1Z0-184-25 Question 76
Select 2Your e-commerce platform on Oracle Cloud Infrastructure stores a large dataset of product images, each represented by vector embeddings in an Oracle Database 23c environment. You need to guarantee that any query for a given image retrieves only truly identical matches, with no missed results. Which two design choices would ensure your system performs an exact similarity search for the image embeddings?
- A
Use only an approximate nearest neighbor index to achieve quick lookups
- B
Perform a re-check by scanning the full dataset after retrieving approximate matches
- C
Create an index designed for exact similarity lookups that filters on precise distance values
- D
Rely on Oracle Text indexes created on raw image data for high-precision similarity
Show answer and explanation
Correct answers: B, C
Explanation
Exact similarity search in Oracle Database 23c typically involves either creating precise vector indexes that compare distance values exactly or re-verifying approximate results via a full dataset scan. While approximate indexes are valuable for speed in most use cases, they do not guarantee perfect recall. References: Oracle Database 23c documentation on vector search and indexing best practices outlines methods for creating and tuning exact KNN (k nearest neighbors) indexes to achieve both performance and accuracy requirements.
- A. Incorrect.
Option 1: Use only an approximate nearest neighbor index to achieve quick lookups. INCORRECT. Approximate nearest neighbor (ANN) indexes almost always trade off accuracy for speed, meaning they risk missing some exact matches or returning near duplicates. Relying solely on an approximate index does not guarantee perfect recall.
- B. Correct.
Option 2: Perform a re-check by scanning the full dataset after retrieving approximate matches. CORRECT. One viable way to ensure no exact matches are missed is to use an ANN index for initial filtering, then perform a full table scan to confirm the results precisely. This final pass verifies that only truly identical vectors are returned.
- C. Correct.
Option 3: Create an index designed for exact similarity lookups that filters on precise distance values. CORRECT. Oracle Database 23c can maintain specialized vector indexes that support exact distance-based lookups. This allows for high-accuracy searching without approximations, ensuring no false negatives.
- D. Incorrect.
Option 4: Rely on Oracle Text indexes created on raw image data for high-precision similarity. INCORRECT. Oracle Text indexes are meant primarily for text-based data. Storing and indexing raw images for similarity search within Oracle Text is not suitable for exact vector-based matching. Image vectors need specialized index structures.