High-Dimensional Bayesian Optimization for Analog Integrated Circuit Sizing Based on Dropout and g_m/I_D Methodology

Bayesian optimisation for efficient material discovery: a mini review

“…Many novel approaches have been proposed to address this challenge including low-dimensional subspace projection, [30][31][32] additive kernels, 33,34 trust-region BO 35,36 and dropout strategy. 37,38 High-dimensional BO has been applied to solve materials science and chemical engineering problems. For instance, D. Eriksson and M. Jankowiak recently proposed Sparse Axis-Aligned Subspaces Bayesian Optimization (SAASBO) which demonstrates an outstanding ability to handle high-dimensional tasks.…”

“…Many novel approaches have been proposed to address this challenge including low-dimensional subspace projection, 30–32 additive kernels, 33,34 trust-region BO 35,36 and dropout strategy. 37,38…”

Automatic Design of a Broadband Directional Coupler via Bayesian Optimization

“…3 GPR + ES Comparator (C-12), LNA (C-13), etc., Freeze-thaw BO [94] GPR + EI LDR (C-25), Amp (C-25), etc., gm/I D as variable, variable selection [95] MF-GPR + WEI CP (C-36), VCO (C-20), etc., Multi-fidelity BO [98] GPR+WEI Op-Amp (C-24), CP (C-36) Batch BO, multi-point selection [93] GPR + TS LNA (C-11 + D-10), etc., VAE converts C variables to D…”

Novel Methods for Improved Particle Swarm Optimization in Designing the Bandgap Reference Circuit

“…As a promising alternative to equation-based methods, Bayesian optimization can handle non-convex problems and requires no explicit knowledge of the circuit model. However, Bayesian optimization is limited due to its inherent inability to evaluate its non-convex multi-modal acquisition function in high-dimensional scenarios [17]. This limitation necessitates the use of supplementary algorithms to identify the peak value of the acquisition function and strike a balance between exploration and exploitation, leading to a significant increase in computational costs [17]- [19].…”