GPGL

epoch-evolving Gaussian process guided learning

Citing

@ARTICLE{9779793,
author={Jiabao Cui and Xuewei Li and Bin Li and Hanbin Zhao and Bourahla Omar and Xi Li},
journal={IEEE Transactions on Neural Networks and Learning Systems},
title={Epoch-Evolving Gaussian Process Guided Learning for Classification},
year={2022},
volume={},
number={},
pages={1-12},
doi={10.1109/TNNLS.2022.3174207}}

Dependencies

The project is built to work using MindSpore and MindSpore Vision.

Train and Test

on Cifar10 Dataset

Open folder move_cifar10_ms and run ms1_new.py script.

Generated files:

train_err_2GP_err_32.txt records AVG_TRAIN_LOSS3 of each epoch.
train_err_2GP_loss_32.txt records AVG_Top1_error3 of each epoch.
test_err_2GP_err_32.txt records OUT_TEST_Top1_error3 every 200 steps.
test_err_2GP_loss1_32.txt records test loss every 200 steps.

on Cifar100 Dataset

Open folder move_cifar100_ms and run ms_top5_test.py script.

Output:

train_batch_total 391
test_batch_total 79
val_batch_total 100
lr= 0.1

Epoch: 0
build_val_time:  5.165887355804443

Test:  0 i_num:  0
OUT_TEST_Top1_error3:  0.99011075

Test:  0 i_num:  200
OUT_TEST_Top1_error3:  0.93047863
-------------------
AVG_TRAIN_LOSS3:  1.9875456
AVG_Top1_error3:  0.46451208
this_train_epoch_time:  83.14549994468689
Saving..
this_epoch:  83.15923070907593
total:  83.1623797416687 



lr= 0.1

Epoch: 1
build_val_time:  2.8037099838256836
...

Generated files (see move_cifar100_ms/logs/):

train_err_2GP_err_32.txt records AVG_TRAIN_LOSS3 of each epoch.
train_err_2GP_loss_32.txt records AVG_Top1_error3 of each epoch.
test_err_2GP_err_32.txt records OUT_TEST_Top1_error3 every 200 steps.
test_err_2GP_loss1_32.txt records test loss every 200 steps.

Checkpoints

See move_cifar10_ms/checkpoint/cifar10_32/ or move_cifar100_ms/checkpoint/5/. For Cifar10, ckpt_2_155.ckpt performs best; for Cifar100, ckpt_top5_res20_102.ckpt performs best.