Title: Global Optimization for Big Data Classification using Simulated Annealing
Abstract: problem. Simulated annealing is a generic method of finding the global optimum of functions with many local minima. Unlike deterministic local search algorithms that always go downhill and thus get trapped in any local minimum, simulated annealing and other adaptive random search algorithms make random proposals, which sometimes accepts uphill steps with probability that decreases with a parameter called so that it will not get trapped and keep searching. Simulated annealing methods in the literature do not adjust the proposal distribution as the temperature changes, resulting in almost all proposals being rejected as the temperature goes to zero. Here we show that simulated annealing has better performance if the proposal variance is a linear function of temperature because this keeps the proposal acceptance rate about the same. Simulated Annealing
Publication Year: 2014
Publication Date: 2014-04-01
Language: en
Type: article
Access and Citation
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot