1Z0-184-25 Question 83
Single answerYour data science team has stored text embeddings in an Oracle Autonomous Database table as a vector column named EMBEDDING and created a vector index for approximate similarity searches. However, early query tests show slower than expected response times when retrieving top matches. Which approach best improves performance while balancing accuracy for these approximate similarity searches?
- A
Increase the number of requested nearest neighbors for each query to widen the result set
- B
Use an indexing parameter or configuration that optimizes the trade-off between recall and query performance
- C
Convert approximate vector searches to exact searches to ensure deterministic results
- D
Increase the dimensionality of the vector embeddings to improve precision
Show answer and explanation
Correct answer: B
Explanation
To optimize approximate similarity searches, it is crucial to tune the vector index configuration to balance recall and performance. Oracle Database and Autonomous Database support parameters that allow fine-grained control over index accuracy versus speed. By adjusting these parameters, you can achieve faster query times without sacrificing critical accuracy. For more information, refer to Oracle documentation on Vector Indexes in Autonomous Database.
- A. Incorrect.
Option 1: Increasing the number of nearest neighbors (e.g., from 5 to 50 or more) often increases query-time complexity because the system must search deeper into the index. While you might gain more comprehensive results, it typically slows down queries rather than speeding them up.
- B. Correct.
Option 2: Correct. Oracle� vector index supports parameters (such as defining search thresholds or control parameters) that balance search accuracy (recall) and query performance. Tuning these parameters can significantly improve performance while maintaining an acceptable level of similarity accuracy.
- C. Incorrect.
Option 3: Converting to exact searches removes the approximations that speed up queries. Exact searches typically require more exhaustive lookups, leading to slower queries, which contradicts the need for improved performance.
- D. Incorrect.
Option 4: Increasing the vector dimensionality generally makes indexing and query operations more computationally intensive. This approach would likely degrade performance rather than improve it.