Machine Learning Large Language Models Intermediate 1 min read

What is recall at k?

A metric used to evaluate the performance of a recommendation system or information retrieval model

recall at k explained in plain English

Recall at k measures the proportion of relevant items that are retrieved by a model within the top k recommendations or results

Analogy

Imagine you're searching for a specific book in a library, and the librarian gives you a list of 10 book recommendations. Recall at k is like checking how many of the relevant books (e.g., books by your favorite author) are in that list of 10

Example

A music streaming service uses recall at k to evaluate its song recommendation algorithm, checking how many songs by an artist that a user likes are included in the top 10 recommended songs

How is recall at k used?

Recall at k is used to evaluate the effectiveness of a model in retrieving relevant information or making accurate recommendations, and to compare the performance of different models

Common misconceptions about recall at k

Recall at k is often confused with precision at k, which measures the proportion of relevant items among the top k recommendations, rather than the proportion of relevant items that are retrieved

History

Recall at k originated in the field of information retrieval, where it was used to evaluate the performance of search engines and other retrieval systems

Also known as

recall rate recall metric r@k