Skip to content

Latest commit

 

History

History
54 lines (35 loc) · 1.9 KB

README.md

File metadata and controls

54 lines (35 loc) · 1.9 KB

Weight_Standardization

Weight Standardization Analysis with Keras

Weight standardization is a new regularization technique that can outperform Batch Normalisation on micro batches, while reaching the same performance on large batches.

The original research paper : Siyuan Qiao Huiyu Wang Chenxi Liu Wei Shen Alan Yuille : “Weight Standardization” (2019) https://arxiv.org/pdf/1903.10520v1.pdf

Experiments

We tested the performance of Group Normalization (GN) + Weight Standardization (WS) against Batch Normalization (BN) with a multiclass classifier on the MNIST-FASHION dataset.

We are using Python 3.6 and the Keras library.

The code for the original classifier is this official Google Example : https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/tutorials/keras/basic_classification.ipynb

To implement GN we used the code available at this repo : https://github.com/titu1994/Keras-Group-Normalization

To implement WS we used a custom keras kernel regulizer funcitons :

def ws_reg(kernel):
  kernel_mean = tf.math.reduce_mean(kernel, axis=[0, 1, 2], keepdims=True, name='kernel_mean')
  kernel = kernel - kernel_mean
  kernel_std = tf.keras.backend.std(kernel, axis=[0, 1, 2], keepdims=True)
  kernel = kernel / (kernel_std + 1e-5)
  return kernel

Results

Final accuracy per size and model :

Batch size 2 64 128
BN 18.78 92.49 92.93
GN+WS 91.86 92.82 92.9

Files

  • "/src" : source code, the python notebook is meant to be load in colab, with your drive mounted locallt? The python code will run for a default batch size of 128 and 10 apochs.

  • "weight_standardization_analysis.pdf" : detailed report of the project.

  • "/data" : model architectures, model weights, model histories, and plots for several dfferent parameters.

This project was made by Clement Rouault and Thomas Ehling