Deep Networks with Stochastic Depth

Rewriting code for "Deep Networks with Stochastic Depth" in Keras.

Original code is at yueatsprograms/Stochastic_Depth.

Fork by @aaamourao : adriano-dev branch

mourao.aaa@gmail.com

Encrypted mail: adrianomourao@protonmail.com

Usage

System requirements

You need to install the following libraries:

• Developer version of libblas
• Developer version of liblapack
• Developer version of libatlas
• GNU Fortran

Fedora install instructions

sudo dnf install blas blas-devel lapack lapack-devel atlas atlas-devel gcc-fortran

Ubuntu install instructions

sudo apt-get install libblas-dev liblapack-dev libatlas-base-dev gfortran

[optional] Create a virtual environment

Run the following commands and replace <virtual-dir-path> for your target directory on your workspace

virtualenv --distribute <virtual-dir-path>
source <virtual-dir-path>/bin/activate

Install python requirements

There are some python libraries required to run this project

pip install -r requirements.txt

Hint : if pip doesn't find Keras library install it manually:

pip install -r requirements.txt```

Install python libraries from repo

Just run python train.py

Results

• Initial learning rate == 0.1
• Number of layers == 50
• (other configs are same as train.py)