signSGD: Compressed Optimization for Non-Convex Problems

File Structure

• FashionMNIST_logisticRegression.py: impelments signSGD on logistic Regression using Fashion MNIST dataset
• MNIST_signSGD.py: implements signSGD on neural network using MNIST Handwritten digits as dataset
• FashionMINST_signSGD.py: impements signSGD on neural network using MNIST Fashion as dataset
• hostfile: contains the maximum number of cores that our program need to be run on

How to run

• To run FashionMNIST_logisticRegression.py

python3 FashionMNIST_logisticRegression.py

• To run other two files

mpirun --hostfile <hostfile_name> -np <no_of_cores> python3 <filename>.py

Hostfile contents

localhost slots=<maximum_no_of_cores_as_required>