In order to reduce the cost of annotation and solve the problems of inaccurate target localization and omission of detail information, a semi-weakly supervised object detection method with bi-attention-guided feature fusion is proposed. Based on the method which fully labelled and weakly labelled data, the detection performance and annotation cost are balanced, and the spatial attention the low-level feature maps with the high-level feature maps with pixel-level weighting are fused, so that the high-level feature maps have rich low-level information, and performs channel-weighting operations on the fused feature maps to obtain high-level feature maps having rich details and location information. In order to get more accurate pseudo-labelled boxes, a more robust candidate box selection strategy is proposed. The proposed algorithm has better detection performance and reduce the amount of full-labeled image data and additional image-level labeling.

Aiming at the problems of intra-class imbalance and inter-class imbalance and high sparsity of microbial data, a data augmentation method that fuses matrix factorization and space partition is proposed. Matrix factorization technology is used to decompose the original data space into object subspace and feature subspace to extract the latent space representation. The object subspace is divided into multiple data subspaces to alleviate the intra-class imbalance problem. Synthetic samples are then generated in each data subspace to address the inter-class imbalance. Synthetic samples are filtered using Euclidean distance to obtain high-quality samples. The experiment is conducted on 9 microbial data sets, and the performance is compared with 9 sampling algorithms. The results show that the samples generated by the proposed method have great advantages in diversity, and more positive samples can be identified under multiple classifiers.

We combine the natural neighbor search algorithm to improve a series of problems of the density peaks clustering(CFSFDP)algorithm, and propose the NaN-CFSFDP algorithm. First, an outlier samples detection method is proposed based on the natural neighbor search algorithm. Then, for the problem that the truncation distance dc is difficult to be taken accurately manually in the CFSFDP algorithm, the calculation of d_c is improved in combination with the natural neighbor search algorithm, and the automatic taking of dc is realized. The metric rule of the sample density of the CFSFDP algorithm is redesigned and unified to make it pay more attention to the local information of each sample. Finally, to address the problem that the density peak points in the dataset may be concentrated in dense clusters due to the large density difference between clusters, which leads to cluster loss, the concepts of shared natural neighbors for samples and shared natural neighbors for clusters are proposed to construct a new cluster fusion algorithm. Experimental results on synthetic and real datasets show that the algorithm outperforms or is at least comparable to the comparative method in terms of clustering performance in most cases and has fewer parameters compared to CFSFDP algorithm and its improvements.

Aiming at the problem of the complete subtraction and division problems under picture fuzzy environments, the subtraction and division operations of PFNs(picture fuzzy numbers, PFNs)are defined by constructing equations for addition and multiplication. Based on the nearest distances, the optimal solution of the equation is derived in various situations, and more complete subtraction and division operation algorithms are obtained. The properties of the two newly defined operators and the mixed operation of addition and multiplication are analyzed in detail. The two improved operators are extended in a pointwise manner to the picture fuzzy set, and the essential properties are discussed. The applicability and feasibility of this operator is verified through numerical analysis.

For issues where the absence of essential spatial structure signals makes it highly challenging to estimate 3D bounding boxes accurately from a single picture, a monocular 3D object detection algorithm is proposed based on a multi-scale channel attention mechanism plus depth guidance to conquer these challenges. To introduce 3D data and effectively capture spatial information from different scales of feature maps, the depth maps and monocular image feature maps are pre-processed in the feature extraction module using a pyramid split algorithm, respectively, and then on the basic of the weight using the channel-wise attention module to calibrate the corresponding feature vectors to generate a refined feature map which is richer in multi-scale feature information. A depth-guided dynamic local convolution network is suggested for applying depth maps as specific kernels that contain spatial structure signals to monocular image feature maps. This method mitigates error accumulation from direct fusion and addresses the scale sensitivity issue of objects looking larger or smaller with distance. The models performance is assessed and also compared using various evaluation metrics. Experimental results demonstrate that the method proposed in this paper improves the 3D detection accuracy for cars,pedestrians and cyclists in the autonomous driving datasets when compared to other algorithms.

A fuzzy C-means(fuzzy C-means, FCM)clustering algorithm based on five-region shadowed sets is proposed in this paper. The membership degree of the object to the cluster is obtained by the FCM algorithm. The object is divided into core region, semi-core region, shadow region, semi-negative region and negative region according to the membership degree by introducing the five-region shadowed sets. Then, a threshold value ω is obtained by analyzing the semi-core region. The objects whose membership degree μ≥ω in the core region and semi-core region are classified into this cluster to get the final clustering result. Experiments are carried out on 8 public data sets with other 3 clustering algorithms, compared with the other 3 algorithms, the algorithm proposed in this paper achieves the best clustering results on 7 data sets. The experimental results show that the proposed algorithm in this paper is superior to 3 other algorithms.

Rule acquisiton in strongly consistent formal decision context is studied by using the idea of concept reduction of keeping the binary relation unchanged in the formal context. According to the concept reduction of decision subcontext, the conditional attribute set induced by concept reduction is defined, and it is proved that it is the attribute consistent set of the strongly consistent formal decision context, so as to simplify the conditional attribute set. The relationship between concept reduction and other concepts is given, and then it is proved that the conditional relation set induced by concept reduction contains all the information of the simplified rule antecedent. The concrete steps of obtaining non-redundant rules for the strongly consistent formal decision context by concept reduction are given.

Based on the fuzzy rough approximate operator, the concept of upper and lower approximate fuzzy knowledge states is proposed. A notation of upper and lower approximate fuzzy knowledge states is discussed. A necessary and sufficient conditions for upper and lower approximate fuzzy knowledge states family to form the fuzzy knowledge structure is obtained respectively. The lower approximate fuzzy knowledge state set family is a necessary and sufficient condition for the fuzzy closure space, and the upper approximate fuzzy knowledge state set family is a necessary and sufficient condition for the fuzzy knowledge space. It is also proven that the fuzzy closure space induced by the same pair of fuzzy rough approximation operators is dual to the fuzzy knowledge space, and the fine relationship of the upper and lower approximate fuzzy knowledge structures is explored.

Generalized interval-valued q-rung orthopair hesitant fuzzy soft sets(GIVq-ROHFSSs)are proposed, and their operations of union, intersection, complement, and, or are defined to gain relevant properties. Regarding GIVq-ROHFSSs, average membership and non-membership degrees associated with interval values are extracted, and correlations and correlation coefficients are defined to acquire an extended principle and several basic properties. Correlation coefficients of GIVq-ROHFSSs are adopted, and optimal sorting is performed to motivate multi-attribute decision making. The practical example of medical resources and the energy project investment case show the effectiveness of new decision method. The related modeling, measurement and decision facilitate uncertainty analysis and applications.

Theory of intuitive fuzzy soft expert set(IFSES)is expanded, the concept of intuitive hesitation fuzzy soft expert set is improved. The basic operation of the fuzzy soft expert set is defined and a series of important properties is studied. The intuitive hesitation fuzzy soft expert set decision algorithm is given. In the algorithm, evaluation information to the alternative is considered based on the decision maker and weight to the parameter set is given. And the intuitive hesitation fuzzy soft expert set decision algorithm is applied to practical recruitment problems. The composite score of each candidate is calculated by the score function. The best candidate is obtained from large to small. Compared with other soft set decision models, soft expert set decision model is simple and accurate. Through comparative analysis, intuitive hesitation fuzzy soft expert set decision model has better combination, decision-making and application.

To handle q-rung orthopair fuzzy multi-attribute decision making problems, an information fusion method is proposed based on the q-rung orthopair fuzzy self-dual aggregation operator, which is induced by the weighted power mean operator with its power being rung q. Some regular properties of this q-rung orthopair fuzzy aggregation operator are investigated, such as the idempotency, monotonicity and boundedness. The limiting case of this operator is examined as q approaches infinity, and the boundedness is precisely characterized by the monotonicity of weighted power means. The aggregation operator based approach is developed to deal with multi-attribute decision making problems under q-rung orthopair fuzzy environment. An illustrative example related to the venue selection for sporting events is proposed to show the effectiveness and feasibility of this approach. The influence of the parameter therein on the ranking results is discussed to demonstrate the robustness, and comparisons with some existing methods are presented, which implies the current method can maintain the final results with a simpler calculation.