🎁  Self-Information of Popularity

Self-Information of Popularity#

An information-theoretic weighting that rewards recommending less popular items.

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

What it is#

A novelty measure borrowed from information theory. The self-information (surprisal) of recommending an item is the negative base-2 logarithm of its popularity — the probability that a random user has interacted with it. Rare events carry more information, so rare items score high.

The formula#

\[\text{self-information}(i) = -\log_2\!\left(\frac{\text{count}(i)}{|U|}\right),\]

where \(\text{count}(i)\) is the number of users who consumed item \(i\) and \(|U|\) is the total number of users. A metric averages this over a top-\(N\) list and across users.

What it captures#

Popular items (high probability) have low self-information — they are unsurprising; long-tail items have high self-information — they are novel. It quantifies how much a recommendation tells the user something new.


Theme: Recommender Systems  ·  All terminology



See also

Source article Adapted (context, re-expressed) in our own words from: Self-Information of Popularity (insightful-data-lab.com).

Tags: purpose: reference topic: terminology level: advanced