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:

Contents

Indices and tables