readme

To run the code,

prerequisites are torch>1.0 and numpy

##########. INSTALLATIONS We assume that the latest version of anaconda is installed.

Create a virtual environment torch which can be accessed by source activate torch We use the following instructions to install and setup our environment.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% conda create --name torch python=3.6 source activate torch source torch/bin/activate pip install numpy scipy cython matplotlib pandas ipython seaborn jupyter Jinja2 pip install sklearn conda install pytorch torchvision -c pytorch %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

##########. EXECUTION Execute any scripts with sh in the scripts folder. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% For example: sh scripts/permuted_MNIST_incremental_class.sh

would execute all the methods for permuted Mnist case. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%