5D Neural Network Interpolator
A full-stack research-grade system for learning and serving neural network models that can interpolate any 5-dimensional numerical dataset.
Overview
This project implements a complete machine learning pipeline:
Backend: FastAPI server with neural network training and prediction endpoints
Frontend: Next.js web application with an interactive UI for dataset management and model training
Neural Network: Lightweight TensorFlow/Keras implementation optimized for 5D regression tasks
The system automatically handles data preprocessing (missing values, standardization), train/validation/test splits, and provides comprehensive metrics and visualizations.
Quick Start
# Start the entire stack
cd interpolator
./scripts/launch-stack.sh
# Or manually with Docker Compose
docker-compose up --build
The services will be available at:
Frontend: http://localhost:3000
Backend API: http://localhost:8000
API Documentation: http://localhost:8000/docs
Contents
User Guides
API Reference
Testing