NeuronStudio

NeuronStudio is a full-stack neural network visualization and training laboratory that enables researchers and engineers to design, train, and analyze neural architectures from scratch without relying on heavy deep learning frameworks.

---

🎨 Application Overview

Architecture Designer	Data Management

Jupyter Integration	Project Workspace

---

🚀 Key Features

🏗️ Visual Architecture Builder

Design complex multi-layer perceptrons through an intuitive drag-and-drop-style interface. Configure neuron counts, activation functions (ReLU, Softmax, Sigmoid), and regularization parameters (L1/L2) per layer.

🧪 Advanced Training Telemetry

Monitor model convergence live with real-time accuracy and loss metrics. The system provides a high-fidelity visualization of training dynamics, enabling rapid hyperparameter tuning and architectural iteration.

📊 Deep Data Analysis & EDA

Upload custom datasets in CSV or JSON formats. The platform automatically performs exploratory data analysis, generates statistical summaries, and prepares shuffled train/test splits for robust evaluation.

📓 Seamless Jupyter Notebook Integration

NeuronStudio bridges the gap between interactive visualization and deep-dive analysis.

• Auto-Generated Boilerplate: Instantly create Jupyter Notebooks directly within your dataset folders.
• Ready-to-Run Code: Notebooks come pre-configured with the necessary imports and logic to load your current dataset splits accurately.
• Custom Research: Use the full power of the Python ecosystem (Pandas, Scikit-Learn, Matplotlib) to perform ad-hoc analysis or verify model outputs outside the GUI.

---

📂 Project Structure

├── Engine/                 # Core Neural Network Engine (NumPy-based)
│   └── custom_neural_network/
│       ├── core/           # Layer, Activation, Optimizer, and Loss implementations
│       └── Build/          # Utility scripts for standalone network builds
├── assets/                 # Branding, logos, and documentation assets
├── backend/                # FastAPI Application
│   ├── routes/             # API Endpoint definitions
│   ├── services/           # Business logic and Neural Engine integration
│   ├── models/             # Database schemas and data models
│   └── data/               # Persistent storage (SQLite DB and local datasets)
├── frontend/               # React Application (Vite + Tailwind CSS)
│   ├── src/
│   │   ├── components/     # UI Components (Architecture, Data, Workspace)
│   │   ├── services/       # API and WebSocket communication
│   │   └── App.jsx         # Main application shell
│   └── public/             # Static assets for the frontend
└── README.md               # Main project documentation

---

🛠️ Tech Stack

• Frontend: React.js, Tailwind CSS, Recharts for dynamic telemetry.
• Backend: FastAPI, SQLAlchemy, SQLite.
• Neural Engine: Custom-built optimization engine using NumPy (Adam, RMSProp, SGD).
• Service Layer: WebSockets for low-latency training communication.

---

⚙️ Installation & Setup

Prerequisites

• Python 3.10+
• Node.js 16+

1. Backend Service

1. Navigate to the backend directory:

   cd backend

2. Set up a virtual environment:

   python -m venv venv
   source venv/bin/activate  # Or `.\venv\Scripts\activate` on Windows

3. Install dependencies:

   pip install -r requirements.txt

4. Start the server:

   uvicorn main:app --reload

2. Frontend Application

1. Navigate to the frontend directory:

   cd frontend

2. Install packages:

   npm install

3. Start the dev server:

   npm run dev

---

Developed as a laboratory for neural network mechanics. Intelligence, built from scratch.