ANN Vector Similarity Search | Dataford Interview Questions

Write a function to perform efficient vector similarity search using an approximate nearest neighbor (ANN) approach.

Implement cosine-similarity search with locality-sensitive hashing (LSH) using random hyperplanes. The function receives an array of vectors and one query vector, then returns the indices of the k nearest vectors in descending cosine-similarity order. Use the supplied seed to make the index deterministic. If fewer than k candidates are found in matching buckets, evaluate all vectors as a correctness fallback.

Signature: def ann_search(vectors, query, k, num_tables, hash_size, seed):

Vectors are non-empty numeric arrays with equal dimensions. Return a list of k integer indices.

Constraints

1 <= len(vectors) <= 100000
1 <= len(query) <= 128
All vectors have the same positive dimension as query
1 <= k <= len(vectors)
1 <= num_tables <= 32
1 <= hash_size <= 16
seed is an integer
Vector components are finite numbers

Function Signature

def ann_search(vectors, query, k, num_tables, hash_size, seed):

Your question is ANN Vector Similarity Search. 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