Michael Bender
Stony Brook University, NY·Computer Science
Modern computing depends on organizing data so it can be found, updated and searched fast, even at huge scale. Two projects tackle this from different directions.
- Fields
Computer Science
Mathematics & Statistics
- Topics
algorithms
data structures
vector databases
- Grants
- 2 known grants, through Jun 2030
What they work on
- Design efficient methods to build graph-based vector search indexes such as HNSW and DiskANN, reducing their currently very high construction costs.
- Develop query algorithms for graph-based nearest-neighbor search with proven performance guarantees, and study graph navigability and how well local search works in graphs.
- Optimize vector databases for hierarchical memory and parallel hardware and combine graph search with vector compression, checking results in experiments and releasing open source code.
- Use history independence, storing data so its layout hides past operations, as an analysis tool for randomized list-labeling and dictionary data structures in RAM and external memory.
- Apply history-independence techniques to online algorithms, resource allocation, load balancing and scalable concurrent data structures.
More about the research
One asks how to search the numerical summaries that AI systems produce for documents and images. The other asks how a randomness-based idea called history independence can help design faster algorithms. Both mix mathematical proofs about algorithms with practical systems.
Why it matters
Cheaper, faster and more reliable vector search could let small businesses, scientists and other AI users without large cloud or hardware budgets run vector databases. It could also guide systems that handle the largest commercial and scientific datasets. New randomized data-structure techniques may speed up storage, load balancing and concurrent software.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Vinodchandran Variyam
University of Nebraska-Lincoln, NE·Computer Science
Initiate the study of a "right to forget" data stream model, in which items can be required to be deleted, motivated by privacy concerns and the growth of data generation.
Satish Puri
Missouri University of Science and Technology, MO·Computer Science·Early-career
Find which computing tasks and design settings gain the most speed on SmartSSD computational storage devices, which pair a drive with an FPGA chip, through a design space exploration on multicore systems.
Andrew McGregor
University of Massachusetts, MA·Computer Science
Develop adversarially robust data stream algorithms that stay accurate when inputs are chosen adaptively in response to the algorithm's earlier outputs, as happens in interactive data analysis.
The school
Stony Brook University: 32 labs on PI Pages, $319M a year on research, 289 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.