Skip to content

Differentially private training of residual networks with scale normalization

License

Notifications You must be signed in to change notification settings

mrchntia/ScaleNorm

Repository files navigation

Differentially Private Training of Residual Networks with Scale Normalization

Exploring the training of residual networks under differential privacy (DP). Development of an architectural adaptation for residual networks that increases the validation accuracy when trained with DP called ScaleNorm.

This is a selection of scripts written for my Master's Thesis. I conducted this work at the Medicine in AI at the Technical University of Munich. This paper is a short version of the thesis.

About

Differentially private training of residual networks with scale normalization

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published