Control and Machine Learning by E. Zuazua

On Thursday July 21, 2022 our Head Prof. Enrique Zuazua will talk on “Control and Machine Learning” as invited speaker at the Workshop on Algorithmic Optimization and Data Science from July 20 to July 22th, 2022 at the Trier University.

In this lecture we shall present some recent results on the interplay between control and Machine Learning, and more precisely, Supervised Learning and Universal Approximation.
We adopt the perspective of the simultaneous or ensemble control of systems of Residual Neural Networks (ResNets). Roughly, each item to be classified corresponds to a different initial datum for the Cauchy problem of the ResNets, leading to an ensemble of solutions to be driven to the corresponding targets, associated to the labels, by means of the same control.

We present a genuinely nonlinear and constructive method, allowing to show that such an ambitious goal can be achieved, estimating the complexity of the control strategies.
This property is rarely fulfilled by the classical dynamical systems in Mechanics and the very nonlinear nature of the activation function governing the ResNet dynamics plays a determinant role. It allows deforming half of the phase space while the other half remains invariant, a property that classical models in mechanics do not fulfill.

The turnpike property is also analyzed in this context, showing that a suitable choice of the cost functional used to train the ResNet leads to more stable and robust dynamics.

This lecture is inspired in joint work, among others, with Borjan Geshkovski (MIT), Carlos Esteve (Cambridge), Domenec Ruiz-Balet (IC, London) and Dario Pighin (Sherpa.ai)

WHERE

On-site & online (Zoom)
Lecture Hall HS 9.
Universität Trier. E. Building
Universitätsring 15. 54296 Trier, Germany

Optimization algorithms are a key part of almost every solution approach in data science. On the other hand, data science or machine learning concepts can be utilized as components of optimization methods to improve state-of-the-art optimization technologies. These viewpoints will be addressed in lively discussions during the workshop, which will bring together experts in the fields of data science, machine learning, algorithmic optimization, and everything related to these fields.

Fees

Fees apply: Online (10€) and On-site (50€ + includes refreshments, the conference dinner complete with a wine tasting, and a city tour of Trier)

Program & Book of Abstract of the event

Check all details at the official page of the event

The event is finished.

No Responses

© 2019-2022 FAU DCN-AvH Chair for Dynamics, Control and Numerics - Alexander von Humboldt Professorship at Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany | Imprint | Contact