Invariance, Geometry and Deep Neural Networks with Pavan Turaga - #386
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
00:00
Designing Robust Loss Functions in Deep Learning
This chapter examines the intricate design of loss functions and constraints in deep learning networks, particularly in visual and time series tasks. It underscores the importance of integrating mathematical principles like Riemannian geometry and topology for enhanced problem-solving while addressing the limitations of data reliance. The discussion also emphasizes prioritizing robustness and repeatability in machine learning models to navigate challenges such as one-shot and few-shot learning.
Play episode from 36:21
Transcript
