Code for reproducing the results in the following paper:

Decorrelated Batch Normalization
Lei Huang, Dawei Yang, Bo Lang, Jia Deng
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018. arXiv:1804.08450

• Install MAGMA (you can find the instructions in  'Install MAGMA.md' ). Note: MAGMA is required for SVD on GPU. Without MAGMA, you can run the code on CPU only, while all the CNN experiments in the paper are run on GPU.
• Install Torch with CUDA (for GPU). Note that cutorch should be compiled with MAGMA support if you have installed MAGMA and set the environments correctly.
• Install cudnn v5.
• Install the dependency optnet by:

luarocks install optnet

• Run:

bash execute_MLP_0debug_MNIST.sh

This script will download MNIST automatically and you should put the mnist.t7/ under ./dataset/. The experiment results will be saved at ./set_result/MLP/.

• Prepare the data: download the YaleB dataset here, and put the data files under /dataset/ so that the paths look like ./dataset/YaleB/YaleB_train.dat and ./dataset/YaleB/YaleB_test.dat.
• Run:

bash execute_MLP_1FIM_YaleB_best.sh

The experiment results will be saved at directory: 'set_result/MLP/'.

You can experiment with different hyperparameters by running these scripts -- execute_MLP_1FIM_YaleB_HyperP.sh and execute_MLP_1FIM_YaleB_HyperP_nnn.sh.

• Prepare the data: download the PIE dataset here, and put the data file under ./dataset/ such that the paths look like ./dataset/PIE/PIE_train.dat and ./dataset/PIE/PIE_test.dat.
• To experiment with different group sizes, run:

bash execute_MLP_2PIE_DBNGroup.sh

• To obtain different baseline performances, execute:

 bash execute_MLP_2PIE.sh
 bash execute_MLP_2PIE_nnn.sh

Note that the experiments until this point can be run on CPU, so MAGMA is not needed in above experiments.

• Prepare the data: follow the instructions for CIFAR-10 in this project . It will generate a preprocessed dataset and save a 1400MB file. Put this file cifar_provider.t7 under ./dataset/.
• Run:

bash execute_Conv_1vggA_2test_adam.sh
bash execute_Conv_1vggA_2test_base.sh
bash execute_Conv_1vggA_2test_ELU.sh
bash execute_Conv_1vggA_2test_var.sh

Note that if your machine has fewer than 4 GPUs, the environment variable CUDA_VISIBLE_DEVICES should be changed accordingly.

• Prepare the data: same as in VGG-A experiments.
• Run:

bash exp_Conv_4Splain_1deep.lua
bash exp_Conv_4Splain_2large.lua

• Prepare the data: download CIFAR-10 and CIFAR-100, and put the data files under ./dataset/.
• Run:

bash execute_Conv_2residual_old.sh
bash execute_Conv_3residual_wide_Cifar100_wr_BN_d28_h48_g16_b128_dr0.3_s1_C2.sh
bash execute_Conv_3residual_wide_Cifar100_wr_DBN_scale_L1_d28_h48_g16_b128_dr0.3_s1_C3.sh
bash execute_Conv_3residual_wide_Cifar10_wr_BN_d28_h48_g16_b128_dr0.3_s1_C2.sh
bash execute_Conv_3residual_wide_Cifar10_wr_DBN_scale_L1_d28_h48_g16_b128_dr0.3_s1_C3.sh

• Clone Facebook's ResNet repo here.
• Download ImageNet and put it in: /tmp/dataset/ImageNet/ (you can also customize the path in opts.lua)
• Install the DBN module to Torch as a Lua package: go to the directory ./models/imagenet/cuSpatialDBN/ and run luarocks make cudbn-1.0-0.rockspec.
• Copy the model definitions in ./models/imagenet/ (resnet_BN.lua, resnet_DBN_scale_L1.lua and init.lua) to ./models directory in the cloned repo fb.resnet.torch, for reproducing the results reported in the paper. You also can compare the pre-activation version of residual networks introduced in the paper (using the model files preresnet_BN.lua and preresnet_DBN_scale_L1.lua).
• Use the default configuration and our models to run experiments.

Email: [email protected]. Any discussions and suggestions are welcome!