Bayesian Optimization 详细讲义PPT.pdf

发布时间：2022-06-09 发布人：admin 分类：说明书资料大小：26.44M 资料格式：pdf 举报版权申诉

vickybai58-11155958-4744302543387847614.pdf-第1页.png

第1页 / 共78页

vickybai58-11155958-4744302543387847614.pdf-第2页.png

第2页 / 共78页

vickybai58-11155958-4744302543387847614.pdf-第3页.png

第3页 / 共78页

vickybai58-11155958-4744302543387847614.pdf-第4页.png

第4页 / 共78页

vickybai58-11155958-4744302543387847614.pdf-第5页.png

第5页 / 共78页

vickybai58-11155958-4744302543387847614.pdf-第6页.png

第6页 / 共78页

vickybai58-11155958-4744302543387847614.pdf-第7页.png

第7页 / 共78页

vickybai58-11155958-4744302543387847614.pdf-第8页.png

第8页 / 共78页

文本预览

Introduction to Bayesian Optimization Javier Gonz´alez Masterclass, 7-February, 2107 @Lancaster University

Big picture “Civilization advances by extending the number of important operations which we can perform without thinking of them.” (Alfred North Whitehead) We are interested on optimizing data science pipelines: Automatic model conﬁguration. Automate the design of physical experiments.

Agenda of the day 9:00-11:00, Introduction to Bayesian Optimization: What is BayesOpt and why it works? Relevant things to know. 11:30-13:00, Connections, extensions and applications: Extensions to multi-task problems, constrained domains, early-stopping, high dimensions. Connections to Armed bandits and ABC. An applications in genetics. 14:00-16:00, GPyOpt LAB!: Bring your own problem! 16:30-15:30, Hot topics current challenges: Parallelization. Non-myopic methods Interactive Bayesian Optimization.

Section I: Introduction to Bayesian Optimization What is BayesOpt and why it works? Relevant things to know.

Data Science pipeline/Autonomous System Challenges and needs for automation Data CollectionOptimal designFeatures extractionFilters, dimensionality reductionModellingModel tuning and configuration, code optimisationResults Interpretation/DecisionData visualisation,Sequential experimentationPipeline improvement / Interaction with environment

Experimental Design - Uncertainty Quantiﬁcation Can we automate/simplify the process of designing complex experiments? Emulator - Simulator - Physical system

Global optimization Consider a ‘well behaved’ function f : X → R where X ⊆ RD is a bounded domain. xM = arg min x∈X f (x). f is explicitly unknown and multimodal. Evaluations of f may be perturbed. Evaluations of f are expensive.

Expensive functions, who doesn’t have one? Parameter tuning in ML algorithms. Number of layers/units per layer Weight penalties Learning rates, etc. Figure source: http://theanalyticsstore.com/deep-learning

分享到：

赞收藏

资料库

Bayesian Optimization 详细讲义PPT.pdf

相关推荐

课程资源

热门标签

最新资料