Filters








464,953 Hits in 4.7 sec

Bayesian Optimization with a Finite Budget: An Approximate Dynamic Programming Approach

Rémi Lam, Karen Willcox, David H. Wolpert
2016 Neural Information Processing Systems  
We consider the problem of optimizing an expensive objective function when a finite budget of total evaluations is prescribed.  ...  We show how to approximate the solution of this dynamic programming problem using rollout, and propose rollout heuristics specifically designed for the Bayesian optimization setting.  ...  We propose to build such a heuristic using existing suboptimal Bayesian optimization strategies, in particular maximization of the expected improvement and minimization of the posterior mean.  ... 
dblp:conf/nips/LamWW16 fatcat:go6kfrqyl5dqxfvvmlcramilm4

Adapting control policies for expensive systems to changing environments

Matthew Tesch, Jeff Schneider, Howie Choset
2011 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems  
In this paper, we focus on the problem of learning a policy to adapt a system's controller based on the value of these external conditions in order to always perform well (i.e., maximize system output)  ...  We formally define the problem setup and the notion of an optimal control policy. We propose two algorithms which aim to find such a policy while minimizing the number of system output evaluations.  ...  By maximizing (6), we are choosing a point to evaluate which maximizes expectation of improvement in the policy score function, rather than choosing a point which maximizes improvement in one differential  ... 
doi:10.1109/iros.2011.6095039 dblp:conf/iros/TeschSC11 fatcat:3am733hhbvddpfjh4hqt2cb2lm

Adapting control policies for expensive systems to changing environments

M. Tesch, J. Schneider, H. Choset
2011 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems  
In this paper, we focus on the problem of learning a policy to adapt a system's controller based on the value of these external conditions in order to always perform well (i.e., maximize system output)  ...  We formally define the problem setup and the notion of an optimal control policy. We propose two algorithms which aim to find such a policy while minimizing the number of system output evaluations.  ...  By maximizing (6), we are choosing a point to evaluate which maximizes expectation of improvement in the policy score function, rather than choosing a point which maximizes improvement in one differential  ... 
doi:10.1109/iros.2011.6048689 fatcat:tt5g6jb4yjeb5j2hzmqnszl4qe

A Novel Single-DBN Generative Model for Optimizing POMDP Controllers by Probabilistic Inference

Igor Kiselev, Pascal Poupart
2014 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
The proposed approaches to policy optimization by probabilistic inference are evaluated on several POMDP benchmark problems and the performance of the implemented approximation algorithms is compared.  ...  As a promising alternative to using standard (often intractable) planning techniques with Bellman equations, we propose an interesting method of optimizing POMDP controllers by probabilistic inference  ...  Establishing this equivalence makes it further possible to maximize the objective function and compute the optimal policy parameter θ approximately using the Expectation-Maximization algorithm by varying  ... 
doi:10.1609/aaai.v28i1.9100 fatcat:qcwoem5sarb6fidqmp6ujgqbqa

Weed Control Decision Rules under Uncertainty

William Deen, Alfons Weersink, Calum G. Turvey, Susan Weaver
1993 Review of Agricultural Economics  
Under the assumption of maximizing expected utility, there are instances in which herbicide use increases with risk aversion as per conventional wisdom.  ...  and the optimal rate of application.  ...  maker is assumed to maximize expected utility on the basis of the maximin principle.  ... 
doi:10.2307/1349710 fatcat:vbtlklo6hfbfvabfzm4bhmrcae

Directional sensor control for maximizing information gain

Shankarachary Ragi, Hans D. Mittelmann, Edwin K. P. Chong, Oliver E. Drummond, Richard D. Teichgraeber
2013 Signal and Data Processing of Small Targets 2013  
We develop heuristic methods to solve the problem approximately and provide lower and upper bounds on the optimal information gain.  ...  We develop tractable solutions to the problem of controlling the directions of 2-D directional sensors for maximizing information gain corresponding to multiple targets in 2-D.  ...  We approximate the expectation by a Monte Carlo method.  ... 
doi:10.1117/12.2022451 fatcat:nixlggsr6ff4th4mfrvtmncpi4

Planning for robotic exploration based on forward simulation

Mikko Lauri, Risto Ritala
2016 Robotics and Autonomous Systems  
We address the problem of controlling a mobile robot to explore a partially known environment. The robot's objective is the maximization of the amount of information collected about the environment.  ...  Experimental results in simulated and real environments show that, depending on the environment, applying POMDP based planning for exploration can improve performance over frontier exploration.  ...  Olli Suominen for helpful discussions and comments on the manuscript, and Mr. Joonas Melin for assistance in carrying out the experimental work.  ... 
doi:10.1016/j.robot.2016.06.008 fatcat:t3ewax4shndudk2hsv2ygrvjay

Bayesian Design of Experiments Using Approximate Coordinate Exchange

Antony M. Overstall, David C. Woods
2016 Technometrics  
This methodology uses a Gaussian process emulator to approximate the expected utility as a function of a single design coordinate in a series of conditional optimization steps.  ...  The construction of decision-theoretical Bayesian designs for realistically complex nonlinear models is computationally challenging, as it requires the optimization of analytically intractable expected  ...  , December 2015; http://www.bode2015.wordpress.com) for useful discussions on extensions and future work.  ... 
doi:10.1080/00401706.2016.1251495 fatcat:j56h7dlw6ffr5lp7i45nuq4pvu

Risk-Sensitive Model Predictive Control [article]

Nicholas Moehle
2022 arXiv   pre-print
In both cases, we obtain a lower bound on the optimal cost as a by-product of solving the planning problem.  ...  unfavorably, depending on a risk aversion parameter.  ...  Evaluating our bound requires solving an optimization problem, which we use as the basis for a control policy we call risk-sensitive model predictive control (RS-MPC).  ... 
arXiv:2101.11166v2 fatcat:r5bqoqj3pfbz7gpisducru4qjq

Bayesian Optimization for Parameter of Discrete Weibull Regression

Adesina, Olumide Sunday, Onanaye, Adeniyi Samson, Okewole, Dorcas Modupe
2020 Journal of Advances in Mathematics and Computer Science  
This study aim at optimizing the parameter θ of Discrete Weibull (DW) regression obtained by maximizing the likelihood function.  ...  On implementing Bayesian optimization to obtain parameter new parameter θ of discrete Weibull regression using the known β, the results showed promising applicability of the technique to the model, and  ...  The general Bayesian optimization technique for a maximization problem aim at optimizing the following problem: The objective is to compute a maximizer of expensive-to-evaluate function In order to apply  ... 
doi:10.9734/jamcs/2019/v34i630233 fatcat:axlz54jvhbf7jnpkmcmy7sqzqa

Variational Bayesian Optimization for Runtime Risk-Sensitive Control

Scott Kuindersma, Roderic Grupen, Andrew Barto
2012 Robotics: Science and Systems VIII  
We extend recent work on variational heteroscedastic Gaussian processes to the optimization case to achieve efficient minimization of very noisy cost signals.  ...  We present a new Bayesian policy search algorithm suitable for problems with policy-dependent cost variance, a property present in many robot control tasks.  ...  Both (5) and its gradient, ∂EI(θ)/∂θ, are efficiently computable, so we can apply standard nonlinear optimization methods to maximize EI to select the next policy.  ... 
doi:10.15607/rss.2012.viii.026 dblp:conf/rss/KuindersmaGB12 fatcat:chig2eanzzhczaw5cyjwoixlce

Robust PID design by chance-constrained optimization

Pedro Mercader, Kristian Soltesz, Alfonso Baños
2017 Journal of the Franklin Institute  
The proposed method constitutes a stochastic extension to the well-studied maximization of integral gain optimization (MIGO) approach, i.e., maximization of integral gain under constraints on the H∞-norm  ...  The underlying chance-constrained optimization problem is solved using a gradient-based algorithm once it has been approximated by a deterministic optimization problem.  ...  Evaluating the mentioned cost, constraints and associated gradients, relies on the evaluation of multivariate expectation integrals, of the form (11) .  ... 
doi:10.1016/j.jfranklin.2017.10.017 fatcat:x346sngsknb5xeqfkp6gxciyku

Page 9760 of Mathematical Reviews Vol. , Issue 2003m [page]

2003 Mathematical Reviews  
Experimental results on large hypergraphs from VLSI applications show that the run time is reduced, on av- erage, by a factor approximately 2, while memory occupation is reduced, on average, by a factor  ...  The control action means introducing the predator into the habitat under a permanent cost per unit time, optimality means minimizing the expected long-run average cost per unit time, and a ‘control-limit  ... 

Active Contextual Entropy Search [article]

Jan Hendrik Metzen
2015 arXiv   pre-print
expects to learn the most.  ...  However, learning, when performed on real robotic systems, is typically restricted to a small number of trials.  ...  Bayesian optimization for contextual policy search (BO-CPS) is based on applying ideas from Bayesian optimization to contextual policy search [20] .  ... 
arXiv:1511.04211v2 fatcat:hkwzusi4mbghxayvevsh7scgjy

Soft Actor-Critic Algorithms and Applications [article]

Tuomas Haarnoja and Aurick Zhou and Kristian Hartikainen and George Tucker and Sehoon Ha and Jie Tan and Vikash Kumar and Henry Zhu and Abhishek Gupta and Pieter Abbeel and Sergey Levine
2019 arXiv   pre-print
Model-free deep reinforcement learning (RL) algorithms have been successfully applied to a range of challenging sequential decision making and control tasks.  ...  In this framework, the actor aims to simultaneously maximize expected return and entropy. That is, to succeed at the task while acting as randomly as possible.  ...  Maximum entropy reinforcement learning optimizes policies to maximize both the expected return and the expected entropy of the policy.  ... 
arXiv:1812.05905v2 fatcat:dw4325vci5ezzpfyfrtulkpko4
« Previous Showing results 1 — 15 out of 464,953 results