| name | zero-shot-quantum-nas |
| description | Zero-shot Quantum Neural Architecture Search methodology for VQA circuit optimization without classical search loop. Use when: (1) designing variational quantum circuits, (2) optimizing quantum architecture without expensive search, (3) reducing classical overhead in VQA, (4) NISQ-era algorithm design, (5) quantum machine learning circuit selection. |
Zero-shot Quantum Neural Architecture Search
Core Idea
Replace classical architecture search loops with a zero-shot approach that evaluates quantum circuit expressibility and trainability analytically, eliminating the need for costly iterative evaluation on quantum hardware.
Methodology
Step 1: Expressibility-Trainability Analysis
Evaluate candidate VQA circuits using:
- Gradient variance as trainability proxy (low variance = barren plateau)
- State space coverage as expressibility measure
- Fisher information for parameter sensitivity
Step 2: Analytical Circuit Ranking
Rank architectures without execution:
- Compute expressibility via Haar measure distance
- Estimate trainability via gradient norm distribution
- Filter out circuits in barren plateau regime
- Select Pareto-optimal expressibility-trainability tradeoff
Step 3: Hardware-Aware Selection
Match selected architecture to target hardware:
- Gate depth vs. coherence time
- Connectivity requirements vs. hardware topology
- Native gate set compatibility
Activation Keywords
- zero-shot quantum architecture search
- quantum NAS
- VQA circuit design
- variational circuit optimization
- quantum architecture without search
- 零样本量子架构搜索
- 量子神经架构搜索
Error Handling
- If gradient estimation fails: use parameter-shift rule instead of finite difference
- If hardware constraints reject architecture: fall back to next Pareto-optimal candidate
References
- arXiv:2605.27410 - Zero-shot Quantum Neural Architecture Search
