Room 03.124, Building 5, St Priest campus Machine Learning in Montpellier, Theory & Practice The Procrustes-Wasserstein problem consists in matching two high-dimensional point clouds in an unsupervised setting, and has applications in natural language processing and computer vision. This talk will first introduce and motivate this problem, before considering a planted model with two random datasets $X,Y$ that consist of $n$ datapoints in $\R^d$, where $Y$ is a noisy version of $X$, up to an orthogonal transformation and a relabeling of the data points. This setting is related to the graph alignment problem in geometric models as we will show. Focusing on the Euclidean transport cost between the point clouds as a measure of performance for the alignment, we first establish information-theoretic results, in the high ($d \gg \log n$) and low ($d \ll \log n$) dimensional regimes, and provide geometrical and probabilistic insights to explain the dichotomy between these two regimes. We then study computational aspects and propose intuitive algorithms to approximate solutions for this problem, alternatively estimating the orthogonal transformation and the relabeling, initialized via a convex relaxation. Machine Learning in Montpellier, Theory & Practice

Room 02.022, Building 5, St Priest campus Machine Learning in Montpellier, Theory & Practice In this talk, we will focus on an important security concern in modern day machine learning, namely adversarial example attacks. The vulnerability of state-of-the-art models to these attacks has genuine security implications especially when models are used in AI-driven technologies, e.g., for self-driving cars or fraud detection. Besides security issues, these attacks show how little we know about the models used every day in the industry, and how little control we have over them. The problem of adversarial example attacks is still open and constitutes an active area of research. We will provide some insights allowing to navigate through this research field essentially presenting the current state-of-knowledge on how these attacks work, and how to mitigate them by using some notions of learning theory and optimization. Machine Learning in Montpellier, Theory & Practice