US 11,727,250 B2
Elastic-centroid based clustering
Timoleon Moraitis, Glattpark (CH); and Abu Sebastian, Adliswil (CH)
Assigned to INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed by INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed on Sep. 6, 2019, as Appl. No. 16/563,811.
Prior Publication US 2021/0073616 A1, Mar. 11, 2021
Int. Cl. G06F 17/00 (2019.01); G06N 3/049 (2023.01); G06N 3/063 (2023.01); G06N 3/08 (2023.01); G06F 16/28 (2019.01)
CPC G06N 3/049 (2013.01) [G06F 16/285 (2019.01); G06N 3/063 (2013.01); G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method of pattern recognition using an elastic clustering algorithm for use in the recognition of occluded objects or of data from partial patterns, the method comprising:
assigning one or more input datapoints of a sequence of datapoints representing a group of data to a particular cluster of K clusters, based on a distance from a centroid k representing a center of the particular cluster;
in each of the K clusters:
clustering the datapoints based on their location relative to the centroid k;
shifting the centroid k from a first position to a second position determined to be closer than the first position to the sequence of input datapoints assigned to the particular cluster, over a predetermined time period, wherein the shifting is based on a distance between the sequence of input datapoints and the centroid k; and
relaxing a location of the centroid k from the second position toward an equilibrium point of the particular cluster, wherein relaxing the location of the centroid k from the second position toward the equilibrium point of the particular cluster occurs according to an elasticity pull factor.