Introduction to Binarized Neural Networks

Explore the fundamentals of binarized neural networks in this comprehensive lecture by Prof. Gerardo I. Simari from UNS. Delve into the motivation behind BNNs, gain insights into quantization techniques and their benefits, and master core concepts essential for understanding these networks. Learn about the straight-through (ST) estimator, discover effective training methods for BNNs, and examine experimental results. Conclude with an overview of recent developments and follow-on work in the field of binarized neural networks. This informative session, part of the Neuro Symbolic Channel's offerings, bridges the gap between symbolic methods and deep learning, providing valuable knowledge for those interested in cutting-edge artificial intelligence and machine learning techniques.

Overview of lecture
Motivation for BNN's
A brief introduction to quantization
Benefits of quantization
Core concepts
The straight-through ST estimator
Training BNN's
Experimental results
Further developments survey of follow-on work on BNN's