Implement the MAP@K function by iterating through each query, slicing the top K retrieved items, checking if they exist in the ground truth, and calculating the mean average precision.

MAP@K is a variant of MAP that considers only the top K results for each query. The procedure is similar to MAP, but the ranked list is truncated to the first K items.

Algorithm:

1. For each query:

a. Take the top K items from the ranked list.

b. Initialize relevant_seen = 0, sum_precisions = 0.

c. For each rank k (1‑based) in the truncated list:

  - If the item at rank k is in the ground‑truth set, increment relevant_seen.
  - If relevant_seen > 0, add (relevant_seen / k) to sum_precisions.

d. If there are relevant items in the top K, AP@K = sum_precisions / min(total_relevant, K); otherwise AP@K = 0.

1. MAP@K = (sum of AP@K values) / number_of_queries.

Complexity: O(K * number_of_queries) time, as we only scan the first K items per query. Space usage is O(1) per query. This function is useful when only the top results matter, such as in search engines or recommendation systems.