Vector Search in PolyDBMS systems (Bachelor Thesis, Ongoing)
Author
Description
Classical database systems use boolean search. An tuple, document or element either
matches a predicate or it does not. Multimedia data has seen an exponential growth over
recent years and cannot be efficiently queried with only boolean search, but also requires
vector search. In vector search, the database system returns vectors that are close to a
target vector. Vector search has many important applications such as multimedia retrieval
and retrieval augmented generation.
PolyDBMSs offer unique advantages for these applications by not only combining different
search types (boolean and vector) but also by being able to combine different data models.
For example a multimedia retrieval system can store common metadata in the relational
model while additional metadata is stored in the document model.
The objective of this thesis is to design, implement and extend vector search support
tailored to the specific challenges posed by PolyDBMS, with a focus on the Polypheny
system.
Start / End Dates
2026/03/02 - 2026/07/01