Exploring the Turning Points for Intelligent Traversal with Q-Learning Models

The real reason behind a success or failure need not be the decisions or actions taken in the immediate past, but can be an event occurred in the distant past, as mentioned in Blame Attribution problems. The crucial action points capable of influencing or even deciding the ultimate nature of future outcomes are known as Turning Points. The turning points in real life or gaming situations are similar to junctions of roads diverting to one among the more favorable, less favorable or unfavorable destinations. Intelligent systems with adaptive learning capabilities can help to identify the real turning points using score values assigned to various states transition possibilities during previous epochs. This is a study on attaining knowledge about the turning points through repeated alternate exploration epochs using a basic Q-Learning model of a maze and the scope of exploiting this knowledge during the decision making situations in the future which demand crucial selection among multiple alternative options.