liveBook · Manning

Gain

As explained in AI Recommender Systems

Gain is a relevance value assigned to an item, often binary or graded, that contributes to a ranking-quality score.

Gain is the relevance value assigned to a ranked item. In the simplest case, consumed items receive gain 1 and other items receive 0; graded gain can use profit or ratings. section 4.6.3

Gain supports rank-aware evaluation by:

  • Contributing to cumulative gain across recommendation positions. section 4.6.3
  • Combining with position discounts so relevant items near the top contribute more. section 4.6.3
  • Providing the relevance input for DCG and normalized DCG. section 4.6.3

Same gain, different ranking quality

Two recommendation lists can contain the same single relevant item and therefore share cumulative gain of 1. The list placing it at position 1 is better than one placing it at position 7 because discounting rewards earlier positions. section 4.6.3

Figure 4.14 A simple way to calculate Gain is simply saying an item is relevant if the user consumed it or not. This means that the two lists of recommendations would have the same cumulative gain.

Use the figure to compare lists with identical cumulative gain before position discounting is applied. section 4.6.3

How AI Recommender Systems covers it

How deep it goes
Defines binary and graded gain, then connects it to cumulative and discounted ranking scores.
Examples
Uses consumed items, ratings, profit, and recommendation lists with relevant items at different positions.
What it assumes
Assumes a relevance judgment such as consumption, rating, or profit and teaches binary gain as the starting point.
What you can do after
Apply binary or graded relevance when constructing rank-aware recommendation metrics.
How it connects
Connects relevance to cumulative gain, discounted cumulative gain, and normalized discounted cumulative gain.

Related concepts

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