/miniflow

Recreation of core tensorflow functionality: linear and sigmoid layers, forward and back prop, stochastic gradient descent

Primary LanguagePython

Miniflow: a recreation of the core functionality of tensorflow

Overview

Create layers, sigmoid activation nodes, forward prop, backward prop, and Stochastic Gradient Descent implementations. Much of the starter code given via Udacity's Self-driving Car nanodegree. I implemented key pieces for each of the above-mentioned pieces.

Files

  1. miniflow.py - Implementation of the layers, forward and back prop, stochastic gradient descent.
  2. nn.py - Creates a layer and tests that stochastic gradient descent works as expected (with gradients calculated via correct back prop)