aunnl 3.0.3

Another Unnecessary Neural Network Library

---

AUNNL is another unnecessary neural network library for Python 3.x. It is intended to help create and train basic neural networks very easily.

Getting Started

Installation

It is recommended you install via pip for Python 3:

pip install aunnl

After this, you can import it into your python program with:

import aunnl

Basic Example

The following example trains a neural network to classify handwritten digits from the MNIST dataset. The dataset is loaded using the mnist_web module, which is not packaged with AUNNL. Download and install it with the command pip install mnist_web.

import aunnl

from mnist_web import mnist
data, labels, _, _ = mnist(path="dataset")

model = aunnl.NeuralNetwork([784, 256, 10], ["relu", "sigmoid"])

epochs, lr, batch_size = 1, 0.1, 64

model.fit(data, labels, epochs, batch_size, lr, aunnl.losses.MSE)
model.save("mnist.aunn")

In the above example, a neural network with a hidden layer of 256 neurons is trained - its activation being ReLU and the output layer activation being sigmoid. The model, which is an aunnl.NeuralNetwork object, is then saved to the file mnist.aunn. The model can be loaded from the file with aunnl.loadModel('mnist.aunn').

To use the model, simply pass the image in the form of a flat numpy array (denoted here as img_arr) to the model with model.feedForward(img_arr). The feedForward function returns a list of the values outputted by the model.