US 12,367,427 B2
Methods, systems, and media for robust classification using active learning and domain knowledge
Gopi Krishnan Rajbahadur, Kingston (CA); Haoxiang Zhang, Kingston (CA); and Jack Zhenming Jiang, Richmond Hill (CA)
Assigned to HUAWEI TECHNOLOGIES CO., LTD., Shenzhen (CN)
Filed by Gopi Krishnan Rajbahadur, Kingston (CA); Haoxiang Zhang, Kingston (CA); and Jack Zhenming Jiang, Richmond Hill (CA)
Filed on Sep. 3, 2021, as Appl. No. 17/466,845.
Prior Publication US 2023/0071240 A1, Mar. 9, 2023
Int. Cl. G06N 20/10 (2019.01); G06F 18/21 (2023.01); G06F 18/2431 (2023.01)
CPC G06N 20/10 (2019.01) [G06F 18/2178 (2023.01); G06F 18/2431 (2023.01)] 18 Claims
OG exemplary drawing
 
1. A method for automatically generating, using a computing system, global feature data for use in performing a classification task, comprising:
obtaining a first set of global features generated based on decisions made by one or more human users representing expert knowledge by:
training an active learning-based classification model using active learning based on the expert knowledge with respect to an active learning training dataset, thereby obtaining a trained active learning-based classification model; and
processing the trained active learning-based classification model to generate the first set of global features;
obtaining a second set of global features generated based on one or more trained classification models, each trained classification model being trained using a respective training dataset; and
combining the first set of global features and the second set of global features to generate the global feature data.