{"id":13738422,"url":"https://github.com/GlassyWing/gon_emb","last_synced_at":"2025-05-08T16:33:56.281Z","repository":{"id":169668181,"uuid":"281834717","full_name":"GlassyWing/gon_emb","owner":"GlassyWing","description":"Using fourier feature mapping Strengthen GON (《Gradient Origin Networks》)'s performance","archived":false,"fork":false,"pushed_at":"2020-07-23T02:54:24.000Z","size":845,"stargazers_count":7,"open_issues_count":0,"forks_count":1,"subscribers_count":3,"default_branch":"master","last_synced_at":"2024-11-15T07:34:26.844Z","etag":null,"topics":["fourier-feature-network","gon","image-embedding","implict","siren"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/GlassyWing.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null}},"created_at":"2020-07-23T02:53:00.000Z","updated_at":"2024-04-17T10:59:28.000Z","dependencies_parsed_at":"2024-01-07T17:11:06.643Z","dependency_job_id":"3747fac3-7a0d-42f0-9eb5-bd8dba8f0b5c","html_url":"https://github.com/GlassyWing/gon_emb","commit_stats":null,"previous_names":["glassywing/gon_emb"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/GlassyWing%2Fgon_emb","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/GlassyWing%2Fgon_emb/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/GlassyWing%2Fgon_emb/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/GlassyWing%2Fgon_emb/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/GlassyWing","download_url":"https://codeload.github.com/GlassyWing/gon_emb/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":253105565,"owners_count":21855056,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["fourier-feature-network","gon","image-embedding","implict","siren"],"created_at":"2024-08-03T03:02:22.036Z","updated_at":"2025-05-08T16:33:55.950Z","avatar_url":"https://github.com/GlassyWing.png","language":"Python","funding_links":[],"categories":["Python"],"sub_categories":[],"readme":"# gon_emb\n\nUsing fourier feature mapping Strengthen GON (《Gradient Origin Networks》)'s performance\n\n## Introduction\n \nThe origin GON could quickly learn a latent representation by introducing a second-order derivative and SIREN.\n\nIt get a good performance on MNIST, but not good for Celeba:\n\n\n|          Recon for Celeba           |\n| :-----------------------: |\n| ![celeba](assets/recon_celeba.png \"Celeba\")|\n\nAs shown in the figure above, the recon is very blurry, and it's very easy to cause gradient explosion.\nWhich can be considered that it cannot learn high-frequency functions.\n\n## Solution\n\nAs mentioned earlier, current models need a mechanism to learn high-frequency functions.\nFortunately, recently proposed [《Fourier Features Let Networks Learn\nHigh Frequency Functions in Low Dimensional Domains》](https://arxiv.org/pdf/2006.10739.pdf) Just do it.\n\nSimply apply fourier mapping on coordinates, let's see the result:\n\n\n|          Recon for Celeba after apply fourier            |\n| :-----------------------: |\n| ![celeba](assets/recon_celeba_fourier.png \"Celeba\")|\n\nWow! Considerable improvement!\n\n## Conclusion\n\nThe appearance of SIREN makes it possible to calculate the second derivative, \nso the emergence of GON is natural. \nIt's a good way to learning image embedding, \nand could be used with Image retrieval, face recognition, and so on.\nBut calculation speed and memory footprint are still an issue.\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FGlassyWing%2Fgon_emb","html_url":"https://awesome.ecosyste.ms/projects/github.com%2FGlassyWing%2Fgon_emb","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FGlassyWing%2Fgon_emb/lists"}