Cross Modal Adaptive Few-Shot Learning Based on Task Dependence

Few-shot learning (FSL) is a new machine method that applies the prior knowledge from some different domains tasks. The existing FSL models of metric-based have drawbacks, such as extracted features cannot reflect true data distribution and generalization ability weak. In order to solve problem in present, we developed model named cross modal adaptive few-shot based on task dependence (COOPERATE for short). A feature extraction representation condition network auxiliary co-training proposed. Semantic added each by combining both visual textual features. measurement scale adjusted change property parameter update algorithm. experimental results show COOPERATE has better performance comparing with all approaches monomode alignment FSL.