liveBook · Manning

Linear model

As explained in Applied Reinforcement Learning

A linear model represents an outcome as a weighted linear combination of input features.

A linear model replaces separate state or state–action value entries with a compact, adjustable function. It combines feature descriptions with learned parameters to estimate values across many states, rather than memorizing each one. section 8.1

Choose a model to match the input representation:

  • Use linear models when you can engineer interpretable features. section 8.6
  • Use deep networks for raw, high-dimensional inputs such as images or sensor data. section 8.6

Why approximate values?

Radiotherapy planning illustrates why tabular storage becomes impractical when realistic state spaces cannot be explicitly enumerated. section 8.1

Figure 8.2 Learning as parameter adjustment: fitting a linear model by turning the θ knobs

The parameter-adjustment diagram presents learning as turning model parameters rather than filling separate table entries. section 8.1

How Applied Reinforcement Learning covers it

How deep it goes
Explains how parameterized value estimates replace lookup tables and generalize across states.
Examples
Uses radiotherapy planning and image-valued states to motivate the limits of tabular storage.
What you can do after
Evaluate when engineered features favor linear models over deep networks for raw inputs.
How it connects
Connects linear function approximation to tabular reinforcement learning and deep neural networks.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Applied Reinforcement Learning. The section links point to the parts of the book it draws on.