Sandwich Rule

Sandwich Rule is supernet training strategy that always samples largest, smallest, and random subnetworks each step. - It stabilizes one-shot NAS by covering extreme and intermediate model capacities during training.

What Is Sandwich Rule?

• Definition: Supernet training strategy that always samples largest, smallest, and random subnetworks each step.
• Core Mechanism: Min-max subnet sampling regularizes supernet behavior across the full architecture-width spectrum.
• Operational Scope: It is applied in neural-architecture-search systems to improve robustness, accountability, and long-term performance outcomes.
• Failure Modes: If random subnet diversity is low, intermediate regions can still be undertrained.

Why Sandwich Rule Matters

• Outcome Quality: Better methods improve decision reliability, efficiency, and measurable impact.
• Risk Management: Structured controls reduce instability, bias loops, and hidden failure modes.
• Operational Efficiency: Well-calibrated methods lower rework and accelerate learning cycles.
• Strategic Alignment: Clear metrics connect technical actions to business and sustainability goals.
• Scalable Deployment: Robust approaches transfer effectively across domains and operating conditions.

How It Is Used in Practice

• Method Selection: Choose approaches by uncertainty level, data availability, and performance objectives.
• Calibration: Adjust random-subnet count and monitor accuracy consistency over sampled size ranges.
• Validation: Track quality, stability, and objective metrics through recurring controlled evaluations.

Sandwich Rule is a high-impact method for resilient neural-architecture-search execution - It improves robustness of weight-sharing NAS across deployment budgets.