Showing 1–1 of 1 results for author: Pieter, V

Parameters for the best convergence of an optimization algorithm On-The-Fly

Authors: Valdimir Pieter

Abstract: What really sparked my interest was how certain parameters worked better at executing and optimization algorithm convergence even though the objective formula had no significant differences. Thus the research question stated: 'Which parameters provides an upmost optimal convergence solution of an Objective formula using the on-the-fly method?' This research was done in an experimental concept in w… ▽ More What really sparked my interest was how certain parameters worked better at executing and optimization algorithm convergence even though the objective formula had no significant differences. Thus the research question stated: 'Which parameters provides an upmost optimal convergence solution of an Objective formula using the on-the-fly method?' This research was done in an experimental concept in which five different algorithms were tested with different objective functions to discover which parameter would result well for the best convergence. To find the correct parameter a method called 'on-the-fly' was applied. I run the experiments with five different optimization algorithms. One of the test runs showed that each parameter has an increasing or decreasing convergence accuracy towards the subjective function depending on which specific optimization algorithm you choose. Each parameter has an increasing or decreasing convergence accuracy toward the subjective function. One of the results in which evolutionary algorithm was applied with only the recombination technique did well at finding the best optimization. As well that some results have an increasing accuracy visualization by combing mutation or several parameters in one test performance. In conclusion, each algorithm has its own set of the parameter that converge differently. Also depending on the target formula that is used. This confirms that the fly method a suitable approach at finding the best parameter. This means manipulations and observe the effects in process to find the right parameter works as long as the learning cost rate decreases over time. △ Less

Submitted 23 September, 2020; originally announced September 2020.

Comments: 28 pages