https://github.com/GlassyWing/gon_emb

Using fourier feature mapping Strengthen GON (《Gradient Origin Networks》)'s performance
https://github.com/GlassyWing/gon_emb

fourier-feature-network gon image-embedding implict siren

Last synced: 5 months ago
JSON representation

Using fourier feature mapping Strengthen GON (《Gradient Origin Networks》)'s performance

• Host: GitHub
• URL: https://github.com/GlassyWing/gon_emb
• Owner: GlassyWing
• License: apache-2.0
• Created: 2020-07-23T02:53:00.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2020-07-23T02:54:24.000Z (about 5 years ago)
• Last Synced: 2024-11-15T07:34:26.844Z (11 months ago)
• Topics: fourier-feature-network, gon, image-embedding, implict, siren
• Language: Python
• Homepage:
• Size: 825 KB
• Stars: 7
• Watchers: 3
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# gon_emb

Using fourier feature mapping Strengthen GON (《Gradient Origin Networks》)'s performance

## Introduction

 

The origin GON could quickly learn a latent representation by introducing a second-order derivative and SIREN.

It get a good performance on MNIST, but not good for Celeba:

|          Recon for Celeba           |

| :-----------------------: |

| ![celeba](assets/recon_celeba.png "Celeba")|

As shown in the figure above, the recon is very blurry, and it's very easy to cause gradient explosion.

Which can be considered that it cannot learn high-frequency functions.

## Solution

As mentioned earlier, current models need a mechanism to learn high-frequency functions.

Fortunately, recently proposed [《Fourier Features Let Networks Learn

High Frequency Functions in Low Dimensional Domains》](https://arxiv.org/pdf/2006.10739.pdf) Just do it.

Simply apply fourier mapping on coordinates, let's see the result:

|          Recon for Celeba after apply fourier            |

| :-----------------------: |

| ![celeba](assets/recon_celeba_fourier.png "Celeba")|

Wow! Considerable improvement!

## Conclusion

The appearance of SIREN makes it possible to calculate the second derivative, 

so the emergence of GON is natural. 

It's a good way to learning image embedding, 

and could be used with Image retrieval, face recognition, and so on.

But calculation speed and memory footprint are still an issue.