Diversity on the Go! Streaming Determinantal Point Processes under a Maximum Induced Cardinality Objective

De novo generated combinatorial library design

“…Recent research in usage of DPP has primarily been focused on the algorithmic efficiency while keeping a close-to-greedy performance in selecting diverse items. 118,120–123 For scenarios such as ours, the only selections of relevance are sets of fixed size, such as the same sizes as screening plates, i.e. , 96, 384 or 1536 for parallel library synthesis, or other fixed sizes determined by a project demand.…”

“…Recent research in usage of DPP has primarily been focused on the algorithmic efficiency while keeping a close-to-greedy performance in selecting diverse items. 118,[120][121][122][123] For scenarios such as ours, the only selections of relevance are sets of xed size, such as the same sizes as screening plates, i.e., 96, 384 or 1536 for parallel library synthesis, or other xed sizes determined by a project demand. k-DPPs are an extension of general DPPs that are conditioned to selected sets of size exactly k. Gharan and Rezaei 123 introduced a computationally efficient method for sampling k-DPPs using a Gibbs sampling scheme shown to have fast mixing properties.…”