-
Notifications
You must be signed in to change notification settings - Fork 1
Home
tm-schwartz edited this page Oct 16, 2020
·
5 revisions
The main goal of this project is to help people gain a mental model of what is happening during model training, and to visualize/verify how the parameter/architectural changes they make impact their dl model.
This technique is unique in that it allows anyone to visualize their model, not just a toy example or by plugging values into a loss function.
This wiki (hopefully) be updated over time. Right now there is not much here besides a basic explanation of the project.
If you create an interesting visual, feel free to share it and I'll credit you if I post it!
This project is built based off of the fantastic research and implementation found at https://github.com/tomgoldstein/loss-landscape