In recent years, increasing attention has been drawn to the internal mechanisms of representation models. Traditional methods are inapplicable to fully explain the feature representations, especially if the images do not fit into any category. In this case, employing an existing class or the similarity with other image is unable to provide a complete and reliable visual explanation. To handle this task, we propose a novel visual explanation paradigm called Fea-ture Activation Mapping (FAM) in this paper. Under this paradigm, Grad-FAM and Score-FAM are designed for vi-sualizing feature representations. Unlike the previous approaches, FAM locates the regions of images that contribute most to the feature vector itself. Extensive experiments and evaluations, both subjective and objective, showed that Score-FAM provided most promising interpretable vi-sual explanations for feature representations in Person Re-Identification. Furthermore, FAM also can be employed to analyze other vision tasks, such as self-supervised represen-tation learning.

Face clustering has been widely studied, due to its broad applications in academia and industry. Regarding researches on clustering, two vital parts are involved: one is the feature representation space of the image, and the other is the clustering algorithm. However, most of the current researches lay more emphasis on the latter and overlook the need for an appropriate feature representation space for clustering. Therefore, we propose a novel clustering framework, to address the problem of insufficiently compact features in clustering. Our method is comprised of a feature topology learning module (GCN-FT) and an auto-search clustering module (GCN-AS), which called GCN-F&A (GCN-FT & GCN-AS). Specifically, GCN-FT adds a self-adaptive learning structure to the traditional GCN to capture the internal correlation of features, so that the module can better aggregate information from itself and its neighbor nodes, and provide more gathered features for clustering. Our GCN-AS consists of two parts, one is ‘1-NN’, and the other is clustering module based on linkage prediction, namely ‘GCN-LP’. 1-NN is an effective pre-clustering method, which can automatically search hyper-parameters required in GCN-LP, thereby improving the scalability of the clustering module. Experiments on two large-scale face benchmarks and one clothing dataset demonstrate that our method significantly outperforms the state-of-the-arts.

Given a binary sequence with length$n$, determining its minimal rational fraction representation (MRFR) has important applications in the design and cryptanalysis of stream ciphers. There are many studies of this problem since Klapper and Goresky first introduced an adaptive rational approximation algorithm with a time complexity of$O(n^{2}\log n\log \log n)$. In this paper, we revisit this problem by considering both adaptive and non-adaptive efficient algorithms. Compared with the state-of-art methods, we make several contributions to the problem of finding MRFR. Firstly, we find a general and precise recursive relationship between the minimal bases for two adjacent lattices generated by successive truncation sequences. This enables us to improve the currently fastest adaptive algorithm proposed by Liet al.. Secondly, by optimizing a time-consuming step of the well-known Lagrange reduction algorithm for 2-dimensional lattices, we obtain a non-adaptive, and yet practically faster MRFR-solving algorithm namedglobalEuclidean algorithm. Thirdly, we identify theoretical flaws on some non-adaptive methods in the literature by counter-examples and correct the problems by designing modified Euclidean algorithm namedpartialEuclidean algorithm. Meanwhile, we further reduce the time complexity of existing algorithm from$O(n^{2})$to$O(n\log ^{2}n\log \log n)$by invoking the half-gcd algorithm. We also conduct a comprehensive experimental comparative analysis on the above algorithms to validate our theoretical analysis.

A novel method for image segmentation is proposed. Considering the segmentation results obtained by using the multiphase Chan-Vese model are dependent on initial conditions, an edge link method is used to obtain initial curve firstly. Based on the Four Color Theorem, we can assume that in general, at most two level set functions are sufficient to detect and represent distinct intensities. Then we use an improved four phase Chan-Vese model to segment the image. Experimental results of multi-objects image demonstrate the efficiency and accuracy of the algorithm in its segmentation operations.