1Z0-184-25 Question 81
Select 2Your e-commerce platform uses an Oracle Database 23c instance on Oracle Cloud Infrastructure to store high-dimensional product embedding vectors. You want to implement approximate similarity search for real-time product recommendations. Which two statements are correct about optimizing a vector index for approximate searches in Oracle Database?
- A
Use an approximate indexing method (such as HNSW) with a dimension setting that matches your embedding size.
- B
Store vectors as integer arrays to reduce memory usage because floating-point types are not supported by approximate vector indexes.
- C
Refresh or rebuild the vector index when a large portion of the underlying data is updated or new vectors are added in bulk.
- D
Replicate the same vector data evenly across multiple identical indexes in different schemas to achieve the fastest query performance.
Show answer and explanation
Correct answers: A, C
Explanation
Approximate similarity search in Oracle Database 23c relies on dedicated indexing mechanisms (such as HNSW) to handle high-dimensional vector data efficiently. Matching the index dimension to your actual vector size and periodically refreshing the index after substantial data changes ensures both speed and accuracy. For more information, refer to the official Oracle Database 23c documentation on vector search and maintaining approximate vector indexes.
- A. Correct.
Correct. Oracle Database 23c supports approximate vector indexes (e.g., HNSW) where matching the index dimension to your embedding size enables efficient nearest neighbor lookups. This aligns with best practices for approximate vector search.
- B. Incorrect.
Incorrect. Oracle Database 23c supports floating-point vector columns, which are typically preferred over integer conversions. Integer-only storage can lose precision and is not required by approximate vector indexes, making this option a poor choice.
- C. Correct.
Correct. Approximate indexing structures may require rebuilding or refreshing after significant data updates or bulk inserts to maintain accuracy and performance. Failing to do so can degrade search quality over time.
- D. Incorrect.
Incorrect. Merely replicating identical indexes across different schemas does not improve individual query performance. Instead, it can add unnecessary overhead.