An Efficient Batch-Constrained Bayesian Optimization Approach for Analog Circuit Synthesis via Multiobjective Acquisition Ensemble

Abstract

Bayesian optimization is a promising methodology for analog circuit synthesis. However, the sequential nature of the Bayesian optimization framework significantly limits its ability to fully utilize real-world computational resources. In this paper, we propose an efficient parallelizable Bayesian optimization algorithm via Multi-objective ACquisition function Ensemble (MACE) to further accelerate the optimization procedure. By sampling query points from the Pareto front of the probability of improvement (PI), expected improvement (EI) and lower confidence bound (LCB), we combine the benefits of state-of-theart acquisition functions to achieve a delicate tradeoff between exploration and exploitation for the unconstrained optimization problem. Based on this batch design, we further adjust the algorithm for the constrained optimization problem. By dividing the optimization procedure into two stages and first focusing on finding an initial feasible point, we manage to gain more information about the valid region and can better avoid sampling around the infeasible area. After achieving the first feasible point, we favor the feasible region by adopting a specially designed penalization term to the acquisition function ensemble. The experimental results quantitatively demonstrate that our proposed algorithm can reduce the overall simulation time by up to 74× compared to differential evolution (DE) for the unconstrained optimization problem when the batch size is 15. For the constrained optimization problem, our proposed algorithm can speed up the optimization process by up to 15× compared to the weighted expected improvement based Bayesian optimization (WEIBO) approach, when the batch size is 15.

Area Optimisation of Two Stage Miller Compensated Op-Amp in 65 nm Using Hybrid PSO

“…It can be seen that the proposed algorithm gives a better optimal area than the ones reported in [12], [19]. The proposed method has been used for the design of the op-amp reported in [7] for an unconstrained optimisation in 180 nm technology and the comparison with the reported Bayesian optimisation method is given in Table V. It shows that the proposed technique converges to a higher fitness function value showing a better performance.…”

“…Since the state-of-the-art transistors have really complex mathematical models, the problem becomes even more challenging. In recent studies [6], [7], Bayesian optimisation methods have been reported for automated analog circuit sizing. Another class of algorithms, called the evolutionary algorithms (EAs) have been reported to be used in analog circuit optimisation.…”

Automatic Design of a Broadband Directional Coupler via Bayesian Optimization

“…Local BO, sparse GPR model [85] GNN + WEI Amp (C-2, D-8), driver (C-1, D-6) Parasitic-aware, GNN w/ dropout [86] GPR + Acq ensemble Amp-1,2,3 (C-10, C-12, C-24), etc., Batch BO enabled by the ensemble [87] Add-GPR + UCB Amp, DC (C-na) LDE-aware, high-dimensional [89] GPR + WEI Op-Amp (C-na) Bi-level BO, compensation design [90] GPR + EI Amp-1,2 (C-10, C-12) Local penalization 1 [92] GPR + modified TS Amp-1,2 (C-11, C-43) Applied to technology migration [105] Online GPR Op-Amp-1,2 (C-11, C-21) Self-adaptive incremental learning [102] GPR + wPESC Amp-1,2 (C-10, C-11) Automatically choose test benches [103] GPR + EIM Op-Amp-1,2 (C-11, C-21) Asynchronous BO [104] Online GPR + EIM Op-Amp-1,2 (C-11, C-26), etc., Self-adaptive incremental learning [100] GPR + LCB/EI CP (C-36), Amp (C-12) Search in one-dimensional subspace [101] MT-GPR + EI Transformer (C-4), LNA (C-15), etc., Multitask NN as GPR kernel [229] GPR + WEI 5 OTAs, 2 VCOs, 2 SCFs (D-na), etc., Wire sizing, GPR guided by GNN [96] GPR + LCB Voltage regulator (C-17 + D-10), etc., Novel evolutionary algorithm [97] GPR + TS LNA (C/D-17) 2…”

Gaussian Process Regression Based Multi-Objective Bayesian Optimization for Power System Design

“…While a broad variety of MOO approaches in black box environments deals with effective algorithms, only few of them meet the efficiency criterion [2]. Specifically Bayesian Optimization (BO) [3][4][5][6] algorithms based on Gaussian Process Regression (GPR) [3,[7][8][9] appear as interesting effective and efficient MOO candidates.…”

“…By Theorem 3, we obtain a Gaussian Process for each i. In case the covariance function involves the choice of some hyperparameter, we determine that parameter by solving Equation (5). Next, we condition each mean m i and covariance function C i to T using Equations ( 2) and (3), respectively.…”