Global Optimization of Deceptive Functions With Sparse Sampling
نویسندگان
چکیده
This paper introduces a new method of calculating the expected improvement infill criterion, which does not rely on accurate model parameter estimation. The parameter estimation is embedded within the search of the infill criterion, wherein parameter changes are assessed using likelihood ratio tests. Unlike the traditional expected improvement, a new formulation we present cannot be ‘fooled’ by unlucky sampling or deceptive functions. The new method is introduced both mathematically and illustratively using a one-variable test function. It is then shown to outperform traditional expected improvement when optimizing the geometry of a passive vibration isolating truss.
منابع مشابه
Fundamental Principles of Deception in Genetic Search
This paper presents several theorems concerning the nature of deception and the central role that deception plays in function optimization using genetic algorithms. A simple proof is ooered which shows that the only problems which pose challenging optimization tasks are problems that involve some degree of deception and which result in connicting k-arm bandit competitions between hyperplanes. T...
متن کاملMaximal elements of sub-topical functions with applications to global optimization
We study the support sets of sub-topical functions and investigate their maximal elements in order to establish a necessary and sufficient condition for the global minimum of the difference of two sub-topical functions.
متن کاملImproving Exploration in Evolution Strategies for Deep Reinforcement Learning via a Population of Novelty-Seeking Agents
Evolution strategies (ES) are a family of blackbox optimization algorithms able to train deep neural networks roughly as well as Q-learning and policy gradient methods on challenging deep reinforcement learning (RL) problems, but are much faster (e.g. hours vs. days) because they parallelize better. However, many RL problems require directed exploration because they have reward functions that a...
متن کاملGaussian Processes for Active Data Mining of Spatial Aggregates
Active data mining is becoming prevalent in applications requiring focused sampling of data relevant to a high-level mining objective. It is especially pertinent in scientific and engineering applications where we seek to characterize a configuration space or design space in terms of spatial aggregates, and where data collection can become costly. Examples abound in domains such as aircraft des...
متن کاملA Global Solution to Sparse Correspondence Problems
We propose a new methodology for reliably solving the correspondence problem between sparse sets of points of two or more images. This is a key step in most problems of computer vision and, so far, no general method exists to solve it. Our methodology is able to handle most of the commonly used assumptions in a unique formulation, independent of the domain of application and type of features. I...
متن کامل