Few-shot learning in image classification is developed to learn a model that aims to identify unseen classes with only few training samples for each class. Fewer training samples and new tasks of classification make many traditional classification models no longer applicable. In this paper, a novel few-shot learning method named multi-scale metric learning (MSML) is proposed to extract multi-scale features and learn the multiscale relations between samples for the classification of few-shot learning. In the proposed method, a feature pyramid structure is introduced for multi-scale feature embedding, which aims to combine high-level strong semantic features with low-level but abundant visual features. Then a multi-scale relation generation network (MRGN) is developed for hierarchical metric learning, in which high-level features are corresponding to deeper metric learning while low-level features are corresponding to lighter metric learning. Moreover, a novel loss function named intra-class and inter-class relation loss (IIRL) is proposed to optimize the proposed deep network, which aims to strengthen the correlation between homogeneous groups of samples and weaken the correlation between heterogeneous groups of samples. Experimental results on miniImageNet and tieredImageNet demonstrate that the proposed method achieves superior performance in few-shot learning problem.

Z-number is an effective model to describe uncertainty in the real world. Under the condition that uncertainty reasoning is an important issue to process information, how to achieve Z-valuation uncertainty reasoning is a problem. As Z-number involves both fuzzy and probabilistic uncertainty, main difficulty in the problem to be solved is accomplishing both two uncertainties' reasoning. In this paper, a novel Z-network model and its associated reasoning algorithm are proposed to overcome the difficulty. Structure of the proposed Z-network which contains three basic structures is DAG, and this is similarly with Bayesian network. Process of reasoning algorithm involves two parts: First, Bayesian reasoning is applied to establish an optimization model for probabilistic uncertainty reasoning in Z-number; Second, the arithmetic approach of discrete Z-number on if-then rule and maximum entropy approach are proposed for fuzzy uncertainty reasoning. Z-network is essentially an extended model on the basis of Bayesian network and properties of Z-number for Z-valuation uncertainty reasoning. In application, a novel framework of dependence assessment in human reliability analysis is proposed based on Z-network, and case study demonstrates its effectiveness.