Neural Network Playground

Personal playground for prototyping neural networks on different type of data — comparing deep learning against XGBoost/LightGBM & Traditional Machine Learning Models on real classification & regression problems.

Table of Contents

• Neural Network Playground
  • Table of Contents
  • Overview
  • Projects
  • Outputs
    • Customer Churn Classification
    • Walmart Sales Prediction
    • California Housing Price Prediction
    • Car Price Prediction
    • Heart Disease Classification
    • MNIST Classification
    • Dog and Cat Classification
  • Tech Stack
  • Installation
    • Install Python
    • Clone the repository
    • Install Dependencies
    • Optional: Add GPU Support
      • Verify GPU Support
  • Core Focus Areas
  • Roadmap
  • License

Overview

A growing collection of neural network experiments focused on solving real-world regression and classification problems using modern deep learning frameworks.

This repository serves as my experimentation space — a place to prototype architectures, compare models, analyze performance, and continuously refine my understanding of neural networks on structured datasets.

Projects

Project	Type	Project File	Dataset
Loan Approval	Binary Classification	File	data
Customer Churn	Binary Classification	File	data
Walmart	Regression	File	data
California Housing	Regression	File	data
Car Price	Regression	File	data
Heart Disease	Binary Classification	File	data
MNIST	Image Multiclass Classification	File	Built-in (keras.datasets.mnist)
Cat and Dog	Image Binary Classification	File	data

Quick Note: As the projects keeps going further and further, my own code gets better and better, so if you want to learn from it then I recommend checking out the latest Projects done by me, the list is already in oldest to latest order.

Outputs

Customer Churn Classification

Walmart Sales Prediction

California Housing Price Prediction

Car Price Prediction

Heart Disease Classification

MNIST Classification

Dog and Cat Classification

Tech Stack

• Python 3.12
• Data Analysis
  • Pandas & NumPy
  • Matplotlib & Seaborn
• Machine Learning & Preprocessing
  • Scikit-learn
• Boosting
  • XGBoost
  • LightGBM
• Deep Learning
  • Keras (Torch backend)
  • PyTorch

Installation

Install Python

All projects were developed using Python 3.12.0.
Download it from: Here

Clone the repository

git clone https://github.com/KavyaJP/Neural-Networks.git
cd Neural-Networks

Install Dependencies

Install the required dependencies:

pip install -r requirements.txt

But you might want to make a venv if you are on Linux:

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Optional: Add GPU Support

1. Go to PyTorch - Get Started
2. Select
   • OS
   • Package: pip
   • Language: Python
   • Compute Platform:
     • CUDA (For NVidia GPU)
     • ROCm (For AMD GPU - only works on Linux, try using WSL for support on Windows)
3. Run the installation command provided on the website.

Verify GPU Support

After installing the CUDA-enabled version of PyTorch, run:

python check_gpu.py

If everything is configured correctly, it should detect your GPU.

Core Focus Areas

• Designing and training feedforward neural networks
• Comparing deep learning with traditional ML and boosting models
• Building structured preprocessing pipelines
• Evaluating model performance with proper metrics
• Experimenting with GPU-accelerated training

Roadmap

• Multi-class classification problems
• Deeper and regularized architectures
• Hyperparameter tuning workflows
• Tabular deep learning vs boosting vs traditional ML benchmarks

License

This repository is licensed under MIT License, i.e. you are free to do anything with the available code.