Gradient Origin Networks - Paper Explained with Live Coding

Explore the concept of Gradient Origin Networks in this comprehensive video lecture. Dive into the world of implicit representations and learn how this innovative approach extends to entire datasets. Discover how latent vectors can be obtained without an explicit encoder by examining the negative gradient of the zero-vector through the representation function. Follow along as the presenter breaks down the paper, explains key concepts, and demonstrates live coding examples. Gain insights into implicit generative models, dataset representation, and the relationship to gradient descent. Explore practical applications, including using GONs as classifiers, and review experimental results. This in-depth tutorial covers everything from the basics to advanced topics, making it suitable for both beginners and experienced practitioners in the field of neural networks and machine learning.

- Intro & Overview
- Implicit Generative Models
- Implicitly Represent a Dataset
- Gradient Origin Networks
- Relation to Gradient Descent
- Messing with their Code
- Implicit Encoders
- Using GONs as classifiers
- Experiments & Conclusion