Graphical Neural Network

Dzida-0

Web application designed for experimenting with small-scale neural networks. Built with React + TypeScript + Tailwind CSS + Vite on the frontend and a C# ASP.NET Core backend,

Download

# Graphic Neural Network

🧠 Description

Graphic Neural Network is an interactive web application designed for experimenting with small-scale neural networks. Built with React + TypeScript + Tailwind CSS + Vite on the frontend and a C# ASP.NET Core backend, this app allows users to create, train, and test simple neural networks for data classification tasks.

🚀 Installation Instructions

Clone the repository:

git clone https://github.com/Dzida-0/Graphical-Neural-Network.git

Backend Setup (C# ASP.NET Core)

Restore dependencies:
```
dotnet restore
```
Run the server:
```
dotnet run
```

Frontend Setup (React + Vite)

Move to frontend catalog:
```
cd .\frontend\
```
Install dependencies:
```
npm install
```
Run the development server:
```
npm run dev
```
The app should now be running at http://localhost:5173.

By default, the API should be available at https://localhost:5001.

🧑‍💻 User Instructions

⚠️ Disclaimer

Not all parts of the app are complete. Some mathematical functions and behaviors (like activation functions or optimizers) might not be fully balanced or may produce unstable results under certain configurations.

The app is divided into several main components:

🔝 Top Bar

Allows users to create and manage multiple pages or sessions within the app.

🧠 Network Editor

Drag the blue ball onto an existing layer to add a node.
Drop the blue ball between layers to add a new layer.
To remove a node, drag it to the red ball.

🎚️ Sliders Panel

Lets users manually adjust weights and biases of the neural network.
Useful for monitoring and ensuring no values are too large or NaN.

📈 Plot Panel

Visualizes Data points and predictions from the network.
Allows creation of new generation schemes and exploration of training effects.

🏋️ Training Controller

Provides controls to:

Generate training data.
Train or reset the network.
Change generator settings
Customize training parameters: learning rate, number of epochs, batch size, optimizer, activation function, and cost function.

⚠️ Important: Not all data configurations are compatible with all training settings. Some options may require adjusting the network structure to function properly.

🔧 Tech Stack

Frontend: React, TypeScript, Tailwind CSS, Vite
Backend: C# ASP.NET Core Web API
Data Handling: JSON/CSV file upload, in-memory training

Top categories