Search

Existing state-of-the-art unsupervised person re-identification (Re-ID) methods rely on clustering to generate pseudo labels for training. The reliability of pseudo labels directly affects the performance of these methods. Due to unsatisfactory feature embedding and imperfect clustering, pseudo labels are not always reliable that significantly hinder the representation learning of person. To address this issue, an Uncertainty-guided Robust labels refinement method is proposed. It employs uncertainty as guidance to selects and optimizes unreliable pseudo labels, which alleviates the impact of incorrect pseudo labels. Specially, an uncertainty estimation module is constructed, which identifies the samples with low reliability applies base on consistency between the ideal distribution and the predicted distribution. To mitigate the impact of unreliable samples, the labels refinement module is designed, which rebuilds labels for low reliability samples by measuring similarity from the closest centroid. Thanks to the reliability of pseudo labels provided by uncertain estimation module, the proposed method enhances robustness to noisy labels and learns discriminative representations of person. Extensive experiments demonstrate that the proposed method is effective and achieves advanced performance for unsupervised person Re-ID. [ABSTRACT FROM AUTHOR]

The traditional data envelopment analysis (DEA) model is widely used for efficiency evaluation of decision making unit composed of past or current input–output data in production system. We propose a new efficiency prediction model which for the first time combines information granulation (IG) and support vector machine (SVM) with DEA model, to evaluate the future efficiency of decision making unit in this paper. The model first uses fuzzy information granulation to granulate the input–output data over time series and then describe the characteristics of the data within each time window by the minimum, average and maximum values and establish the IG–SVM model based on the fuzzy information granulation and support vector machine. After training by time series data, the optimal model of regression is built. Based on this model, the minimum, average and maximum values of the next window in the future are predicted, and finally the future efficiency of the decision making unit can be calculated by the DEA model. Case studies show the feasibility and applicability of the proposed method. [ABSTRACT FROM AUTHOR]

Linguistic hesitant intuitionistic fuzzy set, which allows an element having several linguistic evaluation values and each linguistic argument having several intuitionistic fuzzy memberships, is a power tool to model uncertain information existing in multiple attribute decision-making problems. In this paper, we propose new methods by using TOPSIS and VIKOR for multiple attribute decision-making problems, in which evaluation values are in the form of linguistic hesitant intuitionistic fuzzy elements. Different situations of attribute weight information are considered. If attribute weights are partly known, a linear programming model is set up based on the idea that reasonable weights should make the relative closeness of each alternative evaluation value to the linguistic hesitant intuitionistic fuzzy positive ideal solution as large as possible. If attribute weights are unknown completely, an optimization model is set up based on the maximum deviation method. A numerical example is presented to illustrate feasibility and practical advantages of the proposed method. We compare the alternatives’ rankings derived from the linguistic hesitant intuitionistic fuzzy TOPSIS method with those derived from the hesitant fuzzy linguistic TOPSIS and the hesitant intuitionistic fuzzy TOPSIS approach to further illustrate their advantages. [ABSTRACT FROM AUTHOR]