Naive recreation of the core functionality of TensorFlow
- miniflow.py: Implementation of:
- The different nodes (Input, Linear Transform, Sigmoid Activation Function and MSE) of the Computational Graph
- Thin wrapper to ensure the connectivity of the nodes and edges
- Forward and backward propgation mechanisms, and,
- Stochastic Gradient Descent
- nn.py: Creates a 2 layer NN and trains a Linear Regression Model over the Boston Housing dataset using backprop and stochastic gradient descent to predict housing prices.