US 12,455,903 B1
Large-scale density-based clustering
Inbal Sagiv, Kfar Saba (IL); David Dyckman, Bet Shemesh (IL); and Oron Nir, Herzliya (IL)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on May 31, 2024, as Appl. No. 18/679,988.
Int. Cl. G06F 16/28 (2019.01)
CPC G06F 16/285 (2019.01) 20 Claims
OG exemplary drawing
 
1. A system, comprising:
a processing system; and
a memory coupled to the processing system, the memory comprising computer executable instructions that, when executed by the processing system, causes the system to perform operations comprising:
receiving a dataset including embeddings of at least one of words, objects in an image, or objects in a video;
selecting, for the dataset, a first upper bound value and a first lower bound value of a neighborhood radius parameter of a density-based clustering algorithm;
identifying, using a modified ternary search algorithm based on near-unimodality of the neighborhood radius parameter as a variant of ternary search algorithm, an optimal neighborhood radius parameter value, based on the first upper bound value and the first lower bound value of the neighborhood radius parameter;
providing as output performing at least one of:
outputting the optimal neighborhood radius parameter value; or
outputting an optimal number of clusters within the dataset corresponding to the optimal neighborhood radius parameter value; and
based on the output, clustering the at least one of words, objects in the image, or objects in the video associated with the dataset.