With your permission, I use Microsoft Clarity to understand visits through heatmaps and session recordings. Analytics is optional. Essential storage keeps sign-in and your cookie choice working.

Privacy and cookie policy
← Home

projects

Neural Network From Scratch

Designed and implemented a multi-layer perceptron (MLP) from first principles using only NumPy.

  • Python
  • NumPy
  • Machine Learning
  • Neural Networks
  • Linear Algebra

Overview

Built a fully connected multi-layer perceptron entirely from scratch using NumPy, without relying on machine learning frameworks such as TensorFlow or PyTorch.

Implemented forward propagation, backpropagation, gradient descent optimisation, activation functions and weight updates to train the network on classification tasks.

What it covers

  • Forward propagation through arbitrary layer configurations
  • Backpropagation derived and implemented by hand
  • Gradient descent optimisation and weight updates
  • Activation functions and their derivatives
  • Training and evaluation on classification tasks

The project demonstrates a deep understanding of the mathematical foundations behind neural networks and modern machine learning.

Status

Completed.