{"id":19287861,"url":"https://github.com/deeppavlov/infinite-width_nets","last_synced_at":"2026-05-18T05:07:52.455Z","repository":{"id":83450956,"uuid":"246790295","full_name":"deeppavlov/infinite-width_nets","owner":"deeppavlov","description":null,"archived":false,"fork":false,"pushed_at":"2020-06-05T14:05:23.000Z","size":73176,"stargazers_count":4,"open_issues_count":0,"forks_count":0,"subscribers_count":4,"default_branch":"camera_ready_code","last_synced_at":"2025-01-05T19:22:20.503Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/deeppavlov.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2020-03-12T09:11:10.000Z","updated_at":"2023-09-22T02:59:26.000Z","dependencies_parsed_at":null,"dependency_job_id":"d3826665-f6fb-4ec7-9d19-fa717810f524","html_url":"https://github.com/deeppavlov/infinite-width_nets","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/deeppavlov%2Finfinite-width_nets","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/deeppavlov%2Finfinite-width_nets/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/deeppavlov%2Finfinite-width_nets/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/deeppavlov%2Finfinite-width_nets/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/deeppavlov","download_url":"https://codeload.github.com/deeppavlov/infinite-width_nets/tar.gz/refs/heads/camera_ready_code","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":240392607,"owners_count":19794125,"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":[],"created_at":"2024-11-09T22:07:26.678Z","updated_at":"2026-05-18T05:07:47.419Z","avatar_url":"https://github.com/deeppavlov.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"## Towards a General Theory of Infinite-Width Limits of Neural Classifiers\n\nThis is the code to reproduce results of the paper *\"Towards a General Theory of Infinite-Width Limits of Neural Classifiers\"* \nsubmitted to ICML'20.\n\n### Instructions:\n\nFirst, perform computations:\n* Main body:\n  * Figure 1:\n    * Left and center:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n    * Right:\n    ```\n    $ python fcnet_epoch_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5  \n    ```\n  * Figure 2:\n    * Left:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=2 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n    * Center:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=3 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n    * Right:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=3 --optimizer=rmsprop --lr=0.0002 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n* Supplementary material:\n  * Figure 1:\n    * All plots:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n  * Figure 2:\n    * All plots:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n  * Figure 3:\n    * Top row, left and right:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n    * Top row, center:\n    ```\n    $ python fcnet_epoch_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5  \n    ```\n    * Bottom row, left and right:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.0002 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=50\n    ```\n    * Bottom row, center:\n    ```\n    $ python fcnet_epoch_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.0002 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5  \n    ```\n  * Figure 4:\n    * Top row:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n    * Second row:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=5 --dataset=cifar2_binary --train_size=1000 --batch_size=100 --activation=lrelu --num_seeds=5\n    ```\n    * Third row:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=5 --dataset=cifar2_binary --train_size=10000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n    * Bottom row:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=500 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n  * Figure 5:\n    * Top row:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=cifar2_binary --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n    * Middle row:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=50 --dataset=mnist --train_size=1000 --batch_size=1000 --activation=lrelu --num_seeds=5\n    ```\n    * Bottom row:\n    ```\n    $ python fcnet_width_dependence.py --device=\u003ccuda:0 or cpu\u003e --num_hidden=1 --optimizer=sgd --lr=0.02 --num_epochs=10 --dataset=mnist --train_size=60000 --batch_size=100 --activation=lrelu --num_seeds=5\n    ```\n   \nAfter performing computations, make plots by running cells in notebooks: [width_dependence_plots.ipynb](/width_dependence_plots.ipynb) and [epoch_dependence_plots.ipynb](/epoch_dependence_plots.ipynb).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fdeeppavlov%2Finfinite-width_nets","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fdeeppavlov%2Finfinite-width_nets","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fdeeppavlov%2Finfinite-width_nets/lists"}