ml-neural-net-constructor

Usage

1. Install package from source code (not published to PyPi)

   $ git clone https://github.com/IrinaTerlizhenko/ml-neural-net-constructor.git
   $ cd ml-neural-net-constructor/src
   $ pip install .

2. Use it in your code

   from netconstructor.network import NeuralNetwork
   net = NeuralNetwork()\
      .with_dense_layer(...)\
      .with_logistic_activation()\
      .with_square_error()
   net.train(x, y, num_iterations)
   error = net.test(x_test)

The sample can be found here. numpy and tensorflow are required in order to run this sample (tensorflow is used only to load MNIST data). They can be installed separately or with pip install -r requirements/requirements_dev.txt (note: this also installs a bunch of other dev dependencies).

Benchmark results

Available on this page.

Development docs

1. Technical task
2. Work done

Usage from source code

1. Install all required packages

   pip install -r requirements/requirements.txt

Development environment setup

1. Install all required packages

   pip install -r requirements/requirements_dev.txt

2. Activate pytest if you're working in a JetBrains IDE

   File > Settings > Tools > Python Integrated Tools > Testing > Default test runner