Filters








1,872 Hits in 3.5 sec

Preferential Batch Bayesian Optimization [article]

Eero Siivola, Akash Kumar Dhaka, Michael Riis Andersen, Javier Gonzalez, Pablo Garcia Moreno, Aki Vehtari
2021 arXiv   pre-print
In this work, we present preferential batch Bayesian optimization (PBBO), a new framework that allows finding the optimum of a latent function of interest, given any type of parallel preferential feedback  ...  Most research in Bayesian optimization (BO) has focused on direct feedback scenarios, where one has access to exact values of some expensive-to-evaluate objective.  ...  In this work, we introduce a method called preferential batch Bayesian optimization (PBBO) that allows optimizing black-box functions with BO when one can query preferences in a batch of input locations  ... 
arXiv:2003.11435v3 fatcat:46lk44xhpjb5lbxpjmirztkvza

Efficient Exploration in Binary and Preferential Bayesian Optimization [article]

Tristan Fauvel, Matthew Chalk
2021 arXiv   pre-print
Bayesian optimization (BO) is an effective approach to optimize expensive black-box functions, that seeks to trade-off between exploitation (selecting parameters where the maximum is likely) and exploration  ...  We then generalize these acquisition rules to batch learning, where multiple queries are performed simultaneously.  ...  Figure S6 : Batch preferential Bayesian optimization. A. With batches of size 4. Note that the MUC acquisition rule is more explorative than KSS. B. With batches of size 25.  ... 
arXiv:2110.09361v1 fatcat:agwbbf3dn5anrdh2y6g57mqzq4

Efficient and Scalable Batch Bayesian Optimization Using K-Means [article]

Matthew Groves, Edward O. Pyzer-Knapp
2018 arXiv   pre-print
We present K-Means Batch Bayesian Optimization (KMBBO), a novel batch sampling algorithm for Bayesian Optimization (BO).  ...  We show in empirical experiments that our method outperforms the current state-of-the-art batch allocation algorithms on a variety of test problems including tuning of algorithm hyper-parameters and a  ...  Summary We propose a novel batch sampling algorithm for Bayesian optimization based upon K-means, K-means Batch Bayesian Optimization (KMBBO).  ... 
arXiv:1806.01159v2 fatcat:mgpugddyjbbgnknoq42kjsrkpq

Bayesian Optimisation for Sequential Experimental Design with Applications in Additive Manufacturing [article]

Mimi Zhang, Andrew Parnell, Dermot Brabazon, Alessio Benavoli
2021 arXiv   pre-print
Bayesian optimization (BO) is an approach to globally optimizing black-box objective functions that are expensive to evaluate.  ...  , review all the applications of BO in additive manufacturing, compare and exemplify the features of different open BO libraries, unlock new potential applications of BO to other types of data (e.g., preferential  ...  The prototype achieved near-ideal production of the user-defined geometry within 100 experiments. 4 Review of Bayesian Optimization Software As of this writing, there are a variety of Bayesian optimization  ... 
arXiv:2107.12809v3 fatcat:fvw3dmx2s5azpje5cs6dmhqkni

Deep Bayesian Active Learning for Natural Language Processing: Results of a Large-Scale Empirical Study [article]

Aditya Siddhant, Zachary C. Lipton
2018 arXiv   pre-print
We find that across all settings, Bayesian active learning by disagreement, using uncertainty estimates provided either by Dropout or Bayes-by Backprop significantly improves over i.i.d. baselines and  ...  Our objective in optimizing the variational parameters is to minimize the KL divergence between q(θ) and p(w|D).  ...  For sequences, we look at agreement on the entire sequence tag, noting that this may exhibit a bias to preferentially sample longer sentences.  ... 
arXiv:1808.05697v3 fatcat:jnjghvzurrb37hr2z7be2lkngm

Bayesian Optimisation for Adaptive Experimental Design: A review

Stewart Greenhill, Santu Rana, Sunil Gupta, Pratibha Vellanki, Svetha Venkatesh
2020 IEEE Access  
INDEX TERMS Bayesian methods, design for experiments, design optimization, machine learning algorithms. 13938 VOLUME 8, 2020  ...  Solutions are surveyed for a range of core issues in experimental design including: the incorporation of prior knowledge, high dimensional optimisation, constraints, batch evaluation, multiple objectives  ...  His recent research in optimization using small data (Bayesian optimization) has found applications in efficient experimental design of products and processes in advanced manufacturing, such as alloy design  ... 
doi:10.1109/access.2020.2966228 fatcat:f4vz4jvjczc4dlg2twoqlrbtiy

Deep Bayesian Active Learning for Natural Language Processing: Results of a Large-Scale Empirical Study

Aditya Siddhant, Zachary C. Lipton
2018 Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing  
We find that across all settings, Bayesian active learning by disagreement, using uncertainty estimates provided either by Dropout or Bayes-by-Backprop significantly improves over i.i.d. baselines and  ...  Our objective in optimizing the variational parameters is to minimize the KL divergence between q(θ) and p(w|D).  ...  For sequences, we look at agreement on the entire sequence tag, noting that this may exhibit a bias to preferentially sample longer sentences.  ... 
doi:10.18653/v1/d18-1318 dblp:conf/emnlp/SiddhantL18 fatcat:q35rj2fj55g5halebna4yo7sqq

Generative Melody Composition with Human-in-the-Loop Bayesian Optimization

Yijun Zhou, Yuki Koyama, Masataka Goto, Takeo Igarashi
2020 Zenodo  
In this work, we present an interactive system that supports generative melody composition with human-in-the-loop Bayesian optimization (BO).  ...  This system takes a mixed-initiative approach; the system generates candidate melodies to evaluate, and the user evaluates them and provides preferential feedback (i.e., picking the best melody among the  ...  Related Work Human-in-the-Loop Bayesian Optimization BO is a black-box optimization technique (Shahriari et al., 2016) .  ... 
doi:10.5281/zenodo.4285363 fatcat:wwpy75aolrfuvkf45wjjakfmxi

Generative Melody Composition with Human-in-the-Loop Bayesian Optimization [article]

Yijun Zhou, Yuki Koyama, Masataka Goto, Takeo Igarashi
2020 arXiv   pre-print
In this work, we present an interactive system that supports generative melody composition with human-in-the-loop Bayesian optimization (BO).  ...  This system takes a mixed-initiative approach; the system generates candidate melodies to evaluate, and the user evaluates them and provides preferential feedback (i.e., picking the best melody among the  ...  Related Work Human-in-the-Loop Bayesian Optimization BO is a black-box optimization technique (Shahriari et al., 2016) .  ... 
arXiv:2010.03190v1 fatcat:3mjh476ldreytbea7zruzgz74y

Multi-Objective Bayesian Optimization for Accelerator Tuning [article]

Ryan Roussel, Adi Hanuka, Auralee Edelen
2021 arXiv   pre-print
number of measurements to converge to a useful solution.Here, we introduce a multi-objective Bayesian optimization scheme, which finds the full Pareto front of an accelerator optimization problem efficiently  ...  Usually, accelerator optimization problems are solved offline, prior to actual operation, with advanced beamline simulations and parallelized optimization methods (NSGA-II, Swarm Optimization).  ...  MULTI-OBJECTIVE BAYESIAN OPTIMIZATION We begin the explanation of MOBO by first briefly going over single-objective Bayesian optimization.  ... 
arXiv:2010.09824v2 fatcat:hugy6kp4wzbnhe6lv7vua2xdim

Deep Ensemble Bayesian Active Learning : Addressing the Mode Collapse issue in Monte Carlo dropout via Ensembles [article]

Remus Pop, Patric Fulop
2018 arXiv   pre-print
This method improves upon the current state-of-the-art deep Bayesian active learning method, which suffers from the mode collapse problem.  ...  Deep Bayesian Active Learning.  ...  One key difference between DAL and classical AL is the sampling in batches, which is needed to keep computational costs low.  ... 
arXiv:1811.03897v1 fatcat:4n6psgcze5gwtnv7bnibo6kztq

Structure learning in action

Daniel A. Braun, Carsten Mehring, Daniel M. Wolpert
2010 Behavioural Brain Research  
Dark blue colours indicate early batches, green colours intermediate batches, red colours indicate later batches.  ...  This shows that subjects' performance improves over batches. (C) Mean speed profiles for ±90 • rotations of the same batches.  ... 
doi:10.1016/j.bbr.2009.08.031 pmid:19720086 pmcid:PMC2778795 fatcat:bpuiesqu3jbq7o62dudimezcpq

Iterative experimental design based on active machine learning reduces the experimental burden associated with reaction screening

Natalie Suzanne Eyke, William H. Green, Klavs F Jensen
2020 Reaction Chemistry & Engineering  
Other techniques, including Bayesian optimization and genetic algorithms, have been successfully applied to reaction optimization and related problems.  ...  compared to smaller batch sizes.  ... 
doi:10.1039/d0re00232a fatcat:642jz6qmtvbtrhkay6qxbfpd6q

Choice functions based multi-objective Bayesian optimisation [article]

Alessio Benavoli and Dario Azzimonti and Dario Piga
2021 arXiv   pre-print
We then apply this surrogate model to solve a novel multi-objective Bayesian optimisation from choice data problem.  ...  By placing a Gaussian process prior on f and deriving a novel likelihood model for choice data, we propose a Bayesian framework for choice functions learning.  ...  Preferential Bayesian Optimisation (PBO) In many applications, evaluating g can be either too costly or not always possible.  ... 
arXiv:2110.08217v1 fatcat:m4vgw7hllbe7zfptkmuwlpytxy

Hot Swapping for Online Adaptation of Optimization Hyperparameters [article]

Kevin Bache, Dennis DeCoste, Padhraic Smyth
2015 arXiv   pre-print
We describe a general framework for online adaptation of optimization hyperparameters by 'hot swapping' their values during learning.  ...  This approach can also be contrasted to traditional hyperparameter search strategies such as grid search, random search, or Bayesian optimization which set optimization hyperparameters in an outer loop  ...  Instead, we propose to observe the optimization process under a variety of hyperparameter settings and to preferentially continue to use those settings which have performed best in the past.  ... 
arXiv:1412.6599v3 fatcat:xoem4o3tizekpmmazjnywxjksq
« Previous Showing results 1 — 15 out of 1,872 results