Systematic Generalization with Deep Equilibrium Models

Topic: Systematic Generalization with Deep Equilibrium Models

Time: 17h00, Thursday, March 10th, 2022

Speaker: Dr. Hoang Thanh Tung, Research Scientist – FPT Software

📌 Location: MS Teams.

✅ Link: MS Teams


Deep Equilibrium models (DEQs) are an emerging class of models that have high performance and small memory footprints.
They can be thought of as infinitely deep networks that compute infinite level representations. Empirical evidence shows that compared to traditional neural networks, DEQs produce more robust representations and have better out-of-distribution generalization performance.
In this talk, we review some of the most recent advances and discuss some open problems of DEQs.

Speaker Bio:

Hoang Thanh Tung is a research scientist at FPT Software. He got his PhD in Computer Science from Deakin University in 2022. His research focuses on the theory and applications of Deep Learning. Currently, he is working on Deep Generative Models and Implicit Neural Networks.

Slides: Systematic Generalization Fixed Points are all you need

Recording: here