Kernel Approximation by Locality Sensitive Hashing with Application to Active Learning

Michel Crucianu (1)

1 CEDRIC - Centre d'études et de recherche en informatique et communications

Locality Sensitive Hashing (LSH) methods are being successfully employed for scaling similarity queries or similarity joins to large databases. We show that operations requiring intensive kernel computations can also benefit from the use of LSH. Specifically, this allows to accelerate active learning methods for which the query corresponds to a class boundary.

Mots clés: LSH; active learning; kernel approximation; scalable active learning; scalability

Mots clés: MSi; apprentissage actif; rapprochement du noyau; apprentissage actif évolutif; évolutivité

Domaines: Informatique [cs]; Base de données [cs.DB]

Michel Crucianu. Kernel Approximation by Locality Sensitive Hashing with Application to Active Learning. [Research Report] CEDRIC-10-1898, CEDRIC Lab/CNAM. 2010. ⟨hal-01125750⟩