Display Accessibility Tools

Accessibility Tools

Grayscale

Highlight Links

Change Contrast

Increase Text Size

Increase Letter Spacing

Readability Bar

Dyslexia Friendly Font

Increase Cursor Size

Department of Computational Mathematics, Science & Engineering Michigan State University Comprehensive Exam Notice

August 30, 2023, 10:00 am -12 noon EST

CMSE Conference Room 1502

Zoom link:  https://msu.zoom.us/j/99454514194

Passcode: 123456

What effects generalization in learned models?

By Avrajit Ghosh

Abstract:

The success of deep learning has been attributed to the surprising phenomenon of generalization in neural networks. Although heavily overparameterized, deep neural networks barely tend to overfit on training samples and perform well on unseen problems. In this proposal, I will first explain the role of widely used optimizers (SGD and momentum) and it’s implicit effect to find solutions that generalize well for supervised problems. Then, I will propose to study the notion of generalization for unsupervised learning problems (image reconstruction) and aim to study the determining factors that effect them. I aim to study the role of optimizers in finding low-rank/sparse and simple solutions which is critical for the success in unsupervised learning problems.

 

 

Committee:


Saiprasad Ravishankar

Rongrong Wang

Sijia Liu

John Wright