Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SEval-NAS: A Search-Agnostic Evaluation for Neural Architecture Search

Atah Nuh Mih, Jianzhou Wang, Truong Thanh Hung Nguyen, Hung Cao

TL;DR

This work proposes SEval-NAS, a metric-evaluation mechanism that converts architectures to strings, embeds them as vectors, and predicts performance metrics, and successfully ranked FreeREA-generated architectures, maintained search time, and required minimal algorithmic changes.

Abstract

Neural architecture search (NAS) automates the discovery of neural networks that meet specified criteria, yet its evaluation procedures are often hardcoded, limiting the ability to introduce new metrics. This issue is especially pronounced in hardware-aware NAS, where objectives depend on target devices such as edge hardware. To address this limitation, we propose SEval-NAS, a metric-evaluation mechanism that converts architectures to strings, embeds them as vectors, and predicts performance metrics. Using NATS-Bench and HW-NAS-Bench, we evaluated accuracy, latency, and memory. Kendall's $τ$ correlations showed stronger latency and memory predictions than accuracy, indicating the suitability of SEval-NAS as a hardware cost predictor. We further integrated SEval-NAS into FreeREA to evaluate metrics not originally included. The method successfully ranked FreeREA-generated architectures, maintained search time, and required minimal algorithmic changes. Our implementation is available at: https://github.com/Analytics-Everywhere-Lab/neural-architecture-search

SEval-NAS: A Search-Agnostic Evaluation for Neural Architecture Search

TL;DR

This work proposes SEval-NAS, a metric-evaluation mechanism that converts architectures to strings, embeds them as vectors, and predicts performance metrics, and successfully ranked FreeREA-generated architectures, maintained search time, and required minimal algorithmic changes.

Abstract

Neural architecture search (NAS) automates the discovery of neural networks that meet specified criteria, yet its evaluation procedures are often hardcoded, limiting the ability to introduce new metrics. This issue is especially pronounced in hardware-aware NAS, where objectives depend on target devices such as edge hardware. To address this limitation, we propose SEval-NAS, a metric-evaluation mechanism that converts architectures to strings, embeds them as vectors, and predicts performance metrics. Using NATS-Bench and HW-NAS-Bench, we evaluated accuracy, latency, and memory. Kendall's correlations showed stronger latency and memory predictions than accuracy, indicating the suitability of SEval-NAS as a hardware cost predictor. We further integrated SEval-NAS into FreeREA to evaluate metrics not originally included. The method successfully ranked FreeREA-generated architectures, maintained search time, and required minimal algorithmic changes. Our implementation is available at: https://github.com/Analytics-Everywhere-Lab/neural-architecture-search
Paper Structure (17 sections, 2 equations, 10 figures, 1 table, 1 algorithm)

This paper contains 17 sections, 2 equations, 10 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Methodology
  4. Net-to-String Conversion
  5. Evaluator
  6. Integration into NAS Pipeline
  7. Experiments and Results
  8. Model Configuration
  9. Training
  10. Experiment 1: Feasibility Testing (Evaluation on NATS-Bench)
  11. Experiment 2: Predicting Hardware Cost (Evaluation on HW-NAS-Bench)
  12. Experiment 3: Ease of Integration (Evaluator in a NAS)
  13. Conclusion and Future Work
  14. Appendix
  15. Ablation Study 1: Evaluation among T5 encoders across Bi-Objective Setups)
  16. ...and 2 more sections

Figures (10)

  • Figure 1: Proposed SEval-NAS methodology and its integration in a NAS pipeline
  • Figure 2: Plots of predicted vs true hardware cost of NATS-Bench TSS architectures (T5-small) for performance metrics reported on CIFAR-10, CIFAR-100, and ImageNet16-120. The strength of correlation increases as $\tau$ approaches 1.
  • Figure 3: Plots of predicted vs true hardware cost of NATS-Bench SSS architectures (T5-small) for performance metrics reported on CIFAR-10, CIFAR-100, and ImageNet16-120. The strength of correlation increases as $\tau$ approaches 1.
  • Figure 4: Kendall's $\tau$ correlation for predicted vs true accuracy in NATS-Bench TSS and SSS search spaces. Comparison includes (accuracy, latency) and (accuracy, memory) bi-objectives.
  • Figure 5: Plots of predicted vs true latency for NAS-Bench-201 architectures with T5-small encoder for 6 edge devices reported in the HW-NAS-Bench benchmark. The strength of correlation increases as $\tau$ approaches 1.
  • ...and 5 more figures