Résumé :

In this talk, I will discuss accelerating similarity searches and related proximity search problems using the GPU. Similarity searches are fundamental to many fields and areas, including databases, machine learning, and large scale data analytics applications. These fields require information regarding nearby points (or feature vectors) to a query point in a data space.
But, as the volume and dimensionality of modern datasets continues to increase largely as a result of advances in scientific instrumentation, these searches become intractable using standard sequential algorithms. The GPU has very high memory bandwidth and computational throughput, and therefore, seems to be well-suited to processing similarity searches. However, similarity searches have data-dependent properties and so performance may be limited on the GPU due to its Single Instruction Multiple Thread (SIMT) execution model. I will discuss my work on parallel similarity searches and related algorithms, and describe several lessons learned over the years.
Lien VISIO : : https://teams.microsoft.com/l/meetup-join/19%3ameeting_NDJjODg2ZmYtZjlmNy00MjJkLWExNzEtMTE1MmVmODg5Y2Q5%40thread.v2/0?context=%7b%22Tid%22%3a%22967236d1-9003-4f1a-9556-8afe047945f1%22%2c%22Oid%22%3a%2215339270-20ac-4bfb-90c9-53d2645bc2a6%22%7d