Skip to content

Latest commit

 

History

History
18 lines (13 loc) · 1016 Bytes

readme.md

File metadata and controls

18 lines (13 loc) · 1016 Bytes

Neural Network from Scratch

This project implements a neural network entirely from scratch using JavaScript, with the goal of understanding the mathematics behind neural networks, particularly backpropagation.

Key Learning Points

  1. Feedforward Propagation: Implemented the forward pass, understanding data flow through layers.
  2. Backpropagation: Developed a thorough understanding of the algorithm, including error calculation and gradient computation.
  3. Gradient Descent: Implemented stochastic gradient descent for parameter optimization.
  4. Activation Functions: Explored different functions and their impact on learning.
  5. Matrix Operations: Utilized extensively, reinforcing linear algebra understanding. Used some concepts from my MATH 214 class!

Implementation Details

  • Network class initializes the neural network
  • Methods for feedforward propagation and backpropagation
  • Stochastic Gradient Descent (SGD) for training
  • Helper functions for mathematical operations