Lifelong Visual Episodic Memory
Abstract
This report describes work on Lifelong Learning that develops methods for using Generative Adversarial Networks (GANs) to represent the probability of images. This problem is integral to developing the ability to detect and characterize domain shift. It also explores causality in video. Causal reasoning may play a key role in the ability of an agent to adapt to changing environments. If one understands what components of the environment lead to a particular outcome, one can determine whether changes to the environment affect these causal factors and should affect expected outcomes. It also explores semantic segmentation in the presence of domain shift.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 15, 2020
- Accession Number
- AD1108933
Entities
People
- David R. Jacobs
- Ronen Basri
- Tom Goldstein
Organizations
- University of Maryland
- Weizmann Institute of Science