Gradient X-Ray
by Javier Ideami
Title
Gradient X-Ray
Artist
Javier Ideami
Medium
Digital Art - Digital
Description
Mode Connectivity. Optima of complex loss functions connected by simple curves over which training and test accuracy are nearly constant. Visualization data generated through a collaboration between Pavel Izmailov (@Pavel_Izmailov), Timur Garipov (@tim_garipov) and Javier Ideami (@ideami). Based on the NeurIPS 2018 paper by Timur Garipov, Pavel Izmailov, Dmitrii Podoprikhin, Dmitry Vetrov, Andrew Gordon Wilson: https://arxiv.org/abs/1802.10026 | Creative visualization and artwork produced by Javier Ideami.
This artwork shows the area of the two minima and also higher loss areas which contain irregular morphology.
Real data, resnet-20 no-skip, imagenette, sgd-mom, bs=128wd=3e-4, mom=0.9, bn, train mod, 1 million pts, log scaled (orig loss nums)
This piece has been created through a multi day process composed of different stages that begin with the training of deep learning neural networks and proceeds through a number of phases until arriving to the final artwork.
Final artwork created by Javier Ideami.
For more about the Loss Landscape project visit https://losslandscape.com
Uploaded
November 5th, 2019
Statistics
Viewed 482 Times - Last Visitor from Fairfield, CT on 04/19/2024 at 1:20 PM
Embed
Share
Sales Sheet
Comments
There are no comments for Gradient X-Ray. Click here to post the first comment.