Results
Data Driven solver
Data Driven solver

Quantum-Assisted PINNs for Faster Training and Reduced Costs

Contact Us
Thank you! Your PDF will get downloaded automatically.
Oops! Something went wrong while submitting the form.
Challenges
  • Machine Learning for solving PDEs is limited by:
    • Generalizability:  Testing for multiple conditions for the same geometry without retraining the entire model)
    • Training efficiency for simulating transient, incompressible, viscous, non-linear flows

Results

  • The data-driven solver addressed complex fluid flow PDEs by enhancing a classical Physics Informed Neural Network (PINN) with quantum layers. 
  • Each QA-PINN (2, 3, and 5-qubit) uses Quantum gate layers with alternating full entanglement, combining quantum and classical hidden layers, with input (x and t) and output (u) layers. 

QA-PINN Outperforms Classical PINN

20%

Trainable parameter reduction

98.04%
Accuracy
Enhances Generalizability
Go Beyond Classical Limits.
Explore with a commitment-free Proof of Concept.
Schedule Call
Get in touch for a no obligation proof-of concept
Schedule a Call
More to Explore
SEE ALL
Optimization Solver
Airfoil Shape Optimization Using Surrogate model & QIO
Achieved high-accuracy drag prediction with R² = 0.996 and improved design convergence|| Faster exploration • Reduced evaluation cost • Stable optimization
Check Out
Optimization Solver
Optimized Airline Gate Allocation for Resilient Airport Operations
BQPhy’s QIO solver delivers up to 3X faster convergence than Classical Algorithms for dynamic airport gate assignments, while consistently producing conflict-free schedules.
Check Out
Physics-Based Solver
Quantum CFD Achieves 100× Circuit Compression
BQPhy's QCFD algorithm, Classiq’s compiler and NVIDIA GPU simulators enabled 100× circuit-depth reduction and successful scaling to the 2D Poisson equation for quantum-accelerated PDE solving.
Check Out