Bayesian Optimization with a Finite Budget: An Approximate Dynamic Programming Approach
نویسندگان
چکیده
We consider the problem of optimizing an expensive objective function when a finite budget of total evaluations is prescribed. In that context, the optimal solution strategy for Bayesian optimization can be formulated as a dynamic programming instance. This results in a complex problem with uncountable, dimension-increasing state space and an uncountable control space. 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 present numerical experiments showing that the resulting algorithm for optimization with a finite budget outperforms several popular Bayesian optimization algorithms.
منابع مشابه
Mathematical Programming Approach To Allocate Local Or National Resources For Bridge Maintenance Rehabilitation & Replacement Planning (RESEARCH NOTE)
Today’s, the transportation facilities such as terminals, street, bridge, etc, represent the major investment in highway network. Every year tremendous resources should be invested to maintain these facilities. Among them, the Bridge Management System (B.M.S.) has been necessitated by large imbalance between extensive bridge repair and maintenance needs and limited available budget. So the main...
متن کاملA dynamic programming approach for solving nonlinear knapsack problems
Nonlinear Knapsack Problems (NKP) are the alternative formulation for the multiple-choice knapsack problems. A powerful approach for solving NKP is dynamic programming which may obtain the global op-timal solution even in the case of discrete solution space for these problems. Despite the power of this solu-tion approach, it computationally performs very slowly when the solution space of the pr...
متن کاملOPTIMIZATION OF A PRODUCTION LOT SIZING PROBLEM WITH QUANTITY DISCOUNT
Dynamic lot sizing problem is one of the significant problem in industrial units and it has been considered by many researchers. Considering the quantity discount in purchasing cost is one of the important and practical assumptions in the field of inventory control models and it has been less focused in terms of stochastic version of dynamic lot sizing problem. In this paper, stochastic dyn...
متن کاملConstrained Bayesian Reinforcement Learning via Approximate Linear Programming
In this paper, we consider the safe learning scenario where we need to restrict the exploratory behavior of a reinforcement learning agent. Specifically, we treat the problem as a form of Bayesian reinforcement learning in an environment that is modeled as a constrained MDP (CMDP) where the cost function penalizes undesirable situations. We propose a model-based Bayesian reinforcement learning ...
متن کاملCost Analysis of Acceptance Sampling Models Using Dynamic Programming and Bayesian Inference Considering Inspection Errors
Acceptance Sampling models have been widely applied in companies for the inspection and testing the raw material as well as the final products. A number of lots of the items are produced in a day in the industries so it may be impossible to inspect/test each item in a lot. The acceptance sampling models only provide the guarantee for the producer and consumer that the items in the lots are acco...
متن کامل