README.md

In ifrit98/R2deepR:

R2deepR

R2deepR is an R package designed to make the lives of Deep Learning practitioners a little more conveneint.

Features

Feature Scaling

Standardization

Minmax Normalization

Unit Vector Scaling

Custom callbacks

• lr_history
• save_optimizer_weights

Fully Preprocessed keras datasets

Convenience functions to return fully preprocessed keras datasets as either numpy arrays or tensorflow-dataset objects. - mnist - imdb

Novel activation functions

• gelu
• brelu
• belu

- nalu

Novel layers

• Residual
• GLU
• PRU
• Antirectifier
• Autoencoder

Learning Curve Plots

• training set size
• batch size
• learning rate

Learning rate range test (see: Smith et al.)

One Cycle Learning Rate Policy (see: Smith et al.)

Evalation tools

• F1 score for binary classifiers

Assorted utility and convenience functions

Weight activation curves (WIP)

ifrit98/R2deepR documentation built on June 19, 2020, 7:45 a.m.