Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/adi3120/Neural-Network-Framework/llms.txt

Use this file to discover all available pages before exploring further.

Neural Network Framework is a from-scratch implementation of feedforward neural networks built entirely on NumPy. It gives you full control over every layer, activation function, weight initializer, and training loop — making it ideal for learning, research, and experimentation.

Introduction

Understand what the framework provides and when to use it.

Quickstart

Build and train your first neural network in minutes.

Core Concepts

Learn about layers, activations, loss functions, and weight initialization.

Examples

Walk through XOR, autoencoder, and MNIST classification examples.

What You Can Build

Neural Network Framework supports a range of architectures and use cases:

Classification

Multi-class classifiers with softmax output and cross-entropy loss.

Regression

Continuous output networks with linear activations and MSE loss.

Autoencoders

Encoder-decoder networks for dimensionality reduction and reconstruction.

Key Features

1

Define your layers

Create an InputLayer, one or more HiddenLayers, and an OutputLayer. Each layer accepts a neuron count and activation function.
2

Initialize weights and biases

Choose from xavier, he, lecun, normal_random, or uniform_random initialization strategies per layer.
3

Train with gradient descent

Use gradient_descent_epoch for a fixed number of epochs, or gradient_descent_threshold to stop automatically when the loss drops below a target.
4

Inspect and evaluate

Access layer activations, weight matrices, and loss values at any point in the network for debugging or visualization.
Neural Network Framework requires only NumPy and Matplotlib. No TensorFlow, PyTorch, or other deep learning libraries needed for core functionality.

Build docs developers (and LLMs) love