PhD seminar 06-12-2021

PhD Seminars 06-12-2021

Talk 1


Tareq Si Salem (NEO)


AÇAI: Ascent Similarity Caching with Approximate Indexes




Similarity search is of paramount importance in multimedia retrieval systems and recommender systems, augmented reality, and future machine learning applications. When these systems require to serve large objects for delay-sensitive demands, edge servers close to the end-user can operate as similarity caches to speed up the retrieval.
The paper "AÇAI: Ascent Similarity Caching with Approximate Indexes" proposes a new similarity caching policy which improves on the state of the art by using (1) a fast indexing mechanism to query a large catalog of 1 billion objects in less than 1ms, and (2) an online mirror ascent scheme to optimally cache objects even in the presence of adversarial queries.


  • December 6, 2021, 2:00 pm

Comments are closed.