Εφαρμοσμένα νευρωνικά δίκτυα σε ηλεκτρονικά παίγνια: μια μελέτη του εργαλείου ML – Agents στο περιβάλλον Unity
Applied neural networks in video games: a case study of ML – Agents in Unity
Mατσιόζεκ, Ιζαμπέλ - Βικτωρία
KeywordsUnity ; ML-Agents ; Μηχανική μάθηση ; Machine learning ; Multi-agent game ; Θεωρία παιγνίων ; Πολυπρακτορικά συστήματα ; Τεχνητή νοημοσύνη
Machine learning is one of the most rapidly growing branches of artificial intelligence, which in turn is part of computer science. Machine learning deals with the extraction of models and algorithms from data, in order to improve the computer's behaviour in some task based on experience, with the ultimate goal of providing a learning solution or answer.The subject of the current dissertation is complex and aims to present a part of this new scientific field through its applications in the field of video games. Initially, the concepts and terminology governing the mechanics of machine learning and video games should be understood through a bibliographic review of their theory (Chapters 1 & 2). Following, we introduce the ML-Agents toolkit for Unity, a platform that will be used as the reference point of this work (Chapter 3), present installation and usage instructions (Chapter 4), and then analyze a usage example of the toolkit for the creation of a multi-agent game (Chapter 5). Finally, we analyze the conclusions drawn from the implemented example and propose future improvements (Chapter 6).