Reinforcement Learning: Function Approximation and Deep Q-Networks — Part 4

Author(s): Tan Pengshi Alvin

Originally published on Towards AI.

Reinforcement Learning with continuous state spaces and gradient descent techniques
Image by SpaceX on Unsplash

Since Part 1 of this series, we have framed Reinforcement Learning as a Markov Decision Process environment with discrete states and actions with their corresponding state-action values Q(s,a) stored in a tabular manner. This means that you can imagine a 2-dimensional table, with a row index presented by states and a column index represented by actions. The table entries are then the array of Q(s,a) values. This simplified paradigm has allowed us to gain intuition about how a decision-making agent learns as it traverses through a bite-sized environment.

In real-world applications, however, there can either be an… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI