Local Bayesian Optimization For Analog Circuit Sizing

Touloupas, Konstantinos; Chouridis, Nikos; Sotiriadis, Paul P.

doi:10.1109/dac18074.2021.9586172

Cited by 16 publications

(5 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The common ground between GA MOO and TS efficient optimization is that both algorithms begin by randomly initializing some set of data points (Touloupas and Sotiriadis, 2021). However, the GA produces offsprings from parents through the cross-over operation and mutation, while in the TS algorithm, new data points are generated using sample functions from the Gaussian processes.…”

Section: Discussionmentioning

confidence: 99%

A socio-economic multi-objective optimization approach toward enhancing public–private partnerships’ concession agreements in infrastructure development

Alghamdi

2022

Journal of Strategic Contracting and Negotiation

View full text Add to dashboard Cite

The participants of the public–private partnership projects have different parts in developing the concessions, hence different perspectives and goals. Besides, the public–private partnership concession agreement has many parameters and components, and a change in one component will have a considerable impact on other components. Thus, determining the optimal value for different concession parameters by providing a series of feasible contribution options was investigated in this paper using two different multi-objective optimization models: genetic algorithm and Thompson sampling. Two case studies (the US I-495 and the I-4 Ultimate) were used to construct and validate the results of the model. The model results showed that the socio-economic sustainability performance increases as the private equity increases and the public equity decreases. The results also showed that the socio-economic sustainability performance increases as the concession price (user-fee) decreases. Having these contribution options could facilitate the decision-making process for both the public and private parties. The genetic algorithm model obtained faster optimization results when compared to the Thompson sampling model; however, the Thompson sampling obtained better results. Moreover, the use of the model could improve the socio-economic sustainability performance of the public–private partnership projects.

show abstract

Section: Discussionmentioning

confidence: 99%

A socio-economic multi-objective optimization approach toward enhancing public–private partnerships’ concession agreements in infrastructure development

Alghamdi

2022

Journal of Strategic Contracting and Negotiation

View full text Add to dashboard Cite

show abstract

“…There is a large body of work on BO for single-objective optimization [13,15]. Standard BO methods have been applied to a variety of problems including solving simple analog circuit design optimization and synthesis problems [27,33,31,49,26,34,42,32,41].…”

Section: Related Workmentioning

confidence: 99%

Preference-Aware Constrained Multi-Objective Bayesian Optimization

Ahmadianshalchi¹,

Belakaria²,

Doppa³

2023

Preprint

View full text Add to dashboard Cite

This paper addresses the problem of constrained multi-objective optimization over black-box objective functions with practitioner-specified preferences over the objectives when a large fraction of the input space is infeasible (i.e., violates constraints). This problem arises in many engineering design problems including analog circuits and electric power system design. Our overall goal is to approximate the optimal Pareto set over the small fraction of feasible input designs. The key challenges include the huge size of the design space, multiple objectives and large number of constraints, and the small fraction of feasible input designs which can be identified only after performing expensive simulations. We propose a novel and efficient preference-aware constrained multi-objective Bayesian optimization approach referred to as PAC-MOO to address these challenges. The key idea is to learn surrogate models for both output objectives and constraints, and select the candidate input for evaluation in each iteration that maximizes the information gained about the optimal constrained Pareto front while factoring in the preferences over objectives. Our experiments on two real-world analog circuit design optimization problems demonstrate the efficacy of PAC-MOO over prior methods.

show abstract

“…However, there is no way to yield exact, analytical expressions for the samples of GP models [13]. Typically, TS samples the values of the functions at a predefined, finitelength vector of input points [10], and determines the query point by finding the best value among them. However, in the case of high-dimensional variable spaces, this quantization approach is not practical since exponentially many points need to be considered to achieve good coverage.…”

Section: Algorithm 1: Bo Algorithmmentioning

confidence: 99%

“…Therefore, in this work, we use analytic approximations of the posterior samples using Random Fourier Features (RFF) [26], instead of relying on quantization. RFFs are a set of basis cosine functions φ(x) = [φ i (x] M i=1 that approximate the GP's kernel function as k x, x ≈ φ(x) T φ(x ) and can be also used to to approximate a sample from the GP as a linear model [26] f sample (x) ≈ φ(x) T θ, (10) where θ is an M-dimensional vector drawn from a Gaussian distribution. In this work, at each iteration, N s ≥ 1 approximate analytic samples are created through RFF, for each of the objective and the constraint functions.…”

Section: Algorithm 1: Bo Algorithmmentioning

confidence: 99%

“…Efforts to establish means of automation in the analog design cycle include mostly sizing automation approaches [5][6][7][8][9][10]. In this setting, the sizing procedure of a given analog circuit topology is formulated as a black-box optimization problem, where the design space includes all the unknown design variables and the optimization goals and constraints rely on the desired performance metrics.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Mixed-Variable Bayesian Optimization for Analog Circuit Sizing through Device Representation Learning

Touloupas

Sotiriadis

2022

Electronics

View full text Add to dashboard Cite

In this work, a deep representation learning method is proposed to build continuous-valued representations of individual integrated circuit (IC) devices. These representations are used to render mixed-variable analog circuit sizing problems as continuous ones and to apply a low-budget black box Bayesian optimization (BO) variant to solve them. By transforming the initial search spaces into continuous-valued ones, the BO’s Gaussian process models (GPs), which typically operate on real-valued spaces, can be used to guide the optimization search towards the global optimum. The proposed Device Representation Learning approach involves using device simulation data and training a composite model of a Variational Autoencoder (VAE) and a dense Neural Network. The latent variables of the trained VAE model serve as the representations of the integrated device and replace the discrete-valued parametrizations of particular devices. A thorough explanation of the proposed methodology’s mathematical formulation is given and example sizing applications on real-world analog circuits and integrated devices underline its efficiency.

show abstract

Local Bayesian Optimization For Analog Circuit Sizing

Cited by 16 publications

References 12 publications

A socio-economic multi-objective optimization approach toward enhancing public–private partnerships’ concession agreements in infrastructure development

A socio-economic multi-objective optimization approach toward enhancing public–private partnerships’ concession agreements in infrastructure development

Preference-Aware Constrained Multi-Objective Bayesian Optimization

Mixed-Variable Bayesian Optimization for Analog Circuit Sizing through Device Representation Learning

Contact Info

Product

Resources

About