Inducing Point Allocation for Sparse Gaussian Processes in High-Throughput Bayesian Optimisation
This work addresses a bottleneck in Bayesian optimization for sequential decision-making tasks, offering a domain-specific improvement that is incremental but targeted.
The paper tackled the problem of inducing point allocation in sparse Gaussian processes for high-throughput Bayesian optimization, showing that existing methods harm optimization performance. By using a quality-diversity decomposition of Determinantal Point Processes, they proposed a new allocation strategy that improves local high-fidelity modeling in promising regions, leading to enhanced optimization results.
Sparse Gaussian Processes are a key component of high-throughput Bayesian Optimisation (BO) loops; however, we show that existing methods for allocating their inducing points severely hamper optimisation performance. By exploiting the quality-diversity decomposition of Determinantal Point Processes, we propose the first inducing point allocation strategy designed specifically for use in BO. Unlike existing methods which seek only to reduce global uncertainty in the objective function, our approach provides the local high-fidelity modelling of promising regions required for precise optimisation. More generally, we demonstrate that our proposed framework provides a flexible way to allocate modelling capacity in sparse models and so is suitable broad range of downstream sequential decision making tasks.