mixup_mxnet

This is a unofficial MXNet implementation of mixup method.

Usage

1.simple example to train cifar10

python mixup_scratch.py

2.train cifar10 with multi gpus

usage: mixup_scratch_cmd.py [-h] [--lr LR] [--batch_size BATCH_SIZE]
[--epochs EPOCHS] [--num_classes NUM_CLASSES]
[--use_mixup] [--alpha ALPHA]

mxnet CIFAR10 Training

optional arguments:

--h, --help                    show this help message and exit
--lr LR                        learning rate
--batch_size BATCH_SIZE        batch size
--epochs EPOCHS                train epoches
--num_classes NUM_CLASSES      number of classes
--use_mixup                    whether to use mixup or not
--alpha ALPHA                  Beta distributed parmas

paper: mixup: BEYOND EMPIRICAL RISK MINIMIZATION

Reference:

• [1] implementation by leehomyc with PyTorch

• [2] MXNet image classification example (particularlly train_cifar10.py)