What If Only Batch Normalization Layers Were Trained?
You might be surprised, it works.
Published in
8 min readMar 25, 2020
I, for one, would never bet my money on it.
Recently, I read the paper “Training BatchNorm and Only BatchNorm: On the Expressive Power of Random Features in CNNs”, by Jonathan Frankle, David J. Schwab, and Ari S. Morcos, recently made available at the arXiv platform. The idea immediately…