and can be used in many complex applications of deep learning. We combine theoretical and practical approaches: the Coq proof assistant is used to formalize the robustness of the trimmed mean algorithm and Python Naïve Bayes experiments to illustrate the applicability in AI systems.
Agenda:
• The Context of the Problem
• Theory and Formal verification of Trimmed Mean Algorithm
• Empirical Examples
This webinar is a part of our Academic Corner series where we dive into the newest research from scholars within the field of IT research.
Tilmeld
Login ind for at tilmelde dig.