A feed-forward neural network library that uses the computational graph approach to compute the gradients
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
datasets
doc
LICENSE
README.md
basic_driver.py
driver.py
layers.py
loss.py
network.py
optimizers.py
solver.py
util.py

README.md

Dense-Net

A feed-forward neural network library that uses the computational graph approach to compute the gradients. This library supports ANNs of arbitrary size as defined by the user.

Computational Graph

Getting Started:

Prerequisites:

This implementation makes use of just Python and Numpy. Matplotlib was used for testing the network and plotting graphs to observe it's learning.

Activation functions:

  • Sigmoid
  • ReLU
  • Softmax
  • Linear (No activation, only linear transform)

Loss functions:

  • L1 Loss
  • L2 Loss
  • Cross Entropy
  • SVM Loss

Optimizers:

  • SGD
  • Momentum

Computational Graph

Contributing:

When contributing to this repository, please first discuss the change you wish to make via issue, email, or any other method with the owners of this repository before making a change. Ensure any install or build dependencies are removed before the end of the layer when doing a build. Update the README.md with details of changes to the interface, this includes new environment variables, exposed ports, useful file locations and container parameters.

License:

This project is licensed under the MIT License - see the LICENSE.md file for details

(Computational graph image source: https://colah.github.io/)