Custom Similarity Metric Coding | Dataford Interview Questions

TELUS Digital AI Community uses vector representations when evaluating semantic similarity. Implement a custom weighted cosine similarity metric that gives each vector dimension a configurable importance.

Formal Specification

Write weighted_similarity(a, b, weights) to accept three lists of equal length containing real numbers. For each dimension i, weights[i] must be nonnegative. Compute:

similarity = Σ(weights[i] * a[i] * b[i]) / (sqrt(Σ(weights[i] * a[i]^2)) * sqrt(Σ(weights[i] * b[i]^2)))

Return a Python float. Dimensions with weight 0 must not affect the result. If either weighted vector has zero magnitude, return 0.0 instead of raising an error. The result should be in [-1.0, 1.0] for valid inputs.

Constraints

1 <= len(a) = len(b) = len(weights) <= 10^5
-10^6 <= a[i], b[i] <= 10^6
0 <= weights[i] <= 10^6
At least one weight is positive
Inputs contain finite numeric values

Function Signature

def weighted_similarity(a, b, weights):

Your question is Custom Similarity Metric Coding. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem