Loading [MathJax]/extensions/MathZoom.js
An Evolutionary Strategy Based Training Optimization of Supervised Machine Learning Algorithms (EStoTimeSMLAs) | IEEE Conference Publication | IEEE Xplore

An Evolutionary Strategy Based Training Optimization of Supervised Machine Learning Algorithms (EStoTimeSMLAs)


Abstract:

Evolutionary strategy is increasingly used for optimization in various machine learning problems. It can scale very well, even to high dimensional problems, and its abili...Show More

Abstract:

Evolutionary strategy is increasingly used for optimization in various machine learning problems. It can scale very well, even to high dimensional problems, and its ability to globally self optimize in flexible ways provides new and exciting opportunities when combined with more recent machine learning methods. This paper describes a novel approach for the optimization of models with a data driven evolutionary strategy. The optimization can directly be applied as a preprocessing step and is therefore independent of the machine learning algorithm used. The experimental analysis of six different use cases show that, on average, better results are attained than without evolutionary strategy. Furthermore it is shown, that the best individual models are also achieved with the help of evolutionary strategy. The six different use cases were of different complexity which reinforces the idea that the approach is universal and not depending on specific use cases.
Date of Conference: 28-30 December 2022
Date Added to IEEE Xplore: 24 August 2023
ISBN Information:
Conference Location: Bangkok, Thailand

I. Introduction

The need to produce a better result when creating machine learning models is probably the biggest topic right now in machine learning. Improving models by a few percent is sometimes seen as groundbreaking. Even if it is only the case for a few very specific use cases, like defect object detection in quality assurance processes of manufacturing. No one is denying the advantages of optimizing the models for particular domains, where an improvement of a few percent could make a big difference. Another striking challenge in machine learning is the lack of sufficient training data. In general, it can quickly become a problem to create a good machine learning model if there is only a few training data available. On the one hand, this can lead to the problem of over fitting, if the models are trained too much with a small amount of data. On the other hand, and probably seen more often in practice, there simply is not enough data available to learn the general structure of the problem.

References

References is not available for this document.