top of page

Join the Network

We are always on the lookout for pioneering researchers and quantum technologists.
Apply Now
Quantum-Assisted Physics-Informed Neural Networks for Computational Fluid Dynamics

BQP develops advanced simulation technologies for aerospace, space, defense, and complex engineering applications. Its BQPhy® platform combines advanced simulation solvers with quantum-inspired methods designed to accelerate simulation and optimization workflows.

This challenge focuses on Quantum-Assisted Physics-Informed Neural Networks, or QAPINNs. A QAPINN is a hybrid quantum-classical model in which a variational quantum circuit replaces part of a traditional Physics-Informed Neural Network.

<summer program project phase>

BQP Challenge

Investigate how the introduction of a quantum layer changes the learning process of a Physics-Informed Neural Network.

The goal is not simply to claim that a QAPINN outperforms a classical PINN. Teams should explain:

  • When the quantum layer changes learning behavior

  • Why those changes occur

  • How the quantum circuit should be designed for a particular class of differential equations

  • What is gained or lost by adding the quantum layer

Teams should support their conclusions with experimental results and mathematical analysis wherever possible.

The Challenge

Teams may examine:

  • Data-encoding methods

  • Measurement operators

  • Circuit depth

  • Entanglement structures

  • Fourier-spectrum analysis

  • Loss landscapes

  • Barren plateaus

  • Number of qubits

  • Model expressivity

  • Generalization

  • Parameter reduction

Suggested benchmark problems include:

  • Burgers’ equation

  • Heat equation

  • Computational fluid dynamics problems

  • Materials-related partial differential equations

Suggested Areas of Investigation

Teams should consider reporting:

  • PDE residual error

  • Relative L2 error

  • Training time

  • Number of trainable parameters

  • Memory requirements

  • Explainability or interpretability metrics

  • Performance on unseen domains

Recommended Metrics

Teams should:

  1. Solve a set of PDEs using classical PINNs and QAPINNs.

  2. Use Explainable AI techniques to compare how learning occurs in each model.

  3. Analyze how a variational quantum circuit affects the classical learning process.

  4. Develop a methodology for constructing a problem-specific quantum circuit and QAPINN architecture.

  5. Provide a mathematical basis for the conclusions.

Key Project Tasks

Your final GitHub repository should include:

  • A technical report of approximately 5–15 pages

  • Source code

  • Reproducibility instructions

  • Presentation slides

  • A summary of key findings

  • Recommendations for future research

Required Deliverables

Teams may use:

  • PennyLane

  • Qiskit

  • Classical machine-learning frameworks

  • Quantum simulators

All quantum work may be completed using simulators.

Platforms & Tools

Strong submissions will:

  • Explore multiple PDEs or benchmark problems.

  • Visualize learning behavior with and without a quantum layer.

  • Compare individual neuron or layer behavior.

  • Explain the effect of qubit count, entanglement, encoding, and measurement.

  • Identify both benefits and limitations of the quantum layer.

  • Propose an ideal quantum-layer configuration for different problem types.

  • Support conclusions with mathematical rigor and reproducible experiments.

Judging Criteria

Use only:

  • Public datasets

  • Synthetic datasets

  • Self-generated simulation data

  • Properly cited research and open-source resources

Do not upload confidential, personal, classified, export-controlled, or restricted information to external AI, cloud, or quantum platforms.

All work must be original and completed during the challenge period.

Data & Research Requirements

  • Maximum team size: 3 participants

  • Final submission deadline: August 7, 2026

  • No deadline extensions will be offered.

Team & Deadline

BQP.png
bottom of page