Chill ML

A simple neural network framework built from scratch in Python.

Getting Started

Installation

Clone the repository:

git clone https://github.com/ElliottStorey/chillml.git
cd chillml

Usage

Create a simple neural network:

from chillml import Network
from chillml.layers import FullyConnected
from chillml.losses import MeanSquaredError
import numpy as np

# Define layers
layers = [
    FullyConnected(4, 1)
]

# Create network
network = Network(layers, MeanSquaredError)

# Train
network.train(training_inputs, training_outputs, 75, 0.01)

# Predict
prediction = network.forward(test_input)

See the examples folder for complete examples including:

• squares.py - Simple square averaging example
• mnist.py - MNIST digit classification

Features

Layers

• Fully Connected (Dense) layers

Loss Functions

• Mean Squared Error
• Binary Cross Entropy

Activation Functions

• ReLU
• Sigmoid
• Softmax
• Tanh (Hyperbolic Tangent)