Pre-computed memory or on-the-fly encoding? A hybrid approach to retrieval augmentation makes the most of your compute.
Michiel de JongYury ZemlyanskiyNicholas FitzGeraldJoshua AinslieSumit SanghaiFei ShaWilliam CohenPublished in: CoRR (2023)
Keyphrases
- pre computed
- image database
- information retrieval systems
- database
- information retrieval
- image retrieval
- test collection
- disk space
- triangle inequality
- document retrieval
- efficient retrieval
- text retrieval
- retrieval model
- retrieval systems
- data cube
- storage cost
- video sequences
- memory size
- low memory
- main memory
- random access
- retrieval accuracy
- retrieval process
- multimedia databases
- vector quantization
- query expansion
- neural network