Small sample learning method and system based on absolute-relative learning architecture

• Main Authors: LI DONGYANG, ZHANG HONGGUANG, MA LINRU, BAO JINZHEN, YANG XIONGJUN
• Format: Patent
• Language: Chinese; English
• Published: 13.08.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention provides a small sample learning method and system based on an absolute-relative learning architecture, and the method comprises the steps: calling a representation extraction module, and carrying out representation extraction of each image sample in a training set, so as to obtain a feature vector of each image sample; calling an absolute learning module to train the feature vector of each image sample to determine a category-based first prediction result and a semantic-based second prediction result of each image sample; combining the feature vectors of every two image samples into a group of sample feature pairs, and splicing the two feature vectors in each group of sample feature pairs into a group to form a vector; calling a relative learning module to train the sample feature pairs so as to determine a category-based first similarity and a semantic-based second similarity of two feature vectors in each group of sample feature pairs; and calculating a loss function of the model according to