1Z0-184-25 Question 77
Single answerYou have an Oracle Autonomous Database table storing 50 million image embeddings (each represented as a high-dimensional vector). You initially created an approximate vector index to quickly identify similar images. However, your data science team reports that some valid matches are occasionally missed. They now require a true exact similarity search that returns every relevant embedding. Which approach best fulfills this requirement using Oracle Autonomous Database?
- A
- Increase the recall settings on the approximate index to guarantee perfect matches.
- B
- Rebuild the vector index in exact mode or leverage a full table scan with built-in distance functions to ensure no matches are skipped.
- C
- Convert the vector data into JSON documents and utilize Oracle Search Service to retrieve exact results.
- D
- Move the embeddings to a non-relational store and create a custom similarity algorithm outside of Oracle Autonomous Database.
Show answer and explanation
Correct answer: B
Explanation
For exact similarity searching in Oracle Autonomous Database�especially when approximate indexes are insufficient�either reconfigure your vector index in exact mode or use a full table scan with the built-in distance functions. Approximate indexes trade some accuracy for speed, so if consistently returning all relevant items is critical, an exact search approach or scanning all vectors is necessary. Refer to the official Oracle Database documentation on vector search features to learn how to create or rebuild indexes and use the distance functions for guaranteed accuracy.
- A. Incorrect.
Incorrect. Tuning recall settings on an approximate index can improve coverage, but it cannot strictly guarantee the retrieval of every single relevant match. Approximate indexing, by definition, can skip certain results in favor of faster queries.
- B. Correct.
Correct. Oracle Autonomous Database supports exact vector queries either by creating (or rebuilding) an index with exact search parameters or performing a full table scan using the built-in vector distance functions. Both approaches ensure no potential matches are skipped.
- C. Incorrect.
Incorrect. While Oracle Search Service can help with keyword or metadata lookups, it is not designed for high-dimensional exact similarity searches on vectors. Converting vectors into JSON does not inherently solve the requirement for exact matches.
- D. Incorrect.
Incorrect. Moving embeddings to another store and implementing a custom algorithm is possible but adds unnecessary complexity. Oracle Autonomous Database already provides mechanisms for exact similarity searches, so this option is not an efficient choice.