Optimizing Neural Networks with Gradient Lexicase Selection
Optimizing Neural Networks with Gradient Lexicase Selection, Ding & Spector, ICLR 2022.
Basic usage:
- To train and evaluate baseline architectures, do
python3 base.py
- To train and evaluate gradient lexicase selection, do
python3 lexi.py
Also use the --help flag to see instructions for optional arguments to configure architecture and dataset:
-
The architectures are indexed as follows (from 0 to 6): VGG, ResNet18, ResNet50, DenseNet121, MobileNetV2, SENet18, EfficientNetB0
-
The datasets are specified as follows: 'C10' - CIFAR-10, 'C100' - CIFAR-100, 'SVHN' - SVHN
Please contact the authors for further questions.