US 11,816,593 B2
TAFSSL: task adaptive feature sub-space learning for few-shot learning
Leonid Karlinsky, Mazkeret Batya (IL); Joseph Shtok, Binyamina (IL); and Eliyahu Schwartz, Haifa (IL)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Aug. 23, 2020, as Appl. No. 17/000,319.
Prior Publication US 2022/0058505 A1, Feb. 24, 2022
Int. Cl. G06N 20/00 (2019.01); G06N 7/01 (2023.01); G06F 16/55 (2019.01); G06V 10/762 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01)
CPC G06N 7/01 (2023.01) [G06F 16/55 (2019.01); G06N 20/00 (2019.01); G06V 10/762 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01)] 17 Claims
OG exemplary drawing
 
1. A method, implemented in a computer system comprising a processor, memory accessible by the processor, and computer program instructions stored in the memory and executable by the processor, the method comprising:
training a machine learning system to classify features in images by:
generating a sample set comprising one or a few labeled training samples and one or a few bulk query samples that are used as unlabeled samples or one or a few unlabeled samples for which predictions are not needed, and wherein the generated sample set is Gaussian,
performing dimensionality reduction computed on the samples in the sample set to form a dimension reduced sub-space, wherein the dimensionality reduction is performed by finding dimensions of maximal variance,
generating class representatives in the dimension reduced sub-space using clustering, wherein the clustering is performed using Mean-Shift Propagation processing; and
classifying features in images using the trained machine learning system.