RT Journal Article
SR Electronic
T1 A neural algorithm for a fundamental computing problem
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 180471
DO 10.1101/180471
A1 Dasgupta, Sanjoy
A1 Stevens, Charles F.
A1 Navlakha, Saket
YR 2017
UL http://biorxiv.org/content/early/2017/08/25/180471.abstract
AB Similarity search, such as identifying similar images in a database or similar documents on the Web, is a fundamental computing problem faced by many large-scale information retrieval systems. We discovered that the fly’s olfac-tory circuit solves this problem using a novel variant of a traditional computer science algorithm (called locality-sensitive hashing). The fly’s circuit assigns similar neural activity patterns to similar input stimuli (odors), so that behav-iors learned from one odor can be applied when a similar odor is experienced. The fly’s algorithm, however, uses three new computational ingredients that depart from traditional approaches. We show that these ingredients can be translated to improve the performance of similarity search compared to tra-ditional algorithms when evaluated on several benchmark datasets. Overall, this perspective helps illuminate the logic supporting an important sensory function (olfaction), and it provides a conceptually new algorithm for solving a fundamental computational problem.