1Z0-184-25 Question 3
Select 2Your media analytics company is transitioning from a traditional relational model to a vector-based indexing system in Oracle Cloud Infrastructure (OCI). You have millions of video embeddings that need to be searched quickly for similarity scores. Which two practices should you follow to ensure high-performance and accurate vector similarity queries in this scenario?
- A
Use an approximate nearest neighbor (ANN) indexing strategy configured for your vector data
- B
Store vectors as human-readable JSON documents for faster text-based searches
- C
Normalize embeddings so that vectors have consistent magnitudes before indexing
- D
Rely solely on conventional B-tree indexes for vector similarity queries
Show answer and explanation
Correct answers: A, C
Explanation
In OCI, storing vector embeddings typically involves using data formats and indexing approaches optimized for high-dimensional data. Approximate nearest neighbor techniques enable fast similarity searches, while normalization helps maintain consistent vector magnitudes. For further guidance, refer to Oracle documentation on best practices for deploying vector-based analytics workloads in OCI.
- A. Correct.
Correct. ANN indexing is a common technique for handling high-dimensional vector data efficiently, offering rapid approximate similarity search across large datasets.
- B. Incorrect.
Incorrect. While storing vectors as JSON might be convenient for human readability, it is not efficient for large-scale similarity searches, as parsing overhead and lack of specialized indexing degrade query performance.
- C. Correct.
Correct. Normalizing embeddings (often by L2 normalization) ensures that vectors lie on a uniform scale, improving the accuracy and speed of similarity measurements, such as cosine similarity.
- D. Incorrect.
Incorrect. B-tree indexes are designed for traditional scalar data and cannot efficiently handle high-dimensional vector similarity searches, leading to suboptimal query performance.