Unsupervised domain adaptation aims to transfer knowledge from a labeled source domain to an unlabeled target domain. Existing methods, whether based on distribution matching or self-supervised learning, often focus solely on classifying individual source samples, potentially overlooking discriminative information. To address this limitation, we propose FlowUDA, a novel plugin method that enhances existing UDA frameworks by constructing semantically invariant flows from individual source samples to corresponding target samples, forming cross-domain trajectories. By leveraging a diffusion network guided by ordinary differential equations, FlowUDA ensures these flows preserve the topological structure of the source domain, maintaining their distinguishability. Our method then classifies these flows by sampling points along them and transferring labels from source samples, effectively capturing spatial relationships between domains. In essence, FlowUDA transforms the traditional point-based classification on individual source samples into flow-based classification on flows, allowing the model to learn richer, more discriminative features that bridge the gap between source and target domains. Extensive experiments on standard benchmarks demonstrate that integrating FlowUDA into existing UDA methods leads to notable performance gains, highlighting its effectiveness in addressing domain shift challenges.

Spatial co-location pattern mining aims to discover sets of spatial features that prevalently occur together. However, existing methods for local co-location pattern (LCP) mining often yield prevalent regions with limited spatial coverage, resulting in the overestimation of pattern interest measures and the omission of implicit spatial associations. Furthermore, the distinction between pattern categories is often obscured, as global co-location patterns can be regarded as special cases of LCPs. To address these limitations, we propose a novel framework that can identify prevalent regions with broader coverage and achieve clear pat tern classification. We first prove that the problem of maximizing each prevalent region is NP-hard. Then we formally define the concept of maximal prevalent regions as a viable alternative and develop a heuristic dynamic spatial expansion algorithm for their efficient identification. In addition, we introduce a spatial occu pancy ratio and a three-level classification scheme (rare, common, and significant) to replace the traditional global/local dichotomy. Finally, a Delaunay-triangulation-based method is employed to quantify the coverage of non-convex regions, ensuring accurate occupancy calculations. Extensive experiments on synthetic and real-world datasets demonstrate the effectiveness and efficiency of the proposed framework.

Multi-Object Tracking (MOT) in Unmanned Aerial Vehicles (UAV) aims to continuously and stably detect and track objects in videos captured by UAVs. In existing MOT tracking-by-detection schemes, the tracker with a fixed step size is always employed, and a fixed length of past tracking information is input to the tracker to guide position prediction. However, the limited prediction range of a single-scale tracker leads to frequent tracking losses, and limited historical information also reduces tracking accuracy. To address these limitations, we propose a novel Long-Short Match (LSMTrack) tracking method. The key idea is to use long and short trackers and maintain a long-term motion state to improve tracking performance, thus reducing the likelihood of entering the lost status. To this end, a new Mamba-based tracker and a long-short match strategy are proposed. For long and short trackers, the same architecture is used based on Mamba. Unlike the previous Mamba-based approach, the proposed tracker maintains a long-term state while updating the state and making position predictions in each time step, so we call it a step Mamba tracker. Meanwhile, we devise a long-short match strategy at the inference stage to integrate long and short trackers, and design a lost control operation which updates the long-term states using historical state values. In this way, the matching probability and the inference efficiency are guaranteed. Experimental results on two UAV MOT datasets confirm the state-of-the-art performance. Specifically, the best results are achieved in terms of two popular MOTA and IDF1 tracking evaluation metrics.

Unmanned Aerial Vehicle Object Detection (UAVOD) presents unique challenges due to varying altitudes, dynamic backgrounds, and the small size of objects. Traditional detection methods often struggle with these challenges, as they typically rely on visual feature only and fail to extract the semantic relations between the objects. To address these limitations, we propose a novel approach named Self-Prompting Analogical Reasoning (SPAR). Our method utilizes the vision-language model (CLIP) to generate context-aware prompts based on image feature, providing rich semantic information that guides analogical reasoning. SPAR includes two main modules: self-prompting and analogical reasoning. Self-prompting module based on learnable description and CLIP-text encoder generates context-aware prompt by combining specific image feature; then an objectness prompt score map is produced by computing the similarity between pixel-level features and context-aware prompt. With this score map, multi-scale image features are enhanced and pixel-level features are chosen for graph construction. While for analogical reasoning module, graph nodes consists of category-level prompt nodes and pixel-level image feature nodes. Analogical inference is based graph convolution. Under the guidance of category-level nodes, different-scale object features have been enhanced, which helps achieve more accurate detection of challenging objects. Extensive experiments illustrate that SPAR outperforms traditional methods, offering a more robust and accurate solution for UAVOD.

Test-time adaptation with pre-trained vision-language models, such as CLIP, aims to adapt the model to new, potentially out-of-distribution test data. Existing methods calculate the similarity between visual embedding and learnable class embeddings, which are initialized by text embeddings, for zero-shot image classification. In this work, we first analyze this process based on Bayes theorem, and observe that the core factors influencing the final prediction are the likelihood and the prior. However, existing methods essentially focus on adapting class embeddings to adapt likelihood, but they often ignore the importance of prior. To address this gap, we propose a novel approach, Bayesian Class Adaptation (BCA), which in addition to continuously updating class embeddings to adapt likelihood, also uses the posterior of incoming samples to continuously update the prior for each class embedding. This dual updating mechanism allows the model to better adapt to distribution shifts and achieve higher prediction accuracy. Our method not only surpasses existing approaches in terms of performance metrics but also maintains superior inference rates and memory usage, making it highly efficient and practical for real-world applications.

Unmanned Aerial Vehicle (UAV) object detection faces significant challenges due to complex environmental conditions and different imaging conditions. These factors introduce significant changes in scale and appearance, particularly for small objects that occupy limited pixels and exhibit limited information, complicating detection tasks. To address these challenges, we propose a Multimodel Causal Reasoning framework based on YOLO backbone for UAV Object Detection (MCR-UOD). The key idea is to use the backdoor adjustment to discover the condition-invariant object representation for easy detection. Specifically, the YOLO backbone is first adjusted to incorporate the pre-trained vision-language model. The original category labels are replaced with semantic text prompts, and the detection head is replaced with text-image contrastive learning. Based on this backbone, our method consists of two parts. The first part, named language guided region exploration, discovers the regions with high probability of object existence using text embeddings based on vision-language model such as CLIP. Another part is the backdoor adjustment casual reasoning module, which constructs a confounder dictionary tailored to different imaging conditions to capture global image semantics and derives a prior probability distribution of shooting conditions. During causal inference, we use the confounder dictionary and the prior to intervene on local instance features, disentangling condition variations, and obtaining condition-invariant representations. Experimental results on several public datasets confirm the state-of-the-art performance of our approach. The code, data and models will be released upon publication of this paper.

A spatial co-location pattern is a subset of a spatial feature set whose instances prevalently appear in nearby locations in space. The objective of spatial co-location pattern mining is to detect co-location patterns that are non-obvious, informative, or predictive. Due to the heterogeneity in the distribution of spatial instances, spatial co-location patterns are classified into global co-location patterns (GCPs) and local co-location patterns (LCPs). The technique that discovers both simultaneously is termed multi-level co-location pattern mining (MLCPM). However, existing MLCPM methods have room for improvement in efficiently identifying GCPs and perform poorly in discovering prevalent sub-areas of LCPs. To address these issues, we propose a novel MLCPM framework called ML-CCHDB. This framework enhances GCP mining efficiency by optimizing the column calculation method tailored for MLCPM. Furthermore, it utilizes the HDBSCAN clustering method to identify potential prevalent sub-areas of LCPs and develops an adaptive approach for generating input parameters to enhance detection efficiency and quality. Experimental results on both synthetic and real datasets demonstrate that the column calculation optimizations in ML-CCHDB effectively enhance efficiency. Moreover, HDBSCAN strikes a balance between efficiency and quality in prevalent sub-area mining. These results fully validate the proposed framework's effectiveness and efficiency.

Spatial multi-level co-location pattern mining (MLCPM) is the improvement and further development of traditional spatial co-location pattern mining, which can simultaneously mine both global and local prevalent co-location patterns. Existing MLCPM algorithms are quite time-consuming and are very sensitive to prevalence thresholds. Especially when the prevalence threshold is changed, the entire mining process needs to be redone. Considering the above problems, this paper proposes a novel MLCPM algorithm based on extended maximal cliques (MLCPM-EMC). Firstly, a new materialization model of neighbor relationships between instances, the extended maximal clique (EMC), is proposed to reduce the number of the maximal cliques. Secondly, a novel hash structure is designed to store the EMCs. With the hash structure, the algorithm can recognize global prevalent colocations quickly. Moreover, a DBSCAN clustering method based on participation instances of local patterns is used to efficiently identify their prevalent regions. Lastly, extensive experiments on both synthetic and real-world datasets show that the EMC materialization model reduces the number of maximal cliques by about 80%. And the MLCPM-EMC algorithm has better performance and scalability than two state-of-the-art baseline algorithms, in particular, the efficiency of re-mining increases several times when changing the prevalence threshold.

Due to the capability of Heterogeneous Information Network (HIN) in modeling different types of entities and relations, HIN based recommendation is proposed to enhance performance. In existing methods, meta-paths are extensively used to explore the semantic information in HINs. However, these methods mainly consider some nodes in instances of meta-paths, which may result in information loss and reduce recommendation performance. In the paper, we propose a Pre-training encoder based on Meta-Path for Recommendation (PMPRec) to learn rich and meaningful information between users and items. Specifically, in the pre-training stage, we design a Meta-Path based encoder (MP-encoder) to properly capture both structural and semantic information between users and items by considering all nodes and edges in instances of meta-paths including their types and positions, leading to effective embeddings. In the fine-tuning stage, we utilize a Multi-Layer Perceptron (MLP) to further optimize the embeddings according to user-item ratings. Experiments on four real-world datasets show the proposed PMPRec model outperforms the baseline models.

Generative self-supervised learning (SSL) for heterogeneous graph representation has drawn increasing attention due to its powerful capability of capturing rich structural and semantic information without relying on labeled information. However, the multi-layer network structure with multiple attention mechanisms and masking strategies adopted by existing generative SSL methods based on Auto-Encoder, brings semantic confusion and creates dependence of the reconstruction of masked attributes on attribute distinguishability issues. To address these issues, we propose Multi-masks and Bi-spaces reconstruction based Single-Layer Heterogeneous graph AutoEncoder(MBS-HAE), which employs a single-layer masked autoencoder with only semantic attention to learn node embeddings and uses re-masking by randomly selecting nodes to reduce the model’s reliance on the distinguishability of input attributes. Meanwhile, MBS-HAE performs bi-spaces reconstruction in both attribute and latent space to improve the certainty of inferring masked data. The extensive experiments on multiple datasets demonstrate the effectiveness and efficiency of the proposed MBS-HAE.

We propose to explore an interesting and promising problem, Cloud Object Detector Adaptation (CODA), where the target domain leverages detections provided by a large cloud model to build a target detector. Despite with powerful generalization capability, the cloud model still cannot achieve error-free detection in a specific target domain. In this work, we present a novel Cloud Object detector adaptation method by Integrating different source kNowledge (COIN). The key idea is to incorporate a public vision-language model (CLIP) to distill positive knowledge while refining negative knowledge for adaptation by self-promotion gradient direction alignment. To that end, knowledge dissemination, separation, and distillation are carried out successively. Knowledge dissemination combines knowledge from cloud detector and CLIP model to initialize a target detector and a CLIP detector in target domain. By matching CLIP detector with the cloud detector, knowledge separation categorizes detections into three parts: consistent, inconsistent and private detections such that divide-and-conquer strategy can be used for knowledge distillation. Consistent and private detections are directly used to train target detector; while inconsistent detections are fused based on a consistent knowledge generation network, which is trained by aligning the gradient direction of inconsistent detections to that of consistent detections, because it provides a direction toward an optimal target detector. Experiment results demonstrate that the proposed COIN method achieves the state-of-the-art performance.

Abstract Discovering the correlation relationships of spatial facilities in urban functional regions is of great significance in analyzing urban planning and promoting urban development. However, existing regional co-location pattern mining methods do not fully consider the attributes of urban functional regions. Consequently, they fail to effectively capture the relationships between spatial facilities and regions, resulting in patterns lacking practical application value and interpretability. To address this issue, this paper proposes a novel regional co-location pattern (RCP) mining method based on urban functional region division. First, an ontology-based method for urban functional region division is proposed. On the basis of dividing urban functional regions, the expected mean distance is generated to adaptively calculate the neighbor relationships between instances in each region, and a parallel algorithm is designed to quickly mine RCPs in urban functional regions. Then, the correlation coefficient ( $$CC_{RP}$$ C C RP ) measures the association strength between RCP and function type to provide interpretability for RCPs. Finally, extensive experiments on real-world datasets are conducted to verify the effectiveness of urban functional region partition and to demonstrate the superiority of the proposed RCP mining method.

Attributed network embedding (ANE) maps nodes in a network into a low-dimensional space while preserving the intrinsic essence of node attribute and network topology. Incorporating node attribute and network topology with more deeply and more harmoniously is a critical and challenging issue in the ANE, because node attribute and network topology are two kinds of heterogeneous information. Existing approaches fuse two kinds of heterogeneous information at different stages: i.e. before, during or after the learning process. In fact, fusions at different stages have their own advantages and disadvantages. To maximise the profit of utilising the attributed and networked information in ANE, we propose an Attributed Network Embedding model with Dual Fusion strategies (abbr. ANEDF), which consists of both mutually beneficial components: early fusion component for capturing the latent complementarity and late fusion component for extracting the unique and distinctive information from node attribute and network topology. The two components are co-trained during the learning process, which promotes information interaction and captures the consensus of heterogeneous information. Extensive experiments with the tasks of node classification, node clustering, link prediction and visualisation on eight publicly available networks have been conducted to evaluate the effectiveness and rationality of the proposed model. The experimental results demonstrate that ANEDF obtains the best classification, clustering and link prediction performance on 6–7 of 8 datasets, respectively.

The spatial co-location pattern is a collection of spatial features in which instances of features prevalently appear in neighboring spatial regions. Due to the heterogeneity of spatial data distribution, the instances of some patterns appear prevalently in the global region (i.e., Global Prevalent Co-location Patterns, GPCPs), while some patterns are not prevalent globally, and their instances are clustered only in some local regions (i.e., Local Prevalent Co-location Patterns, LPCPs). Multi-level co-location pattern mining aims to mine these two types of patterns simultaneously, but existing methods cannot accurately judge the spatial distribution of patterns in a certain region, leading to unsuitable judgment of both GPCPs and LPCPs. To overcome this problem, this paper firstly proposes the relative distribution coefficient to identify the spatial distribution form of patterns, and provides a more refined way for discovering both GPCPs and LPCPs. Secondly, a novel multi-level co-location pattern mining algorithm is proposed by using the relative distribution coefficient as the interest metrics, and some pruning strategies are suggested to improve the mining efficiency. Finally, extensive experiments are conducted on both real and synthetic datasets to verify the effectiveness and efficiency of the proposed method.

Community search (CS) is an important research topic in network analysis, which aims to find a subgraph that satisfies the given conditions. A dynamic attribute heterogeneous information network (DAHIN) is a sequence of attribute heterogeneous information network (AHIN) snapshots, where each snapshot consists of multiple types of vertices as well as edges, and each vertex is associated a set of attribute keywords. CS over DAHIN faces many challenges. In this paper, we study the CS problem over DAHINs, aiming to search for cohesive subgraphs containing query vertex and simultaneously satisfying the connectivity, attribute cohesiveness and interaction stability. To this end, we propose a deep learning model with a three-level attention mechanism and the concept of interaction frequency with respect to multiple semantic relationships to measure the similarity of attributes and the stability of interactions between vertices respectively. In addition, we design three search algorithms to locate the target community by optimizing the degree of interaction stability and attribute similarity between vertices. Extensive experiments, including comparison with existing algorithms, ablation analysis, parameter sensitivity examination, and case studies, are conducted on four real-world datasets to validate the effectiveness and efficiency of the proposed model and search algorithms. The code and model of CS-DAHIN will be open source on GitHub.

A co-location pattern indicates a subset of spatial features whose instances are frequently located together in proximate geographical space. Most previous studies of spatial co-location pattern mining concern what percentage of instances per feature are involved in the table instance of a pattern, but neglect the heterogeneity in the number of feature instances and the distribution of instances. As a result, the deviation may be occurred in the interest measure of co-locations. In this article, we propose a novel mixed prevalence index (MPI) incorporating the effect of feature-level and instance-level heterogeneity on the prevalence measure, which can address some dilemmas in existing interest measures. Luckily, MPI possesses the partial antimonotone property. In virtue of this property, a branch-based search algorithm equipped with some optimizing strategies of MPI calculation is proposed, namely, Branch-Opt-MPI. Comprehensive experiments are conducted on both real and synthetic spatial datasets. Experimental results reveal the superiority of MPI compared to other interest measures and also validate the efficiency and scalability of the Branch-Opt-MPI. Particularly, the Branch-Opt-MPI performs more efficiently than baselines for several times or even orders of magnitude in dense data.

Existing unsupervised domain adaptation (UDA) methods rely on aligning the features from the source and target domains explicitly or implicitly in a common space (i.e., the domain invariant space). Explicit distribution matching ignores the discriminability of learned features, while the implicit counterpart such as self-supervised learning suffers from pseudo-label noises. With distribution alignment, it is challenging to acquire a common space which maintains fully the discriminative structure of both domains. In this work, we propose a novel HomeomorphisM Alignment (HMA) approach characterized by aligning the source and target data in two separate spaces. Specifically, an invertible neural network based homeomorphism is constructed. Distribution matching is then used as a sewing up tool for connecting this homeomorphism mapping between the source and target feature spaces. Theoretically, we show that this mapping can preserve the data topological structure (e.g., the cluster/group structure). This property allows for more discriminative model adaptation by leveraging both the original and transformed features of source data in a supervised manner, and those of target domain in an unsupervised manner (e.g., prediction consistency). Extensive experiments demonstrate that our method can achieve the state-of-the-art results. Code is released at https://github.com/buerzlh/HMA.

Abstract Diagnosis prediction exploits electronic health records (EHRs) to predict the future diagnoses of patients, further supporting clinical decision making and personalized treatments. However, a patient's EHR is an irregular sequence of visits that contains a large number of medical concepts. The disease progression patterns are closely related to the visits, as well as the contextual knowledge of each visit. The existing diagnosis prediction methods ignore the complex relationships between the visits and the contextual knowledge, and thus cannot achieve satisfactory performance. Therefore, we develop a knowledge‐aware representation learning method to comprehensively model these complex relationships. Specifically, we first construct a medical knowledge graph to model the correlations between medical concepts in EHRs, and project the contextual knowledge into the pre‐learned vectors. We then devise an enhanced gated recurrent unit (GRU) neural network to extract the long‐term intra‐relationships between visits, and design a novel knowledge attention module to capture the complex inter‐relationships between the visits and the contextual knowledge. Armed with these, we provide a powerful and flexible framework to capture the long‐term discriminative disease progression patterns for diagnosis prediction. Intensive experiments are conducted on two real‐world EHR datasets. And the model achieves a competitive performance with Code‐level Accuracy@20 of 0.7465 and Visit‐level Precision@20 of 0.7547 on MIMIC‐III, and Code‐level Accuracy@20 of 0.9337 and Visit‐level Precision@20 of 0.9427 on MIMIC‐IV.

Unsupervised domain adaptation aims to transfer knowledge from labeled source domain to unlabeled target domain. The semi-supervised method based on mean-teacher framework is one of the main stream approaches. By enforcing consistency constraints, it is hopeful that the teacher network will distill useful source domain knowledge to the student network. However, in practice negative transfer often emerges because the performance of the teacher network is not guaranteed to be always better than the student network. To address this limitation, a novel Adaptive Mutual Learning (AML) strategy is proposed in this paper. Specifically, given a target sample, the network with worse prediction will be optimized by pushing its prediction close to the better prediction. This is in the spirit of traditional knowledge distillation. On the other hand, the network with better prediction is further refined by requiring its prediction to stay away from the worse prediction. This can be regarded conceptually as reverse knowledge distillation. In this way, two networks learn from each other according to their respective performance. At inference phase, the averaged output of these two networks can be taken as the final prediction. Experimental results demonstrate that our AML achieves competitive results.

Multiview subspace clustering has turned into a promising technique due to its encouraging ability to discover the underlying subspace structure. In recent studies, a lot of subspace clustering methods have been developed to strengthen the clustering performance of multiview data, but these methods rarely consider simultaneously the nonlinear structure and multilevel representation (MLR) information in multiview data as well as the data distribution of latent representation. To address these problems, we develop a new Multiview Subspace Clustering with MLRs and Adversarial Regularization (MvSC-MRAR), where multiple deep auto-encoders are utilized to model nonlinear structure information of multiview data, multiple self-expressive layers are introduced into each deep auto-encoder to extract multilevel latent representations of each view data, and diversity regularizations are designed to preserve complementary information contained in different layers and different views. Furthermore, a universal discriminator based on adversarial training is developed to enforce the output of each encoder to obey a given prior distribution, so that the affinity matrix for spectral clustering (SPC) is more realistic. Comprehensive empirical evaluation with nine real-world multiview datasets indicates that our proposed MvSC-MRAR achieves significant improvements than several state-of-the-art methods in terms of clustering accuracy (ACC) and normalized mutual information (NMI).

Source-free object detection (SFOD) needs to adapt a detector pre-trained on a labeled source domain to a tar-get domain, with only unlabeled training data from the tar-get domain. Existing SFOD methods typically adopt the pseudo labeling paradigm with model adaption alternating between predicting pseudo labels and fine-tuning the model. This approach suffers from both unsatisfactory accuracy of pseudo labels due to the presence of domain shift and lim-ited use of target domain training data. In this work, we present a novel Learning to Overlook Domain Style (LODS) method with such limitations solved in a principled man-ner. Our idea is to reduce the domain shift effect by en-forcing the model to overlook the target domain style, such that model adaptation is simplified and becomes easier to carry on. To that end, we enhance the style of each tar-get domain image and leverage the style degree difference between the original image and the enhanced image as a self-supervised signal for model adaptation. By treating the enhanced image as an auxiliary view, we exploit a student- teacher architecture for learning to overlook the style de-gree difference against the original image, also character-ized with a novel style enhancement algorithm and graph alignment constraint. Extensive experiments demonstrate that our LODS yields new state-of-the-art performance on four benchmarks.

Spatial co-location pattern mining (SCPM) aims to discover subsets of spatial features frequently located in nearby geographic space. Previous studies of SCPM only concern the inter-features association of a pattern, but neglect the interesting intra-feature behavior. In this paper, we propose spatial co-location pattern mining with coupling relation consideration (SCPM-CR) to capture complex relations in a co-location. Specifically, InterPCI is proposed to capture the inter-features coupling in a pattern, and IntraCAI is designed to capture the congregating behavior of intra-feature objects. Based on the anti-monotone property of InterPCI, a general framework which searches for patterns in a level-wise manner is suggested. To tackle the participating object search problem, a candidate-and-search algorithm with a heuristic backtracking search is proposed, namely CS-HBS. Extensive experiments are conducted to demonstrate the superiority of SCPM-CR, and also to validate the efficiency and scalability of CS-HBS.

Information diffusion prediction is an important task which studies how information items spread among users. Generally, information diffusion prediction is divided into microscopic, macroscopic, and unified multi-scale information prediction, which aims to predict the next influenced user, the total number of influenced users, or the above two within a unified framework, respectively. However, the existing prediction models just adopted the diffusion sequences information of cascades, ignoring two important information for diffusion prediction, i.e. the specific diffusion timestamp and the wide dispersion. The former contains the dynamic changes of users' repost preferences and the diffusion rate, while the latter reflects the user's aggregation of influenced along diffusion paths. In this paper, we propose a novel Multi-Scale Information Diffusion Prediction model (MSIDP) for predicting microscopic and macroscopic information diffusion simultaneously. MSIDP encodes timestamp information and designs a bidirectional graph convolutional network (BGCN) to jointly learn the representations of users from the deep propagation and the wide dispersion, such that more diffusion information can be captured. Extensive experiments on three real datasets demonstrate the effectiveness of our proposed model compared with 11 baseline algorithms in terms of two evaluation metrics.

As a state-of-the-art family of Unsupervised Domain Adaptation (UDA), bi-classifier adversarial learning methods are formulated in an adversarial (minimax) learning framework with a single feature extractor and two classifiers. Model training alternates between two steps: (I) constraining the learning of the two classifiers to maximize the prediction discrepancy of unlabeled target domain data, and (II) constraining the learning of the feature extractor to minimize this discrepancy. Despite being an elegant formulation, this approach has a fundamental limitation: Maximizing and minimizing the classifier discrepancy is not class discriminative for the target domain, finally leading to a suboptimal adapted model. To solve this problem, we propose a novel Class Discriminative Adversarial Learning (CDAL) method characterized by discovering class discrimination knowledge and leveraging this knowledge to discriminatively regulate the classifier discrepancy constraints on-the-fly. This is realized by introducing an evaluation criterion for judging each classifier's capability and each target domain sample's feature reorientation via objective loss reformulation. Extensive experiments on three standard benchmarks show that our CDAL method yields new state-of-the-art performance. Our code is made available at https://github.com/buerzlh/CDAL.

Recently, black-box domain adaptation attracts a lot of attention, which is a new concept to realize domain adaptation with only a cloud API service instead of the source data or well-trained source model, reflecting the focus on development of cloud services and concerns about data security. However, the existing black-box domain adaptation methods always only use high-confidence samples which limits their performance. We propose a self-alignment approach based on statistic moment matching to realize black-box domain adaptation. We construct a model for target domain in the initial stage of our work. Then, we put target data into source model API to obtain the pseudo-labels and divide the target data into high-confidence and low-confidence parts according to their pseudo-labels confidence. By matching the data distributions between these two parts and self-supervised learning on high-confidence part, the performance on both parts samples can be boosted respectively. Information maximization is also applied to the target data to further improve their classification performance. Experiment results confirm that our method achieves state-of-the-art performance.

Spatial co-location pattern mining (SCPM) aims to discover subsets of spatial features frequently located together in proximate areas. Previous studies of SCPM solely concern the inter-features association of a pattern, but neglect the interesting intra-feature behavior. In this paper, we propose the task of spatial co-location pattern mining with coupling relation consideration (SCPM-CR) to capture complex relations embedded in a co-location. Specifically, InterPCI measure is designed to capture the inter-features coupling by considering the comprehensive interaction of objects for the features in a pattern, and luckily it possesses the anti-monotone property. Another measure, IntraCAI, is proposed to capture the congregating behavior of intra-feature objects under the restriction of a co-location. A general framework is designed for SCPM-CR task and experiments show that a large fraction of computation time is devoted to identifying the participating objects of a candidate pattern. To tackle this calculation bottleneck, a novel candidate-and-search algorithm is suggested, CS-HBS, equipped with heuristic backtracking search. Extensive experiments are conducted on real and synthetic datasets to demonstrate the superiority of SCPM-CR compared with traditional SCPM methods, and also to validate the efficiency and scalability of CS-HBS. Experimental results show that CS-HBS outperforms the baselines by several times or even orders of magnitude.

Unsupervised domain adaptation aims to transfer knowledge from labeled source domain to unlabeled target domain. Recently, multisource domain adaptation (MDA) has begun to attract attention. Its performance should go beyond simply mixing all source domains together for knowledge transfer. In this article, we propose a novel prototype-based method for MDA. Specifically, for solving the problem that the target domain has no label, we use the prototype to transfer the semantic category information from source domains to target domain. First, a feature extraction network is applied to both source and target domains to obtain the extracted features from which the domain-invariant features and domain-specific features will be disentangled. Then, based on these two kinds of features, the named inherent class prototypes and domain prototypes are estimated, respectively. Then a prototype mapping to the extracted feature space is learned in the feature reconstruction process. Thus, the class prototypes for all source and target domains can be constructed in the extracted feature space based on the previous domain prototypes and inherent class prototypes. By forcing the extracted features are close to the corresponding class prototypes for all domains, the feature extraction network is progressively adjusted. In the end, the inherent class prototypes are used as a classifier in the target domain. Our contribution is that through the inherent class prototypes and domain prototypes, the semantic category information from source domains is transformed into the target domain by constructing the corresponding class prototypes. In our method, all source and target domains are aligned twice at the feature level for better domain-invariant features and more closer features to the class prototypes, respectively. Several experiments on public data sets also prove the effectiveness of our method.

A co-location pattern is a subset of spatial features whose instances are frequently located together in proximate areas. Mining co-location patterns can discover spatial dependencies in spatial datasets and have particular value in many applications. However, it is challengeable to discover co-location patterns from massive spatial datasets, due to the expensive computational cost. In this paper, we present a novel parallel co-location pattern mining approach. First, dividing spatial neighbor relationships into some neighbor-dependency partitions enables to perform mining task on each partition independently in parallel. Then, a column-based calculation approach is proposed to replace the time-consuming generation of table instances for calculating the prevalence of patterns. To further reduce the search space of patterns on each partition, two pruning strategies are suggested. We implement the parallel co-location pattern mining algorithm based on neighbor-dependency partition and column calculation via MapReduce, named PCPM-NDPCC. Substantial experiments are conducted on real and synthetic datasets to examine the performance of PCPM-NDPCC. Experimental results reveal that PCPM-NDPCC has a significant improvement in efficiency than baseline algorithms and shows better scalability for massive spatial data processing.

Abstract Multi-view clustering (MVC), which aims to explore the underlying structure of data by leveraging heterogeneous information of different views, has brought along a growth of attention. Multi-view clustering algorithms based on different theories have been proposed and extended in various applications. However, most existing MVC algorithms are shallow models, which learn structure information of multi-view data by mapping multi-view data to low-dimensional representation space directly, ignoring the nonlinear structure information hidden in each view, and thus, the performance of multi-view clustering is weakened to a certain extent. In this paper, we propose a deep multi-view clustering algorithm based on multiple auto-encoder, termed MVC-MAE, to cluster multi-view data. MVC-MAE adopts auto-encoder to capture the nonlinear structure information of each view in a layer-wise manner and incorporate the local invariance within each view and consistent as well as complementary information between any two views together. Besides, we integrate the representation learning and clustering into a unified framework, such that two tasks can be jointly optimized. Extensive experiments on six real-world datasets demonstrate the promising performance of our algorithm compared with 15 baseline algorithms in terms of two evaluation metrics.

Multi-view clustering has attracted increasing attention in recent years since many real data sets are usually gathered from different sources or described by different feature types. Amongst various existing multi-view clustering algorithms, those that are based on non-negative matrix factorization (NMF) have exhibited superior performance. However, NMF decomposing original data directly fails to exploit global relationships between data samples and cannot be applied to datasets that are not strictly non-negative. In this paper, a network-based sparse and multi-manifold regularized multiple NMF (NSM_MNMF) for multi-view clustering is proposed, where multi-view data is transformed into multiple networks, and NMF is used to jointly factorize transformed multiple networks for capturing the shared cluster structure embedded in different views. Furthermore, sparse and multi-manifold regularization are incorporated into NMF to keep the intrinsic geometrical information of the multi-view network manifold space. Networks characterize intra-view similarity, and joint factorization reveals inter-view similarity across distinct views, while using NMF to decompose the networks instead of the original data means NSM_MNMF can be applied to datasets that are not strictly non-negative and the clustering results are interpretable. Extensive experiments are conducted on nine real data sets to assess the method proposed, and the results illustrate that NSM_MNMF outperforms other baseline approaches.

Multi-view clustering aims to group similar samples into the same clusters and dissimilar samples into different clusters by integrating heterogeneous information from multi-view data. Non-negative matrix factorization (NMF) has been widely applied to multi-view clustering owing to its interpretability. However, most NMF-based algorithms only factorize multi-view data based on the shallow structure, neglecting complex hierarchical and heterogeneous information in multi-view data. In this paper, we propose a deep multiple non-negative matrix factorization (DMNMF) framework based on AutoEncoder for multi-view clustering. DMNMF consists of multiple Encoder Components and Decoder Components with deep structures. Each pair of Encoder Component and Decoder Component are used to hierarchically factorize the input data from a view for capturing the hierarchical information, and all Encoder and Decoder Components are integrated into an abstract level to learn a common low-dimensional representation for combining the heterogeneous information across multi-view data. Furthermore, graph regularizers are also introduced to preserve the local geometric information of each view. To optimize the proposed framework, an iterative updating scheme is developed. Besides, the corresponding algorithm called MVC-DMNMF is also proposed and implemented. Extensive experiments on six benchmark datasets have been conducted, and the experimental results demonstrate the superior performance of our proposed MVC-DMNMF for multi-view clustering compared to other baseline algorithms.

Network embedding (NE) is a method that maps nodes in a network into a low-dimensional and continuous vector space while maintains inherent features of the network. Most existing algorithms for NE focus on one or two of the aspects of topological structure, node attributes or community structure information, but without integrating the three in a unified framework. In this study, we develop a deep neural network-based framework for Attributed Network Embedding with Community Preservation (ANECP), which simultaneously incorporates the topological structure, node attributes as well as community structure together to obtain the low-dimensional distributed representations of nodes in the network. The use of deep neural networks captures the underlying high non-linearity in both topology and attribute information, while the incorporation of the community structure resolves the issues of data sparsity from microscopic perspective. Consequently, the obtained node representations can preserve proximity and discriminative. We conducted experimental studies using six real-world datasets. The experimental results show that proposed ANECP has superior performance over the existing methods.

Spatial co-location patterns are groups of spatial features whose instances are frequently located together in spatial proximity. Most existing algorithms of discovering spatial co-location patterns are based on the candidate-test model, which is computationally expensive. When the user adjusts the participation index (PI) threshold, these algorithms have to be re-executed from the size 2 co-location patterns. In this paper, we propose a novel spatial instance partition method for mining co-location patterns which called overlap maximal clique partitioning algorithm (OMCP). The OMCP co-location mining algorithm divides instances of an input spatial dataset into a set of overlap maximal cliques. Table instances of all colocation patterns are collected by the overlap maximal cliques. Prevalent co-location patterns are directly calculated without generating the candidate patterns. The OMCP algorithm only needs to execute once to get the PI of all patterns, without re-executing when the PI threshold is adjusted. Our algorithm is performed on both synthetic and real-world datasets to demonstrate that the OMCP algorithm improvements in efficiency of co-location pattern mining.

Influence is a complex and subtle force that changes the behavior of involved users. Measuring influence can benefit to identify the influential users, and also benefit to provide important insights into the design of social platforms and applications. However, most existing work on social influence analysis has focused on homogeneous information networks. Few studies systematically investigate how to mine the strength of influence between nodes in heterogeneous information networks. In this paper, we present a meta path-based information entropy for modeling social influence in heterogeneous information networks (MPIE). Through setting meta paths, MPIE not only flexibly integrates heterogeneous information, but also obtains potential link information to measure the influence of nodes. Experiments on real data sets demonstrate the effectiveness of our proposed method.

Spatial co-location pattern mining is an interesting and important task in spatial data mining which discovers the subsets of spatial features frequently observed together in nearby geographic space. However, the traditional framework of mining prevalent co-location patterns produces numerous redundant co-location patterns, which makes it hard for users to understand or apply. In this paper we study the problem of reducing redundancy in a collection of prevalent co-location patterns. We first introduce the concept of semantic distance between two co-location patterns, and then define redundant co-locations by introducing the concept of δ-covered, where δ (0≤δ≤1) is a coverage measure. We develop two algorithms RRclosed and RRnull to perform the redundancy reduction for prevalent co-location patterns. Our performance studies on the synthetic and real-world data sets demonstrate that our method effectively reduces the size of the original collection of closed co-location patterns by about 50%. Furthermore, the RRnull method runs much faster than the related closed co-location pattern mining algorithm.

Time series data mining has attracted a lot of attention in the last decade, especially the research on the clustering of time series data. Network-based clustering technology, transforming data of time series into a network and then used community detection methods of network to cluster time series, is a new approach to cluster time series data. This approach takes the advantage that a network can describe the relationship between any pair or any group of data samples, but the effectiveness of clustering heavily dependent on the performance of algorithms of community detection. In this paper, we cluster time series by transforming them into network and detecting communities by non-negative matrix factorization (NMF). Experimental evaluations illustrate the superiority of our approach compared with the state-of-the-arts such as Multilevel.

With the in-depth research of data classification, multi-label classification has become a hot issue of research. Multi-label k-nearest neighbor (ML- k NN) is a classification method which predicts the unclassified instances' labels by learning the classified instances. However, this method doesn't consider the interrelationships between attributes and labels. Considering the relationships between properties and labels can improve accuracy of classification methods, but the diversities of properties and labels will present the curse of dimensionality. This problem make such methods can not be expanded under the background of big data. To solve this problem, this paper proposes three methods, called multi-label k-nearest neighbor based on principal component analysis(PML- kNN), coupled similarity multi-label k-nearest neighbor based on principal component analysis(PCSML- kNN) and coupled similarity multi-label k-nearest neighbor classification based on feature selection (FCSML- kNN), which use feature extraction and feature selection to reduce the dimensions of labels' properties. We test the ML- kNN and the three methods we proposed with two real data, the experimental results show that reduce the dimensions of labels' properties can improve the efficiency of classification methods.

Abstract Influence diagrams (IDs) are one of the most commonly used graphical and mathematical decision models for reasoning under uncertainty. In conventional IDs, both probabilities representing beliefs and utilities representing preferences of decision makers are precise point‐valued parameters. However, it is usually difficult or even impossible to directly provide such parameters. In this paper, we extend conventional IDs to allow IDs with interval‐valued parameters (IIDs) and develop a counterpart method of Copper's evaluation method to evaluate IIDs. IIDs avoid the difficulties attached to the specification of precise parameters and provide the capability to model decision‐making processes in a situation that the precise parameters cannot be specified. The counterpart method to Copper's evaluation method reduces the evaluation of IIDs into inference problems of Bayesian networks with interval‐valued probabilities. An algorithm based on the approximate inference of Bayesian networks with interval‐valued probabilities is proposed, and extensive experiments are conducted. The experimental results indicate that the proposed algorithm can find the optimal strategies effectively in IIDs, and the interval‐valued expected utilities obtained by proposed algorithm are contained in those obtained by exact evaluating algorithms. The newly development approach would significantly extend the range of IDs for managerial decision support applications where parameters of variables can be specifically defined by only estimated intervals.

Spatial co-location pattern mining is an interesting and important task in spatial data mining which discovers the subsets of spatial features frequently observed together in nearby geographic space. However, the traditional framework of mining prevalent colocation patterns produces numerous redundant co-location patterns, which makes it hard for users to understand or apply. To address this issue, in this paper, we study the problem of reducing redundancy in a collection of prevalent co-location patterns by utilizing the spatial distribution information of co-location instances. We first introduce the concept of semantic distance between a co-location pattern and its super-patterns, and then define redundant co-locations by introducing the concept of d-covered, where δ (0 ≤ δ ≤ 1) is a coverage measure. We develop two algorithms RRclosed and RRnull to perform the redundancy reduction for prevalent co-location patterns. The former adopts the post-mining framework that is commonly used by existing redundancy reduction techniques, while the latter employs the mine-and-reduce framework that pushes redundancy reduction into the co-location mining process. Our performance studies on the synthetic and real-world data sets demonstrate that our method effectively reduces the size of the original collection of closed co-location patterns by about 50 percent. Furthermore, the RRnull method runs much faster than the related closed co-location pattern mining algorithm.

Abstract Detecting communities is of great importance in social network analysis. However it is an issue that has not yet been satisfactorily solved, despite the efforts made by interdisciplinary research communities over the past few years, because of the nature of complexity in deciding how community structures should be recognized. In this paper we propose an approach based on cooperative game theory for community detection in social networks. We regard individuals as players, and regard communities as coalitions formed by players, and model community detection problem as the formation and optimization of coalitions. Furthermore, we define coalition profile for players to indicate coalitions that players joined, the order of a coalition profile is defined as the number of coalitions in a coalition profile, and we introduce a utility function to measure preference of coalition profiles. Accordingly, we propose an algorithm to detect a coalition profile with maximal utility function values. We have implemented the algorithms developed in this study and experimental results demonstrate the effectiveness of our approaches.

Online community detection is essential for social network analysis. Modularity is a quality function used to measure the strength of the community structure discovered with social networks. Existing methods detect community structures through modularity analysis. However, the existing modularities are unable to identify community structures correctly when communities are very different in size, in particular, when the size of some communities is very small compared with others. To address this problem, we propose the concept of a coupling coefficient between two communities and define a new modularity, MC modularity, to evaluate the quality of the discovered community structures. This method can provide adequate measures for the quality of community structures. In addition, we develop the DC_MC algorithm to detect community structures based on the concept of MC modularity. An algorithm for shifting a group of nodes, instead of one node at a time, amongst communities also is designed to achieve optimal results. The effectiveness and efficacy of the approach we proposed have been demonstrated through a set of experiments involving real data benchmarks and synthetic data sets that are purpose-built for evaluating different types of networks.

Abstract Many decisions are made in interactive situations in which decision makers interact and may also affect each other's decision outcomes. In order to model decision makings in such interactive situations, this paper incorporates game theory into influence diagrams and presents a new approach, called game theory‐based influence diagrams (GIDs). GIDs consider an extra factor, that of the choice of strategies made by other decision makers, to the list of determinants that influence the decision‐making process of each decision maker, so that decision makers can make more rational decisions. As a result of integrating influence diagrams and game theory, GIDs benefit from the simplicity and efficiency of influence diagrams for modelling complex decision problems as well as the rationality and suitability of applying game theory for making decisions in dynamic interactive scenarios. This paper also introduces genetic algorithm‐based methods to evaluate GIDs. Experimental studies have been performed for validation and evaluation.

Being source of Inland River in arid areas, the mountain landscape pattern influence the hydrological process, and play an important role in ecological balance. It is essential for us to understand landscape pattern dynamic changes and the driving forces to estimate the characteristics of development the landscape pattern and water resource. However, knowledge of the grassland degradation in vertical variation and the dominant factors of in arid mountain areas are largely limited. On the basis of Landsat TM images and DEM data, the changes of landscape pattern and the driving factors were analyzed in different altitude zones in the source area of Shule River. The results indicated that: The grassland in the altitude zone of 3800~3900 m degraded markedly obvious, and the altitude where the grassland degenerated had gone upwards slightly. The most efficient driving force of grassland degradation is that the permafrost degradation caused by rising temperature.

The landscape change of desertification process and human activity impacts on desertification had been analyzed on the case of Yanchi County in Northern China. It is helpful to understand the process of desertification, and can help to understand how human activities contribute on the landscape change that caused to desertification. The results show that the landscape change of desertification process is mainly represented as grassland change to farmland and deserted grassland, and then change to desertifed land or sand land. Human activities can dramatically change land-use patterns and the landscape. Due to this paper's analysis, households' economic behaviors should be the most important human factors causing desertification in Northern China. As the case study of Yanchi County, the human activity factor contribution rate can be more than 60% or even more than 70%. However, human activity can often be affected by state policies, so that cause to negative or positive effects on desertification process.

Oasis is a specific landscape that exists with deserts in arid regions. In this paper, the oasis landscape changes in middle-lower reaches of Shule River basin had been analyzed. Using the Landsat TM imagines of June in 1987 and August in 2003 as the digital sources, the vector and grid landscape maps of distributions of the study area were derived. The dynamics variation of the oasis landscape in that basin was got on the basis of the software of ArcGIS 9.0, and the forecast on the dynamic variation of landscapes was executed using the software of the Matlab 6.3. At the same time, the eco-environmental effects of landscape change had been analyzed. The results showed that the landscape change with the main feature of increasing cropland area led to a series of ecological environment problem, such as deterioration of water quality, natural vegetation decreasing, and the eco-environmental system deteriorating. Human activities, especially in relation to water utilization and population increment, have gradual became the most active and dominant driving factors in the landscape change.

Taking Yanchi County as a case study, the driving factors of desertification had been analyzed. The results show that the remarkable feature of desertification of Yanchi County is the landscape changing from grassland and farmland to desert. At the same time, the fragmentation of grassland landscape is increasing continually, while the farmland is linked up on space with the decreasing landscape fragmentation. Land desertification was influenced mostly by the distribution of settlement, and the affected distance of habitation to the farmland extended gradually. Human activities and policies can dramatically change land-use patterns. Therefore, it is important to analyze the human activities and implement appropriate state policies to combat desertification in China. The conclusions of this research can provide the scientific basis of desertification control for Yanchi County and other similar regions.

Influence diagrams (IDs) are a popular framework representing a decision maker's belief and preferences about a sequence of decisions to be made under uncertainty. The quantification of IDs which consists of defining conditional probabilities for chance nodes and utility functions for value nodes is not always obvious. In fact, decision makers cannot always provide exact utilities and in some cases, it is more convincible to describe utilities with interval values than precise ones. This paper is about extending general IDs (with precise single-valued parameters) into IDs with interval-valued utilities. Such extension is interesting because it enables one to model decision making processes in the situation that utilities are represented by interval values. This paper extend Gibbs sampling algorithm to the Bayesian network (BN) containing interval-valued probabilities for approximate inference and propose a method to evaluate IDs with interval-valued utilities based on the BN inference.

Landscape changes are among the most important socio-economic forces of global as well as local environmental change. Mountain is a source of water in the arid regions, playing an important role in keeping ecological balance. This study focuses on the changes of mountain landscape in Shule River basin located in arid Northwest China. Based on the Landsat TM remote sensing satellite data of the study areas taken in 1986 and 2000, DEM (1:100000), using ArcGIS software, the landscape changes had been analyzed in the different elevation zones. The results show that: (1) Area of the sparse grassland is the largest in all the landscape types, the followed is rock. The vegetation of the study is 50%, and the dense grassland accounted for only 4.5%. (2) During the 15 years, the area of the dense grassland decreased 299.45 km2because of degradation, the mid-density grassland decreased 91.5 km2. On the other hand, the area of the swamp meadow increased 168.92 km2, the freezing-throwing rock increased 101.28 km2, and the sparse grassland increased 95.3 km2. (3) The grassland in lower elevation with 3500 m-3800 m were undergoing degeneration because of the global warming and over-grazing, and the altitude of 3500 m-3800 m was just the lower limit of permafrost distribution.

Arid and semi-arid areas are very easily influenced by global warming because of the simple structure of landscape. The change of climate conditions affected the land vegetation coverage and its distribution in arid areas. As an indicator of vegetation growth and its spatial distribution, NDVI has a close relationship with climate conditions. Based on the remote sensing data GIMMS NDVI and software of ArcGIS, it was analyzed that the spatial-temporal correlations of NDVI with the precipitation and temperature in the Qilian Mountain located in the arid Northwest China in this paper from 1982 to 2003. The results showed that: (1) There was significant positive correlation between NDVI and temperature in the areas above the elevation of 2500 m, and significant negative correlation between NDVI and temperature in the areas below the elevation of 2500 m. (2) The value of NDVI was not visibly related with the precipitation in the both elevation areas. It can be concluded from the results that the vegetation coverage in lower elevation areas in Northwest China will be worse with the global warming.

Broadcast authentication is an important security primitive in wireless sensor networks. Costs of Tree-based µTESLA are rising rapidly with more senders in networks, which limit its use in large-scale networks. To removing such a limit, this paper presents a µTPCT-based broadcast authentication protocol which mainly rests on constructions of µTPC (µTESLA Parameters Hash Chain) and µTPCT (µTPC Merkle Tree). The proposed protocol makes an improvement of Tree-based µTESLA. It constructs µTPC to distribute and authenticate the initial parameter of µTESLA. Then it constructs µTPCT to distribute and authenticate the initial parameter of µTPC. Analysis and experiments have shown our proposed protocol has the ability of resisting DoS attacks against the distribution of µTESLA parameters and revoking authenticating powers of compromised nodes on time. It also outperforms the Tree-based µTESLA by reducing the cost of computation, communication, and storage needed, which makes our proposed protocol more applicable to large scale sensor networks.

In arid and semi-arid areas, a large area of soil erosion in the upper reaches will not only aggravate drought of atmospheric and soil, but also induce deterioration of grasslands and desertification. In this paper, as the Shiyang River upper reaches a case, we analyzed how the changes of land use and land cover influence the soil erosion, based on remote sensing satellite date and DEM, using ArcGIS software and soil erosion formula (RUSLE). The results show that: (1) the soil erosion intensity of farmland was the most serious among the vegetation types. The ability of soil and water conservation of forest, shrub and high covered grassland was the best, and which can effectively prevent the soil loss caused by the increase of slope gradient. (2) Lapping the land use changes map, soil erosion changes map and the slope map, the result indicated that reclamation high covered grassland to farmland in the areas of slope gradient exceeding 15 was the most mainly cause.

Conventional fuzzy cognitive maps (FCMs) can only represent monotonic or symmetric causal relationships and cannot simulate the AND/OR combinations of the antecedent nodes. The rule-based fuzzy cognitive maps (RBFCMs) usually suffer from the well-known combinatorial rule explosion problem. A hybrid fuzzy cognitive model based on weighted OWA operators and single-antecedent rules is proposed to eliminate the drawbacks of the existing FCM models. Hybrid fuzzy cognitive maps (HFCMs) represent the causal relationships with single-antecedent fuzzy rules and handle the various AND/OR relationships among the antecedent nodes with weighted OWA aggregation operators. Compared with conventional FCMs, HFCMs have more powerful cognitive capability. Compared with RBFCMs, HFCMs reduce the scale and complexity of the rule bases significantly and have better representation and inference performance. © 2007 Wiley Periodicals, Inc. Int J Int Syst 22: 1189–1196, 2007.

The defect of existing fidelity definition is discussed, by using this definition, it is found that two obviously different image can be determined to be approximate or the same. According to the defect, an improved fidelity definition is presented. The difference and relation between the two definitions are discussed and the analytic expression for deriving one definition from the other is also presented.

Grassland degradation in arid mountain areas of Northwest China mostly attributes to its backward economic situation and mismanagement. In this paper, with Sunan County in Qilian Mountains as a case study, a pasture livestock development pattern has been developed. According to this pattern, the Qilian Mountain areas and its ice-melt water irrigated oasis were regarded as a holistic coupling system. The development pattern has been divided into two sub-patterns, namely grassland management pattern, and animal production pattern. In grassland management pattern, the degraded grassland can be restored step by step by contracting grassland with households and restricting livestock numbers according to pasture carrying capacity. In animal production pattern, mountain pasture system and oasis farming system can be coupled together by developing livestock propagation in mountains and fattening in oasis. As a result, household incomes in mountain pasture area can be increased 2.5-3 times and oasis-farming area can be increased 2-5 times. The development pattern will be beneficial to increase the households' income, lighten grazing pressure, alleviate pasture degradation, and finally realize sustainable development in the arid mountain areas.

We present a Robust Disaster Recovery System Model (RDRSM) for highly security-critical systems to realize continuous service. Through strengthening safe communication ability, RDRSM guarantees secure and reliable command on disaster recovery. It uses selfsupervision function to monitor the integrality and security of system itself. Through GIS, this model makes disaster recovery system easier to use, manage and maintain. RDRSM possesses features of security, robustness and controllability and can be applied to environments such as e-government and bank. This model successfully conducted the construction of an important e-government disaster recovery system, which verified its feasibility and the significance of its components, such as disaster recovery planning, system supervision and robust communication support.

This paper presents the design and implementation of an Internet home automation system. The design is based on an embedded controller which is connected to a PC-based home Web server via RS232 serial port. The home appliances are connected to the input/output ports and the sensors are connected to the analog/digital converter channels of the embedded controller. The software of the system is based on the combination of Keil C, Java Server Pages, and JavaBeans, and dynamic DNS service (DDNS) client. Password protection is used to block the unauthorized from accessing to the server. The system is scalable and allows additional appliances to be added to it with no major changes to its core, the home appliances can be monitored and controlled locally, or remotely through a web browser from anywhere in the world provided that an Internet access is available.

This paper presents a simple yet comprehensive approach that enables the stiffness of a tripod-based parallel kinematic machine to be quickly estimated. The approach can be implemented by two steps. In the first step, the machine structure is decomposed into two substructures associated with the machine frame and the parallel mechanism. The stiffness model of each substructure is formulated by assuming that the components in the other substructure are rigid. This is followed by the second step that enables the stiffness model of the machine structure to be achieved by linear superposition. A 3D representation of the stiffness distributions within the usable workspace are depicted with comparison to those obtained through the finite element analysis.