Meta-learning in Artificial Intelligence: Syatematic Literature Review and Its Application in Simulated Environments

Abstract

Currently, the increasing development and training of artificial intelligence, using various machine learning algorithms present one of the biggest challenges for those involved in their development, namely; selecting the right learning algorithm for each problem. This is because selecting the correct learning algorithm traditionally relies on expert knowledge of many different machine learning algorithms which can take considerable amounts of time to experiment with. although the use of meta-learning has been proposed as an alternative method to automate this task by utilizing past experience. there are still issues related to the feasibility of various meta-learning methods and how theoretical applications relate to practical implementations. therefore, this study will seek to overcome these deficiencies and through a systematic literature review, assess the current state of meta-learning methodologies to determine which methodology(s) have the potential to be both the most effective and feasible for future use. to further support these findings, an experimental prototype application will be proposed that utilizes the methodology(ies) identified as being most effective in simulated environments under various conditions, providing evidence for the importance of autonomous decision making when it comes to AI based learning.