ffnn_implementation

Build for course assignment at university of Freiburg: GitHub Page

Installation

1. Clone repository
2. install requirements (! use python 2.7)

pip install -r requirements.txt

Run on MNIST dataset

1. Open python editor (e.x. jupyter notebook)
2. Import run_mnist.py
3. Train neural network
4. Compute test error and show images that are classified right and wrong

import run_mnist as r

nn = r.train_mnist(0.7,100,100)

r.test_mnist(nn)

Check gradients of neural network

1. Open python editor (e.x. jupyter notebook)
2. Import run_gradient_checking.py

import run_gradient_checking