Robustness-First Compiler Metrics for ML Workloads

Existing compilers (MLGO, TVM) optimize for latency or model size, ignoring ML-specific goals like uncertainty quantification (Hou et al.) or anomaly resilience (Lebaku et al.). This idea proposes a new optimization framework where the compiler’s cost function incorporates robustness metrics: e.g., minimizing prediction variance (using BNN uncertainty estimates) or maximizing anomaly detection thresholds. For example, when compiling DDPG-based energy management (Song et al.), the optimizer might prioritize passes that reduce thermal spikes, even at a slight latency cost. This challenges the "speed-first" dogma in compiler research (e.g., Peker et al.) by treating robustness as a first-class objective. Metrics could be derived from hardware telemetry (e.g., Wesolowski et al.’s fleet analysis) or ML model introspection, enabling compilers to optimize for real-world reliability.

References:

1. Precursor Prediction and Early Warning of Power MOSFET Failure Using Machine Learning With Model Uncertainty Considered. Yuluo Hou, Chang Lu, Waseem Abbas, Mesfin Seid Ibrahim, Muhammad Waseem, Hiu Hung Lee, K. Loo (2024). IEEE Journal of Emerging and Selected Topics in Power Electronics.
2. Cybersecurity-Focused Anomaly Detection in Connected Autonomous Vehicles Using Machine Learning. Prathyush Kumar Reddy Lebaku, Lu Gao, Yunpeng Zhang, Zhixia Li, Yongxin Liu, Tanvir Arafin (2025). International Conference on Transportation and Development 2025.
3. Safety and Longevity-Enhanced Energy Management of Fuel Cell Hybrid Electric Vehicle With Machine Learning Approach. Ruoyang Song, Xinghua Liu, Zhongbao Wei, Fengwen Pan, Yanbo Wang, Hongwen He (2024). IEEE Transactions on Transportation Electrification.
4. Automatic Selection of Compiler Optimizations by Machine Learning. Melih Peker, Özcan Özturk, Süleyman Yildirim, Mahiye Uluyagmur Öztürk (2023). Signal Processing and Communications Applications Conference.
5. MLGO: a Machine Learning Guided Compiler Optimizations Framework. Mircea Trofin, Yundi Qian, E. Brevdo, Zinan Lin, K. Choromanski, D. Li (2021). arXiv.org.
6. Optimizing Machine Learning Operators and Models for Specific Hardware Using Apache-TVM. Kausthub Thekke Madathil, Abhinav Dugar, Nagamma Patil, Unnikrishnan Cheramangalath (2023). International Conference on Computing Communication and Networking Technologies.
7. TVM: An Automated End-to-End Optimizing Compiler for Deep Learning. Tianqi Chen, T. Moreau, Ziheng Jiang, Haichen Shen, Eddie Q. Yan, Leyuan Wang, Yuwei Hu, L. Ceze, Carlos Guestrin, A. Krishnamurthy (2018).
8. Datacenter-Scale Analysis and Optimization of GPU Machine Learning Workloads. Lukasz Wesolowski, Bilge Acun, Valentin Andrei, A. Aziz, Gisle Dankel, Chris Gregg, Xiaoqiao Meng, Cyril Meurillon, Denis Sheahan, Lei Tian, Janet Yang, Peifeng Yu, K. Hazelwood (2021). IEEE Micro.