Amortizing Maximum Inner Product Search with Learned Support Functions
DGX agentMaximum inner product search (MIPS) is a crucial subroutine in machine learning, requiring the identification of a vector taken within a database (the keys) that best aligns with a given query. We pro