Inaugural Lecture by Andrea Paudice
Info about event
Time
Location
INCUBA Lille Aud., building 5510-104, Aabogade 15, 8200 Aarhus N
Title: From Gradient Descent to Learning from Data
Abstract
The transition from deterministic to stochastic optimization plays a central role in modern machine learning. Yet, despite the apparent simplicity of the underlying algorithms, understanding their behavior raises surprisingly subtle mathematical questions, especially when the information guiding their updates is noisy.
In this lecture, we will take a guided tour from gradient descent to learning from data, exploring some fundamental questions along the way. How fast do these algorithms converge? How good is their final iterate? How do randomness and noise affect their reliability?
Drawing on classical insights and recent research, we will illustrate how these questions continue to shape our understanding of optimization algorithms and their role in machine learning.
----------------
Everyone is welcome!
There will be refreshments after the lecture.