liveBook · Manning

Rollback

As explained in Timeless Algorithms, The Foundational Ideas

A procedure for evaluating a decision tree in reverse, combining probabilities and utilities at chance nodes and selecting the best branch at decision nodes.

When choosing whether to treat a patient under uncertainty, a decision tree lays out available actions, possible events, and their consequences. Rollback evaluates this tree from its endpoints toward its beginning: at chance nodes, it combines outcome utilities with their probabilities; at decision nodes, it selects the action with the highest expected utility. section 7.1.2

For a chance node, expected utility is the sum of each outcome’s probability multiplied by its utility. This makes the tree a way to compare specified actions under uncertainty, not a method for learning predictive rules from data. section 7.1.2 section 7.6

Treatment decision

A treatment tree assigns a 0.09 probability to disease and 0.91 to its absence. Rolling back gives expected utilities of −0.1 for treating and 91 for not treating, so the latter is preferred under the assigned probabilities and utilities. section 7.1.2

Figure 7.1 Decision tree for the treatment decision. The square node represents the choice between treating and not treating. Each action leads to a circular chance node representing whether disease is present (𝑝 = 0.09) or absent (𝑝 = 0.91). The terminal branches show the four possible outcomes and their assigned utilities. Rolling the tree back combines the probabilities and utilities to produce expected utilities of −0.1 for treating and 91 for not treating, making the latter the preferred action.

The tree displays the treatment choice, uncertain disease state, assigned utilities, and rollback results. section 7.1.2

How Timeless Algorithms, The Foundational Ideas covers it

How deep it goes
Defines reverse evaluation through probabilities and utilities, then illustrates it with expected utilities for two treatment choices.
Examples
Uses a treatment choice with disease probability 0.09 and compares expected utilities of −0.1 and 91.
What you can do after
Evaluate competing actions by comparing their expected utilities after rollback.
How it connects
Links decision trees to expected utility, Bayesian inference, and value-of-information analysis.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Timeless Algorithms, The Foundational Ideas. The section links point to the parts of the book it draws on.