timlautk / BCD-for-DNNs

Training Deep Neural Networks using Block Coordinate Descent Algorithms

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

BCD-for-DNNs

Codes for An Accelerated Block Coordinate Descent (BCD) Algorithm for Training Deep Neural Networks

For the BCD algorithms, MATLAB is used. Use MATLAB R2017a or later versions.

For the back propagation algorithms, Python (Jupyter Notebook) is used. Use TensorFlow 1.4 or later versions since Keras is included as a module in native TensorFlow 1.4 or later.

Two common datasets are used: MNIST and CIFAR-10

Download the CIFAR-10 dataset from: https://www.cs.toronto.edu/~kriz/cifar.html

About

Training Deep Neural Networks using Block Coordinate Descent Algorithms


Languages

Language:Jupyter Notebook 88.0%Language:MATLAB 11.9%Language:M 0.1%