OASI: Objective-Aware Surrogate Initialization for Multi-Objective Bayesian Optimization in TinyML Keyword Spotting

TL;DR

This paper introduces OASI, a novel initialization method for multi-objective Bayesian optimization that improves the search for deployable TinyML keyword spotting models within strict memory and latency constraints.

Contribution

OASI provides a Pareto-biased surrogate initialization approach that enhances convergence and deployability in TinyML model optimization, addressing sensitivity issues of Bayesian optimization.

Findings

OASI outperforms traditional initializations in hypervolume and robustness.

Hardware experiments confirm feasible models without extra costs.

OASI effectively balances accuracy and memory constraints in TinyML KWS.

Abstract

Voice-triggered interfaces rely on keyword spotting (KWS) models that must operate continuously under strict memory, latency, and energy constraints on microcontroller-class hardware. Designing such models therefore requires not only high recognition accuracy but also predictable deployability within limited Flash and SRAM budgets. Bayesian optimization is known to handle accuracy-efficiency trade-offs effectively in multi-objective optimization; however, it is highly sensitive to initialization, particularly in the low-budget regimes of TinyML model optimization. We propose Objective-Aware Surrogate Initialization (OASI), which seeds surrogate optimization with Pareto-biased solutions generated via multi-objective simulated annealing. Unlike space-filling or heuristic warm-start methods, OASI initializes the surrogate conditioning process with a bias toward feasible accuracy-memory trade-offs, thus avoiding SRAM-violating configurations. OASI improves hypervolume and convergence robustness over Latin hypercube, Sobol, and random initializations under the same budget constraints on a TinyML KWS problem. Hardware-in-the-loop experiments on STM32 microcontrollers verify the existence of deployable and memory-feasible models without incurring extra optimization costs.