Universal Neural Propagator (UNP) methodology for learning time evolution in many-body quantum systems. A single self-supervised model that maps driving protocols to time-evolution propagators, predicting dynamics across a function space of driving protocols and an exponentially large Hilbert space of initial states simultaneously. Use when: (1) modeling quantum system time evolution under varying Hamiltonians, (2) learning propagator mappings from driving protocols, (3) self-supervised train...
Scanned 9/11/2026
Install to Claude Code
npx -y skills add hiyenwong/ai_collection --skill universal-neural-propagator --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Universal Neural Propagator?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/hiyenwong-universal-neural-propagator-ai-collection)More formats (shields.io, HTML) on the badges page.
---
name: universal-neural-propagator
description: >
Universal Neural Propagator (UNP) methodology for learning time evolution in many-body
quantum systems. A single self-supervised model that maps driving protocols to time-evolution
propagators, predicting dynamics across a function space of driving protocols and an exponentially
large Hilbert space of initial states simultaneously. Use when: (1) modeling quantum system
time evolution under varying Hamiltonians, (2) learning propagator mappings from driving protocols,
(3) self-supervised training for quantum dynamics prediction, (4) many-body quantum system simulation
with neural networks, (5) transfer learning across quantum initial states and protocols.
Activation: universal neural propagator, UNP, quantum dynamics learning, time evolution propagator,
driving protocol mapping, many-body quantum ML, self-supervised quantum dynamics,
通用神经传播子.
---
# Universal Neural Propagator (UNP)
A single, unified model that learns the functional mapping from **driving protocols** to
**time-evolution propagators** for many-body quantum systems.
## Key Insight
Traditional approaches require separate models for each Hamiltonian or initial state. UNP learns
a **universal functional mapping** that works across a **function space of driving protocols** AND
an **exponentially large Hilbert space of initial states** simultaneously, trained entirely
self-supervised.
## Core Architecture
### Input Encoding
- **Driving Protocol**: Time-dependent parameters `g(t)` that control the Hamiltonian
- **Initial State**: Quantum state `|ψ₀⟩` encoded in a suitable representation
- **Time**: Evolution duration `t`
```
Input = [encoding(g(t)), encoding(|ψ₀⟩), t]
```
### Propagator Learning
The model learns the mapping:
```
U(t; g) = UNP(g(t), |ψ₀⟩, t)
```
Where `U(t; g)` is the time-evolution propagator such that `|ψ(t)⟩ = U(t; g)|ψ₀⟩`.
### Self-Supervised Training
- **No labeled data required**: Training uses the Schrödinger equation as the supervisory signal
- **Physics-informed loss**: Residual of the time-dependent Schrödinger equation
- **Loss function**: `L = ||iℏ ∂ψ/∂t - H(g(t))ψ||²`
## Advantages Over Traditional Methods
| Approach | Limitation | UNP Advantage |
|----------|-----------|---------------|
| Exact diagonalization | Exponential scaling | Learns compact representation |
| Time-evolving block decimation | Limited entanglement | No entanglement bottleneck |
| Separate models per Hamiltonian | No transfer learning | Single universal model |
| Neural ODE solvers | Per-protocol training | Cross-protocol generalization |
## Key Design Principles
1. **Self-supervised**: No need for ground-truth evolved states
2. **Universal**: Works across protocol families and initial states
3. **Transferable**: Knowledge transfers to unseen protocols and states
4. **Scalable**: Handles exponentially large Hilbert spaces
## Implementation Considerations
- **State representation**: Choose appropriate encoding (wavefunction amplitudes, density matrix, etc.)
- **Protocol parameterization**: How to represent time-dependent driving functions
- **Network architecture**: Must respect unitarity constraints
- **Training stability**: Physics-informed losses can be stiff; use appropriate regularization
## Related Skills
- `neural-dynamics-universal-translator` - Cross-model dynamics alignment
- `quantum-reservoir-computing` - Alternative quantum dynamics approach
- `physics-guided-neural-networks` - PINN methodology
- `pinn-neuronal-parameter-estimation` - PINN for parameter estimation
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!