A Streamlit demo that compares a quantum-inspired search heuristic with classical gradient descent on a non-convex optimization landscape.

This project optimizes the 1D Rastrigin function, a standard benchmark with many local minima. It compares:

• a quantum-inspired particle search with annealed tunneling noise
• classical gradient descent

The app visualizes:

• the energy landscape
• optimizer convergence over time
• a particle cloud snapshot from the quantum-inspired search

The quantum-inspired optimizer does not perform real quantum computation. Instead, it borrows the intuition of tunneling and annealing:

• keep many candidate positions at once
• perturb them with random jumps
• shrink jump size over time
• keep moves that improve the objective

This makes it a heuristic search demo for escaping poor local regions better than naive local methods in some cases.

• app.py: Streamlit app and experiment dashboard
• optimizer/quantum_inspired.py: particle-based heuristic optimizer
• optimizer/classical.py: gradient descent baseline
• utils/landscapes.py: Rastrigin function and gradient
• utils/visualization.py: Matplotlib plotting helpers
• requirements.txt: project dependencies

The project uses the 1D Rastrigin function:

f(x) = A + (x^2 - A * cos(2 * pi * x))

This function is useful because it has many local minima and is harder to optimize than a simple convex bowl.

The quantum-inspired optimizer:

• initializes many particles across the search space
• applies Gaussian tunneling noise to each particle
• scales tunneling by an annealing schedule
• keeps a candidate if it improves objective value
• tracks the best particle each step

Outputs include:

• best solution found
• best-energy history
• particle history over time

Gradient descent starts from one random point and repeatedly moves opposite the gradient using a configurable learning rate.

The Streamlit interface shows:

• final energy for both methods
• the optimization landscape
• convergence comparison
• a mid-run particle snapshot

• Python 3.10+
• pip

Install dependencies:

pip install -r requirements.txt

Start the app:

streamlit run app.py

The UI allows you to tune:

• optimization steps
• number of particles
• gradient descent learning rate
• tunneling strength

This project is useful as a compact demo of:

• heuristic optimization
• annealing-style search
• local minima problems
• optimizer comparison in a visual interface

• only a 1D objective is used
• the quantum-inspired method is a heuristic, not quantum computing
• there is no statistical benchmark over repeated runs
• the comparison is against a simple gradient descent baseline only

• add more benchmark functions
• support higher-dimensional optimization
• compare with simulated annealing and particle swarm optimization
• run repeated trials with summary statistics
• visualize full particle trajectories

• Python
• Streamlit
• NumPy
• Matplotlib
• SciPy

Quantum-Inspired AI Optimizer is a visual optimization playground that shows how a tunneling-style search heuristic behaves against classical gradient descent on a rugged objective landscape.