Towards Better Understanding of Deep Learning: A Perspective from Data and Algorithms

You are cordially invited to the weekly CS Machine Learning Lunch Meetings. This is a chance to get to know machine learning professors, and talk to your fellow researchers.  Our next meeting will be on Tuesday March 28 12:12-1:30pm in CS 1240. Professor Yingyu Liang will explain why deep learning works:

Abstract: Compared to the unprecedented empirical success of deep learning with neural networks, theoretical understanding largely lag behind. Under what conditions of the data and the algorithm can we achieve provable learning guarantees for neural networks? In this talk, we will discuss the unique and novel challenges in analyzing the optimization and generalization in deep learning, for which traditional tools are not adequate. We will then present our work along the line: the neural tangent kernel (NTK) approach, feature learning beyond NTK, and a recent united analysis framework. If time permits, we will also consider the recent dominating learning paradigm of representation pretraining (a.k.a., foundation models), and present our work on two key properties: label efficiency (the ability to learn an accurate model on top of the representation with a small amount of labeled data) and universality (usefulness across a wide range of downstream tasks).

If you would like to be informed of future CS Machine Learning Lunch Meetings, please sign up our mailing list at https://lists.cs.wisc.edu/mailman/listinfo/mllm -- please use your cs or wisc email.  After you enter your email, the system will send you an email for confirmation.  Only after you respond to that email will you be on the mailing list.