Skip to content

aditya-pola/Gradient_Matching

Repository files navigation

All the tests are contained within "all_the_tests.ipynb"

naive_testing.py is now complete. When run, naive_testing finds sunset of size 1000 after every epoch of training the model on full dataset and reports test accuracy metrics for the top_1000 (cosine based), random subset and full dataset models.

Location of saving data needs to be changed in the variables "data" and "test_data" for local usage.

ToDo: [] Store the indices of top 1000 subset and see which samples are recurring

About

Hypothesis testing with gradient matching

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published