US 12,141,715 B2
Machine learning knowledge management based on lifelong boosting in presence of less data
Anil Goyal, Dossenheim (DE); Ammar Shaker, Heidelberg (DE); and Francesco Alesiani, Heidelberg (DE)
Assigned to NEC CORPORATION, Tokyo (JP)
Filed by NEC Corporation, Tokyo (JP)
Filed on Dec. 6, 2023, as Appl. No. 18/530,871.
Application 18/530,871 is a continuation of application No. 16/889,825, filed on Jun. 2, 2020, granted, now 11,995,566.
Prior Publication US 2024/0127088 A1, Apr. 18, 2024
Int. Cl. G06N 5/04 (2023.01); G06N 20/00 (2019.01)
CPC G06N 5/04 (2013.01) [G06N 20/00 (2019.01)] 14 Claims
OG exemplary drawing
 
1. A method for lifelong machine learning using boosting, the method comprising:
learning, by using previously learned knowledge from old tasks, weights over a learning sample for a new task comprising examples using previously learned classifiers;
receiving the learning sample and weights over the learning sample using the previously learned classifiers;
sequentially learning a set of base classifiers and weights over the base classifiers for the new task;
maintaining a distribution of weights over the learning sample in order to learn task-specific classifiers;
pruning some of newly learned task-specific classifiers from the new task based on performance of the task-specific classifiers on the learning sample;
storing hard examples which are hard to classify based on the performance of the task-specific classifiers on the learning sample for both new and old task-specific classifiers; and
updating a knowledge base (KB) such that the weights over the newly learned classifiers for the old tasks are learned using the hard examples from the old tasks.