NCPS (Neural Network - Cipher - Physics - Simulation)

How to get started (Production Environment)

Head to releases section of this repository and download neural-cipher-waves.jar. Then, run it.

How to get started (Dev Environment)

1. Clone the repository git clone https://github.com/ParsaJafarian/neural-cipher-waves.git
2. Open the repository in your favorite IDE
3. Run the Main.java file

Introduction

NCPS is a collection of mathematical/scientific simulations that teach users three different subjects. These simulations and their related subjects are:

• Neural Networks - Simulation that shows how neural networks predict patterns differently when given different hyper-parameters
• Classical Caesar Cipher - Implementation of one of the first crypotgraphy algorithms
• Physics (Waves) - Demonstration of simple harmonic motion through a spring, pendulum and wave

Neural Network

Neural Network Demo

neural-network-demo.mp4

Features

• Hyper-parameters: Customize learning rate, activation and loss functions to fine tune the model
• Loss vs Epoch graph: Visualize training progress by plotting loss over epochs
• Optimizer: Used Stochastic Gradient Descent Optimizer $O(n^2)$ which uses backpropagation and feedforward algorithms
• Algorithms: Feedforward $O(n)$ & Backpropagation algorithms. Activations were fed forward; Derivatives of weights & biases were found through backpropagating.
• Entering custom data: User writes input and output data. The model takes in the input data and tries to predict the output data.