- Sum of the graded relevance values of all results in a search result list. CG at a particular rank position $p$ is $CG_p = \sum_{i=1}^{p} rel_i$, where $rel_i$ is the graded relevance of the result at position $i$.
- $rel_i$ is numeric, e.g., ranging between $[0, 1]$.
- Note: $CG_p$ is independent of each document’s position in the ranked result list. It only characterizes the documents ranked up to position $p$ as relevant or irrelevant to the query.
def calculate_cumulative_gain(relevance_scores):
cumulative_gain = []
current_sum = 0
for score in relevance_scores:
current_sum += score
cumulative_gain.append(current_sum)
return cumulative_gain
Cross-Validation