https://arshadhs.github.io/tags/semi-gradient-td/Recent content in Semi-Gradient TD on Arshad SiddiquiHugoen-ushttps://arshadhs.github.io/docs/ai/045-deep-reinforcement-learning/080-value-based-drl-function-approximation/Mon, 01 Jan 0001 00:00:00 +0000https://arshadhs.github.io/docs/ai/045-deep-reinforcement-learning/080-value-based-drl-function-approximation/<h1 id="value-based-drl-function-approximation"> Value-Based DRL: Function Approximation <a class="anchor" href="#value-based-drl-function-approximation">#</a> </h1> <p>Tabular methods store a separate value for every state or state-action pair. This becomes impractical when the state space is very large, continuous or represented by raw observations. Function approximation replaces this table with a parameterised function that can <strong>learn from observed states, generalise to related states and predict values for states not seen during training</strong>.</p> <p><strong>Course Content covered in this module:</strong></p> <ul> <li>Function approximation</li> <li>Feature Construction for Linear Methods: <ul> <li>Tile Coding</li> <li>Asymmetric Tile Coding</li> </ul> </li> <li>Linear function approximation</li> <li>Semi-Gradient TD methods</li> <li>Off-policy function approximation and TD divergence</li> <li>TD-Gammon</li> </ul> <span style="color: blue;"> <strong>Instead of memorising one value per state, the agent learns parameters shared across many states.</strong> </span> <hr> <h2 id="learning-objectives"> Learning Objectives <a class="anchor" href="#learning-objectives">#</a> </h2> <p>By the end of this module, you should be able to:</p>