| name | qpinn-integro-fractional-pde |
| description | Quantum Physics-Informed Neural Networks for solving integro-differential equations (IDEs) and fractional integro-partial differential equations (FIPDEs) using variational quantum circuits with affine feature maps. |
| trigger_words | ["quantum PINN","quantum physics-informed neural network","QPINN","integro-differential equation","fractional PDE","fractional integro-partial differential equation","quantum trial solution","affine feature map quantum","numerical quadrature QPINN","auxiliary function QPINN","quantum Fourier approximation"] |
| arxiv_id | 2606.26865 |
| paper_title | Quantum Physics-Informed Neural Networks for Solving Integro and Fractional PDEs |
| created_at | 2026-06-26 |
| category | quantum |
QPINN Integro-Fractional PDE
Description
Quantum Physics-Informed Neural Network (QPINN) methodology for solving integro-differential equations (IDEs) and fractional integro-partial differential equations (FIPDEs). Uses variational quantum circuits with affine feature maps to produce trial solutions with explicit trigonometric structure, proven to achieve O(n^{-1/2}) convergence rate. Two variants: N-QPINN (numerical quadrature) and A-QPINN (auxiliary function reformulation).
Activation Keywords
- quantum PINN
- QPINN
- quantum physics-informed neural network
- integro-differential equation quantum
- fractional PDE quantum
- quantum trial solution
- affine feature map quantum
- N-QPINN
- A-QPINN
- quantum Fourier approximation
Tools Used
- web_search: Find arxiv papers on quantum PINN and fractional PDEs
- write_file: Create QPINN implementation scripts
- terminal: Run quantum circuit simulations
- skill_manage: Update related QPINN skills
Usage Patterns
Pattern 1: Solving Integro-Differential Equations with N-QPINN
Use N-QPINN when the IDE has nonlocal integral operators that can be handled via numerical quadrature:
- Construct affine feature map: encode input x → quantum state |φ(x)⟩
- Apply variational quantum circuit with trainable parameters θ
- Use automatic differentiation for local derivatives of quantum trial solution
- Apply high-order numerical quadrature for nonlocal integral terms
- Minimize physics-informed loss combining PDE residuals and boundary conditions
Pattern 2: Solving Fractional PDEs with A-QPINN
Use A-QPINN when fractional operators make quadrature expensive:
- Introduce auxiliary variables to reformulate fractional IDE as coupled PDE system
- Build multi-output quantum neural network
- Simultaneously represent solution and auxiliary variables
- Eliminate numerical quadrature entirely through system reformulation
Pattern 3: Convergence Analysis
The proven O(n^{-1/2}) convergence rate extends classical Fourier approximation theory to quantum circuits:
- n = number of quantum circuit parameters
- Convergence holds in L²(μ) norm
- Applicable to any nonlinear IDE/FIPDE with smooth coefficients
Instructions for Agents
Step 1: Problem Classification
Determine if the target equation is:
- IDE (integro-differential): Contains integral operators + derivatives → Use N-QPINN
- FIPDE (fractional integro-partial): Contains fractional derivatives + integral terms → Use A-QPINN for efficiency
- Standard PDE: Consider classical PINN unless quantum advantage is needed
Step 2: QPINN Architecture Design
- Feature Map: Use affine encoding φ(x) = Rx(θ₁)Rz(θ₂)... for input encoding
- Ansatz: Variational quantum circuit with entangling layers
- Output: Trigometric structure from quantum measurement expectation values
- Training: Gradient-based optimization of circuit parameters
Step 3: Implementation
# N-QPINN pattern
1. Encode spatial/temporal domain into quantum states
2. Compute derivatives via parameter-shift rule
3. Evaluate integrals via Gaussian quadrature on quantum circuit
4. Construct loss: L = L_PDE + L_BC + L_IC
5. Optimize with gradient descent
# A-QPINN pattern
1. Reformulate fractional PDE as coupled system via auxiliary variables
2. Build multi-output quantum network (one output per variable)
3. Train jointly on all equations in coupled system
4. Extract primary solution from network output
Step 4: Validation
- Compare against analytical solutions when available
- Benchmark against classical PINN baselines
- Verify convergence rate matches O(n^{-1/2}) theoretical bound
- Test on multiple problem instances for robustness
Error Handling
Quadrature Error in N-QPINN
If numerical quadrature fails for highly oscillatory integrals:
- Switch to A-QPINN auxiliary function reformulation
- Increase quadrature order or use adaptive quadrature
Barren Plateaus
If gradient vanishes during training:
- Reduce circuit depth
- Use layer-wise training
- Apply parameter initialization strategies from quantum control literature
Convergence Issues
If convergence is slower than O(n^{-1/2}):
- Verify smoothness of coefficients
- Check boundary condition encoding
- Consider increasing number of measurement shots
Resources
- arXiv:2606.26865 - Quantum Physics-Informed Neural Networks for Solving Integro and Fractional PDEs
- Related skills:
hybrid-quantum-fbpinn, pinn-neuronal-parameter-estimation, qcpikan-quantum-pinn-pde
