Search

One of the common techniques to reduce the scalability problem in collaborative filtering (CF)-based recommender systems is the clustering technique, which accelerates finding the nearest neighbor users in the recommendation process. Different clustering algorithms lead to improved accuracy and diversity in recommender systems. It is challenging to develop recommender systems based on clustering with decreasing scalability and simultaneously increasing accuracy. This article proposes a new clustering-based recommender system that takes into account the theoretical properties of the Learning Automata (LA) technique. The presented clustering novelty lies in the fact that employing LA technique for the user clustering in a CF-based recommender system has not been set forth so far in a way that addresses the scalability while improving accuracy. In addition, a novel similarity metric is embedded in the proposed algorithm to measure the similarity value between users. This metric is developed as like/dislike (LD) which can significantly improve the accuracy by reducing the computational cost. Extensive simulations have been performed on real-world datasets such as MovieLens and FilmTrust, which confirm the effectiveness of the proposed algorithm. In this regard, the proposed algorithm has improved the precision between 5 and 16% on average compared to the existing state-of-the-art methods such as GA-GELS, NUSCCF, NNMF, and KL-KM. [ABSTRACT FROM AUTHOR]

In recent years, contrastive learning has been a highly favored method for self-supervised representation learning, which significantly improves the unsupervised training of deep image models. Self-supervised learning is a subset of unsupervised learning in which the learning process is supervised by creating pseudolabels from the data themselves. Using supervised final adjustments after unsupervised pretraining is one way to take the most valuable information from a vast collection of unlabeled data and teach from a small number of labeled instances. This study aims firstly to compare contrastive learning with other traditional learning models; secondly to demonstrate by experimental studies the superiority of contrastive learning during classification; thirdly to fine-tune performance using pretrained models and appropriate hyperparameter selection; and finally to address the challenge of using contrastive learning techniques to produce data representations with semantic meaning that are independent of irrelevant factors like position, lighting, and background. Relying on contrastive techniques, the model efficiently captures meaningful representations by discerning similarities and differences between modified copies of the same image. The proposed strategy, involving unsupervised pretraining followed by supervised fine-tuning, improves the robustness, accuracy, and knowledge extraction of deep image models. The results show that even with a modest 5% of data labeled, the semisupervised model achieves an accuracy of 57.72%. However, the use of supervised learning with a contrastive approach and careful hyperparameter tuning increases accuracy to 85.43%. Further adjustment of the hyperparameters resulted in an excellent accuracy of 88.70%. [ABSTRACT FROM AUTHOR]

$ K $-means problem (KMP) is one of the most well-known clustering problems in unsupervised machine learning. The Lloyd's algorithm, also known as the $ k $-means algorithm, is a popular algorithm for solving KMP, whose each iteration consists of an update step to locate cluster centers and an assignment step to allocate data points. In this paper, we discuss a generalized $ k $-means problem (GKMP) in $ m $-dimensional Euclidean space with considering a new similarity metric, the weights of data points and the affine and locational constraints on cluster centers. The GKMP has extensive applications in practice, but is more difficult to solve than existing $ k $-means problems. In order to solve it efficiently, we propose a new algorithm in the spirit of Lloyd's algorithm. Specifically, we reformulate the update subproblems into separable minimization problems and two customized alternating direction methods of multipliers (ADMMs) are contributed to solving the resulted problems. The global convergence of customized ADMMs is then established. With implementing the customized ADMMs to locate cluster centers and the nearest center reclassification to allocate data points, an ADMM-based $ k $-means algorithm is proposed for GKMP and the convergence of iterative sequence is proved. Some numerical results verify the effectiveness of GKMP and proposed algorithms. [ABSTRACT FROM AUTHOR]

Autoencoders (AEs) are central to hyperspectral anomaly detection, given their impressive efficacy. However, the current methodologies often neglect the global pixel similarity of the hyperspectral image (HsI), thereby limiting reconstruction accuracy. This study introduces an innovative pixel-associated AE approach that leverages pixel associations to augment hyperspectral anomaly detection. First, a dictionary construction methodology was introduced based on superpixel distance estimation to construct distinct dictionaries for background and local anomalies. Second, to recognize pixel similarities, the similarity metric of each pixel from the original HsI to the background dictionary and to the local anomaly dictionary was employed as the AE network input in lieu of the original HsI. Third, a dual hidden-layer feature similarity constraint network was proposed to enhance the reconstruction error of background and anomaly targets. Finally, the reconstruction error was utilized to score the anomaly target. The proposed method was benchmarked against other state-of-the-art techniques using synthetic and real HsI datasets to assess its effectiveness. The experimental results demonstrated the superior performance of the proposed method, outperforming the alternatives.

In recent years, contrastive learning has been a highly favored method for self-supervised representation learning, which significantly improves the unsupervised training of deep image models. Self-supervised learning is a subset of unsupervised learning in which the learning process is supervised by creating pseudolabels from the data themselves. Using supervised final adjustments after unsupervised pretraining is one way to take the most valuable information from a vast collection of unlabeled data and teach from a small number of labeled instances. This study aims firstly to compare contrastive learning with other traditional learning models; secondly to demonstrate by experimental studies the superiority of contrastive learning during classification; thirdly to fine-tune performance using pretrained models and appropriate hyperparameter selection; and finally to address the challenge of using contrastive learning techniques to produce data representations with semantic meaning that are independent of irrelevant factors like position, lighting, and background. Relying on contrastive techniques, the model efficiently captures meaningful representations by discerning similarities and differences between modified copies of the same image. The proposed strategy, involving unsupervised pretraining followed by supervised fine-tuning, improves the robustness, accuracy, and knowledge extraction of deep image models. The results show that even with a modest 5% of data labeled, the semisupervised model achieves an accuracy of 57.72%. However, the use of supervised learning with a contrastive approach and careful hyperparameter tuning increases accuracy to 85.43%. Further adjustment of the hyperparameters resulted in an excellent accuracy of 88.70%.

Medical image registration is an integral part of various clinical applications including image guidance, motion tracking, therapy assessment and diagnosis. We present a robust approach for mono-modal and multi-modal medical image registration. To this end, we propose the novel shape operator based local image distance (SOLID) which estimates the similarity of images by comparing their second-order curvature information. Our similarity metric is rigorously tailored to be suitable for comparing images from different medical imaging modalities or image contrasts. A critical element of our method is the extraction of local features using higher-order shape information, enabling the accurate identification and registration of smaller structures. In order to assess the efficacy of the proposed similarity metric, we have implemented a variational image registration algorithm that relies on the principle of matching the curvature information of the given images. The performance of the proposed algorithm has been evaluated against various alternative state-of-the-art variational registration algorithms. Our experiments involve mono-modal as well as multi-modal and cross-contrast co-registration tasks in a broad variety of anatomical regions. Compared to the evaluated alternative registration methods, the results indicate a very favorable accuracy, precision and robustness of the proposed SOLID method in various highly challenging registration tasks. [ABSTRACT FROM AUTHOR]

High-quality and timely peer review for scientific works has become increasingly acute recently. Scientific reviewers peer review journal articles, conference papers, monographs, PhD-thesis, grants etc. Scientific reviewers are assigned mainly manually due to the lack of time. Having large volumes of content to review, a high-quality peer review is hard to get. In recent years there has been an increase in research on reviewer assignment automation. Also, the formal comparison of a reviewer’s research domains and a proposal’s research domains, or character-level comparison of keywords do not always provide high-quality assignments. In this paper, we perform a review on current methods of automated scientific reviewer assignment. There are 3 stages of reviewer assignment process: 1) creating reviewers’ database and structuring the information about reviewers and proposals; 2) calculating the similarity score between a proposal and a reviewer; 3) finding the appropriate assignment of proposals to reviewers, and maximizing the aggregated similarity and overall topic coverage over all assignments. Two main variations of reviewer assignment problem are considered: single and multiple reviewer assignment problem. Methods for structuring the information about proposals and reviewers based on statistical analysis of text, topic modeling and deep learning are analyzed. In the third stage, we considered possible optimal criteria for optimizing reviewers’ assignments to proposals, and also constraints, that ensure a certain level of reviewers-proposal concordance such as overall topic coverage, fair peer review, reviewers workload etc. The problem of optimal reviewer assignment is NPcomplete, therefore for solving different heuristics and meta-heuristics algorithms are used. We also present perspectives of future research on the automated reviewer assignment. [ABSTRACT FROM AUTHOR]

Cross-lingual text summarization is a highly desired service for overseas report editing tasks and is formulated in a distributed application to facilitate the cooperation of editors. The multilanguage pre-trained language model (MPLM) can generate high-quality cross-lingual text summaries with simple fine-tuning. However, the MPLM does not adapt to complex variations, like the word order and tense in different languages. When the model performs on these languages with separate syntactic structures and vocabulary morphologies, it will lead to the low-level quality of the cross-lingual summary. The matter worsens when the cross-lingual summarization datasets are low-resource. We use a knowledge distillation framework for the cross-lingual summarization task to address the above issues. By learning the monolingual teacher model, the cross-lingual student model can effectively capture the differences between languages. Since the teacher and student models generate summaries in two languages, their representations lie on different vector spaces. In order to construct representation relationships across languages, we further propose a similarity metric, which is based on bidirectional semantic alignment, to map different language representations to the same space. In order to improve the quality of cross-lingual summaries further, we use contrastive learning to make the student model focus on the differentials among languages. Contrastive learning can enhance the ability of the similarity metric for bidirectional semantic alignment. Our experiments show that our approach is competitive in low-resource scenarios on cross-language summarization datasets in pairs of distant languages. [ABSTRACT FROM AUTHOR]

Neighbor-based collaborative filtering is one of the prevalent recommendation approaches that apply similarity algorithms. Using a similarity algorithm to improve the accuracy of the recommendation system is one challenge in collaborative filtering. The computation of similarity now heavily relies on user rating and behavior scores. Users' preferences for each product kind (genre) demonstrate the user behavior's value. The algorithm's weakness is that it only considers genre data when estimating user behavior, regardless of when the user rated the item. Therefore, this study aims to develop a new similarity algorithm by considering the genre and the time weight of the item rating, which is called time loss function-based similarity (TLFSim). Newly assessed items have a higher weight than those estimated for a long time. Our experiment tested the TLFSim's performance compared to the userscore-probability-collaborative-filtering (UPCF) algorithm using the MovieLens 100k dataset. The experimental results demonstrate that the TLFSim algorithm surpasses the prior approach regarding recommendation accuracy, reducing mean absolute error (MAE) by 8.28% and root mean square error (RMSE) by 4.57%. [ABSTRACT FROM AUTHOR]

Multimodal image registration is vital in Deep Brain Stimulation (DBS) surgery. DBS treats movement disorders by implanting a neurostimulator device in the brain to deliver electrical impulses. Image registration between computed tomography (CT) and cone beam computed tomography (CBCT) involves fusing images with a specific field of view (FOV) to visualize individual electrode contacts. This contains important information about the location of segmented contacts that can reduce the time required for electrode programming. We performed a semi-automated multimodal image registration with different FOV between CT and CBCT images due to the tiny structures of segmented electrode contacts that necessitate high accuracy in the registration. In this work, we present an optimization workflow for multi-modal image registration using a combination of different similarity metrics, interpolators, and optimizers. Optimization-based rigid image registration (RIR) is a common method for registering images. The selection of appropriate interpolators and similarity metrics is crucial for the success of this optimization-based image registration process.We rely on quantitative measures to compare their performance. Registration was performed on CT and CBCT images for DBS datasets with an image registration algorithm written in Python using the Insight Segmentation and Registration Toolkit (ITK). Several combinations of similarity metrics and interpolators were used, including mean square difference (MSD), mutual information (MI), correlation and nearest neighbors (NN), linear (LI), and B-Spline (SPI), respectively. The combination of a correlation as similarity metric, B-Spline interpolation, and GD optimizer performs the best in optimizing the 3D RIR algorithm, enhancing the visualization of segmented electrode contacts. Patients undergoing DBS therapy may ultimately benefit from this.

In recent years, electromagnetic spectrum has become an indispensable national strategic resource in the information age and an important strategic support for the development of national informatization. With the rapid proliferation of wireless data today, the demand for electromagnetic spectrum is also growing rapidly, and spectrum resources are limited, leading to the increasingly prominent problem of spectrum scarcity. In response to the traditional spectrum prediction methods for spectrum prediction with extended time and low accuracy, this paper analyzes the spectrum correlation of different channels that are in the same service, performs the similarity measure of frequency dimension data, and proves the correlation between the spectrum; and predicts the spectrum data with different time occupancy, and conducts deep mining of the spectrum target behavior from the time–frequency dimension. The experiments mainly use sequence to sequence (Seq2Seq) based Long Short-Term Memory (LSTM) to predict the data of each occupancy degree, and compare with the traditional LSTM network, Autoregressive Integrated Moving Average model (ARIMA), etc. The prediction accuracy of the method in this paper is higher. At the same time, the abnormal spectrum is monitored using the Mahalanobis distance algorithm, the detection accuracy reaches 100%. [ABSTRACT FROM AUTHOR]

Graph embedding transforms high-dimensional graphs into a lower-dimensional vector space while preserving their structural information and properties. Context-sensitive graph embedding, in particular, performs well in tasks such as link prediction and ranking recommendations. However, existing context-sensitive graph embeddings have limitations: they require additional information, depend on community algorithms to capture multiple contexts, or fail to capture sufficient structural information. In this paper, we propose a novel Graph Embedding with Similarity Metric Learning (GESML). The core of GESML is to learn the optimal graph structure using an attention-based symmetric similarity metric function and establish association relationships between nodes through top-k pooling. Its primary advantage lies in not requiring additional features or multiple contexts, only using the symmetric similarity metric function and pooling operations to encode sufficient topological information for each node. Experimental results on three datasets involving link prediction and node-clustering tasks demonstrate that GESML significantly improves learning for all challenging tasks relative to a state-of-the-art (SOTA) baseline. [ABSTRACT FROM AUTHOR]

The paper considers the application of an algorithm based on a morphological projector for determining structural differences for comparing dermoscopic images. This will allow to identify changes that have occurred in skin lesions over time, for a more accurate diagnosis of their malignancy. The proposed algorithm makes it possible to detect differences in images even if there is a significant difference in the brightness and color levels of the compared images, and also ignores small insignificant details, such as noise, dermatoscope optics marks, hair, etc. A method for correcting the desynchronization of images using the structural similarity index as a similarity metric, and the sinecosine algorithm as an optimization algorithm is proposed. The proposed algorithms were tested on dermatoscopic images and the possibility of their application was demonstrated.

Recently, both ensemble clustering and semi-supervised clustering have emerged as important paradigms of traditional clustering. Ensemble clustering seeks to integrate multiple clustering results from different methods or the same methods with different parameters. Semi-supervised clustering involves using a small amount of class membership information in some samples for the learning process. Meanwhile, Semi-Supervised Ensemble Clustering (SSEC) has attracted increasing attention due to its high performance. However, most SSEC algorithms are configured based on partitional clustering techniques, and there are few attempts on hierarchical clustering techniques. Even in existing hierarchy-based SSEC algorithms, prior knowledge is not sufficiently used and is often applied to create primary partitions. To address these problems, we propose a Semi-supervised Hierarchical Ensemble Clustering framework based on a novel Similarity metric and stratified feature Sampling, which we call SHECSS. SHECSS uses the information of all primary partitions according to their strength to calculate the similarity between samples. Also, SHECSS is equipped with a stratified feature sampling mechanism that can improve the diversity of primary partitions and deal with high-dimensional data. Here, the primary partitions are created based on multiple hierarchical clustering techniques, and the target partition is configured by a consensus function based on the clusters clustering policy. Experimental results show the effectiveness and efficiency of SHECSS compared to representative clustering methods.

Image matching between optical and SAR images remains challenging due to their different imaging principles. Despite recent developments in image-matching methods for optical-to-SAR images, few have seen practical application success due to the significant geometric and radiometric differences between the two modalities. In this letter, we propose a new method called the orientated modality independent region descriptor (OMIRD) to overcome these challenges. Our template-based approach effectively extracts consistent structural and textural features from both optical and SAR images. Additionally, we introduce a principal orientation selection method, using the ratio of exponentially weighted averages (ROEWA) and logarithm histogram of gradient (HOG) to identify the most meaningful feature orientations and further simplify the OMIRD descriptor. The experimental results indicate that the proposed OMIRD exhibits superior matching accuracy and robustness under various geometric distortions and non-linear grayscale variations. [ABSTRACT FROM AUTHOR]

Pedicle screw placement for vertebral fixation is a complicated surgery for orthopaedic surgeons. The main challenge is to estimate the accurate trajectory’s position to minimize post-operative complications related to pedicle screw placement. Different types of 3D to 2D registration techniques have been employed to avoid the misplacement of the screw during the surgery. However, these techniques cannot be applied directly to MR to X-ray registration due to differences in image intensity and tissue non-correspondence. To overcome these limitations, feature-based 3D to 2D registration technique was developed to map a trajectory position in the intra-operative X-ray image on to the pre-operative MR image. The registration framework validated by generating projection images that perfectly matched simulated X-ray images, then back-projecting the trajectory position on the pre-operative MR image using the estimated transformation parameters. The accuracy of the registered trajectory evaluated by measuring the displacement and directional errors between the registered and planned trajectory. The proposed method successfully registered the trajectory position in the simulated X-ray to pre-operative MR to estimate the trajectory position. A number of experiments are performed on the simulated dataset to assess the effectiveness of the proposed method. The Euclidean distance between the entry and end points and the directional error of the registered trajectory from the planned trajectory were below 1mm in AP, Lateral, and a combination of both planes. The mean trajectory length difference between the planned and registered trajectory was less than 1mm.

Copyright of Tobacco Science & Technology is the property of Tobacco Science & Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This article introduces an algorithm for facial expression recognition (FER) using deep Siamese Neural Networks (SNNs) that preserve the local structure of images in the embedding similarity space. We designed the network to reveal the input pairs similarity by comparing features through a designed metric. Furthermore, we developed a novel image pairing (i.e., positive and negative pairs) strategy technique to train our Siamese model. Our Siamese model comprises of a verification framework and an identification framework to learn a joint embedding space. The verification path reduces the intra-class variations by minimizing the distance between the extracted features from the same class, while the identification path increases the inter-class variations by maximizing the distance between the features extracted from different classes. We apply transfer learning to only use the identification model for facial expression classification. We evaluated our algorithm using AffectNet, FER2013, and Compound Facial Expressions of Emotion (CFEE) datasets, where better results are achieved compared to other deep learning-based approaches. [ABSTRACT FROM AUTHOR]

This paper proposes a new similarity metric, Vector Similarity Metric (VSM), which is as simple as the popular Cosine Similarity Metric (CSM). The CSM has a major deficiency. It yields the same value, irrespective of how different the two vectors are in their sizes so long as the angle between them is the same. This deficiency remains intact even when Natural Language Processing is used to associate semantic meanings to the words/phrases and when the term frequency is modified using Inverse Document Frequency. This deficiency becomes a serious concern when one is comparing the risk profile of one company with the risk profile of another company or investigating the changes in the risk profile of a company from one year to another. The VSM is based on the difference of the two vectors. The paper demonstrates the superiority of VSM over CSM analytically and through real-world examples. [ABSTRACT FROM AUTHOR]

Cold-hot nature theory is the core basic theory of the nature of Chinese herbal medicines (CHMs). It is found that the material basis of cold-hot nature is CHM ingredients. In view of this, our group proposed a scientific hypothesis that “CHMs with similar nature should have similar material basis”. To demonstrate this hypothesis, we developed a novel multi-wavelength high performance liquid chromatography (HPLC) fingerprint similarity metric scheme for cold-hot nature identification. We explored a multi-wavelength distance metric learning model to compute the similarity of CHM ingredients, and developed an improved k-nearest neighbor algorithm based on multi-wavelength HPLC fusion (KMHF) to predict cold-hot nature of CHMs. Firstly, multi-wavelength HPLC fingerprints were used to extract the characteristic information of CHM ingredients. Secondly, we defined the similarity of CHM ingredients as semantic relevance and fingerprint similarity. We studied a multi-wavelength distance metric to measure the similarity of CHM ingredients. The learned distance metric could discover complementary characteristics of different wavelength HPLC through an optimization algorithm. Finally, an improved multi-wavelength k-nearest neighbor algorithm KMHF was proposed to analyze the relationship between cold-hot nature and CHM ingredients. Numerous experiments were designed to test the feasibility of the proposed KMHF algorithm. Experimental results indicate that the performance of our KMHF algorithm outperforms that of the compared algorithms. Experimental results demonstrate that the hypothesis that CHMs with similar cold-hot nature have similar material basis. The KMHF model is evaluated to be feasible for nature identification.

The non-uniformity of the readout circuit response in the infrared focal plane array unit detector can result in fixed pattern noise with stripe, which seriously affects the quality of the infrared images. Considering the problems of existing non-uniformity correction, such as the loss of image detail and edge blurring, a multi-scale residual network with attention mechanism is proposed for single infrared image stripe noise removal. A multi-scale feature representation module is designed to decompose the original image into varying scales to obtain more image information. The product of the direction structure similarity parameter and the Gaussian weighted Mahalanobis distance is used as the similarity metric; a channel spatial attention mechanism based on similarity (CSAS) ensures the extraction of a more discriminative channel and spatial feature. The method is employed to eliminate the stripe noise in the vertical and horizontal directions, respectively, while preserving the edge texture information of the image. The experimental results show that the proposed method outperforms four state-of-the-art methods by a large margin in terms of the qualitative and quantitative assessments. One hundred infrared images with different simulated noise intensities are applied to verify the performance of our method, and the result shows that the average peak signal-to-noise ratio and average structural similarity of the corrected image exceed 40.08 dB and 0.98, respectively. [ABSTRACT FROM AUTHOR]

Hypergraph matching has been attractive in the application of computer vision in recent years. The interference of external factors, such as squeezing, pulling, occlusion, and noise, results in the same target displaying different image characteristics under different influencing factors. After extracting the image feature point description, the traditional method directly measures the feature description using distance measurement methods such as Euclidean distance, cosine distance, and Manhattan distance, which lack a sufficient generalization ability and negatively impact the accuracy and effectiveness of matching. This paper proposes a metric-learning-based hypergraph matching (MLGM) approach that employs metric learning to express the similarity relationship between high-order image descriptors and learns a new metric function based on scene requirements and target characteristics. The experimental results show that our proposed method performs better than state-of-the-art algorithms on both synthetic and natural images. [ABSTRACT FROM AUTHOR]

The nature theory of Chinese herbal medicines (CHMs) is the core theory of traditional Chinese medicine (TCM). Cold-hot nature is an important part of CHM nature. It is found that the material basis of cold-hot nature is CHM ingredients. To test the scientific hypothesis that “CHMs with similar cold-hot nature should have similar material basis,” we explored an intelligent method for cold-hot nature identification of CHMs based on the feature similarity of CHM ingredients in this work. Sixty one CHMs were selected for cold-hot nature identification. High performance liquid chromatography (HPLC) was used to separate the chemical ingredients of CHMs and extract the feature information of CHM ingredients. A distance metric learning algorithm was then learned to measure the similarity of HPLC fingerprints. With the learned distance metric, cold-hot nature identification scheme (CHNIS) was proposed to build an identification model to evaluate the cold-hot nature of CHMs. A number of experiments were designed to verify the effectiveness and feasibility of the proposed CHNIS model. The total identification accuracy rate of 61 CHMs is 80.3%. The performance of the proposed CHNIS algorithm outperformed that of the compared classical algorithms. The experimental results confirmed our inference that CHMs with similar cold-hot nature had similar composition of substances. The CHNIS model was proved to be effective and feasible.

Pedicle screw insertion is considered a complex surgery among Orthopaedics surgeons. Exclusively to prevent postoperative complications associated with pedicle screw insertion, various types of image intensity registration-based navigation systems have been developed. These systems are computation-intensive, have a small capture range and have local maxima issues. On the other hand, deep learning-based techniques lack registration generalizability and have data dependency. To overcome these limitations, a patient-specific hybrid 3D-2D registration principled framework was designed to map a pedicle screw trajectory between intraoperative X-ray image and preoperative CT image. An anatomical landmark-based 3D-2D Iterative Control Point (ICP) registration was performed to register a pedicular marker pose between the X-ray images and axial preoperative CT images. The registration framework was clinically validated by generating projection images possessing an optimal match with intraoperative X-ray images at the corresponding control point registration. The effectiveness of the registered trajectory was evaluated in terms of displacement and directional errors after reprojecting its position on 2D radiographic planes. The mean Euclidean distances for the Head and Tail end of the reprojected trajectory from the actual trajectory in the AP and lateral planes were shown to be 0.6–0.8 mm and 0.5–1.6 mm, respectively. Similarly, the corresponding mean directional errors were found to be 4.9 0 and 2 0 . The mean trajectory length difference between the actual and registered trajectory was shown to be 2.67 mm. The approximate time required in the intraoperative environment to axially map the marker position for a single vertebra was found to be 3 min. Utilizing the markerless registration techniques, the designed framework functions like a screw navigation tool, and assures the quality of surgery being performed by limiting the need of postoperative CT. [ABSTRACT FROM AUTHOR]

Unsupervised subsequence outlier detection on multivariate time series (MTS) is a valuable problem in practice that can observably save the cost of labeling and provide interpretability in real applications. For the task, most of the classic methods are under two strong assumptions: (i) stationary MTS. it may have difficulty coping with the phenomenon of time drift. (ii) Attribute-level IIDness (independent and identically distributed), it may ignore the relationship between attribute when measuring the similarity between multivariate subsequences. The above assumptions limit the availability of existing methods in real scenarios. To address this issue, this paper introduces a novel coupled generalized local outlier detection model for MTS, which extends the traditional generalized local outlier detection model to cope with subsequence outlier detection tasks by incorporating a novel Non-IID similarity metric. Specifically, the proposed method mainly includes three aspects: (i) represents the MTS relationship in symbolic space which provides a lower complexity and satisfactory sensitivity. (ii) Proposes a Non-IID coupled similarity metric (TSDis) which considers the intrinsic intra-attribute and inter-attribute coupling between segments. (iii) Extends the traditional generalized local outlier detection model to handle subsequence outlier detection tasks by embedding Non-IID coupled similarity metric. Experimental results show the proposed method can utilize the potential characteristics of MTS effectively and stably. Meanwhile, it detects outliers more accurately than baseline approaches on 12 time-series datasets. [ABSTRACT FROM AUTHOR]

Model-based Test Case Prioritization utilizing similarity metrics has proved effective in software testing. However, the utility of similarity metrics in it varies with test scenarios, hindering its universal effectiveness and performance optimization. To tackle this problem, we propose a Diversity-driven Learn-to-rank model-based TCP approach, named DLTCP, for optimizing early fault detection performance. Our method first employs the whale optimization algorithm to search for a suitable set of similarity metrics from a pool of existing candidates. This search process determines which metrics should be used. According to each selected metric, test cases are then prioritized. The resulting test case rankings are used as the training data for DLTCP. Finally, the proposed method incorporates random forest to train a ranking model for test case prioritization. As such, it can fuse multiple similarity metrics to improve the TCP performance. We conduct extensive experiments to evaluate our method's performance using the average percentage fault detected (APFD) as metric. The experimental results show that DLTCP achieve an average APFD value of 0.953 for seven classic benchmark models, which is 11.37% higher than that of the state-of-the-art algorithms. It can well select a set of similarity metrics for effective fusion, demonstrating competitive performance in early fault detection. • A novel approach, named DLTCP, optimizes model-based test case prioritization. • A fusion framework is proposed to combine multiple similarity metrics. • A Learn-to-rank model is constructed to improve early fault detection performance. • The superiority of DLTCP is verified on five classic benchmark models. [ABSTRACT FROM AUTHOR]

With the growth of various types of media data (e.g., text, image, video, and audio), fine-grained cross-media retrieval which aims to provide flexible and accurate query service has attracted significant attention. Different from traditional keyword-based retrieval, the queries and results in fine-grained cross-media retrieval might be different types. In this work, we demonstrate that the quality of retrieval results can be further improved by additionally considering the media-specified information. Besides, the process of feature extraction should be different for queries with various media types. To this end, we propose a novel network architecture, namely Double Branch Fine-grained Cross-media Net (DBFC-Net), which is the first work that can use the media-specific information to construct the common features by a uniform framework. Furthermore, we devise an effective distance metric (cosine+) for fine-grained cross-media retrieval. Compared with commonly-used metrics (e.g., cosine function), our proposed cosine+ metric is well adaptive to handle fine-grained retrieval scenarios. Extensive experiments and ablation studies on publicly available datasets demonstrate the effectiveness of our proposed approach. [ABSTRACT FROM AUTHOR]

An in‐house hybrid deformable image registration (DIR) method, which combines free‐form deformation (FFD) and the viscous fluid registration method, is proposed. Its results on the planning computed tomography (CT) and the day 1 treatment cone‐beam CT (CBCT) image from 68 head and neck cancer patients are compared with the results of NiftyReg, which uses B‐spline FFD alone. Several similarity metrics, the target registration error (TRE) of annotated points, as well as the Dice similarity coefficient (DSC) and Hausdorff distance (HD) of the propagated organs at risk are employed to analyze their registration accuracy. According to quantitative analysis on mutual information, normalized cross‐correlation, and the absolute pixel value differences, the results of the proposed DIR are more similar to the CBCT images than the NiftyReg results. Smaller TRE of the annotated points is observed in the proposed method, and the overall mean TRE for the proposed method and NiftyReg was 2.34 and 2.98 mm, respectively (p < 0.001). The mean DSC in the larynx, spinal cord, oral cavity, mandible, and parotid given by the proposed method ranged from 0.78 to 0.91, significantly higher than the NiftyReg results (ranging from 0.77 to 0.90), and the HD was significantly lower compared to NiftyReg. Furthermore, the proposed method did not suffer from unrealistic deformations as the NiftyReg did in the visual evaluation. Meanwhile, the execution time of the proposed method was much higher than NiftyReg (96.98 ± 11.88 s vs. 4.60 ± 0.49 s). In conclusion, the in‐house hybrid method gave better accuracy and more stable performance than NiftyReg. [ABSTRACT FROM AUTHOR]

To meet user's functional requirements, cloud computing and big data have become the most commonly used computing and data resources. Based on analysis, conversion, extraction and refinement for the big data, a disease can be prevented and group behavior can be predicted. However, each user's private data is also an element in big data. Users must provide private data to the service providers to meet their functional requirements. To gain economic benefits, some SaaS service providers have not been authorized to collect and analyze the user's sensitive private data, as a result, the user's private data is disclosed. In this paper, we propose a private data chain disclosure discovery method, to prevent a user's sensitive privacy information from being illegally disclosed. First, we measure the similarity degree and cost of the disclosure of the private data. Second, according to the similarity degree and cost of disclosure, the disclosure chain and key private data are detected in the process of interaction between user and SaaS service. Third, we propose a discovery framework for the private data chain and demonstrate its feasibility and effectiveness by experiments. [ABSTRACT FROM AUTHOR]

In this paper, given data with high-dimensional features, we study this problem of how to calculate the similarity between two samples by considering feature interaction network, where a feature interaction network represents the relationship between features. This is different from some traditional methods, those of which learn similarities based on a sample network that represents the relationship between samples. Therefore, we propose a novel network-based similarity metric for computing the similarity between samples, which incorporates the knowledge of feature interaction network, in order to overcome the data sparseness problem. Our similarity metric uses a new Feature Alignment Similarity measure, which does not directly compute the similarities among samples, but projects each sample into a feature interaction network and measures the similarities between two samples using the similarities between the vertices of the samples in the network. As such, when two samples do not share any common features, they are likely to have higher similarity values when their features share the similar network regions. For ensuring that the metric is useful in a real-world application, we apply our metric to discover subtypes in tumor mutational data by incorporating the information of the gene interaction network. Our experimental results from using synthetic data and real-world tumor mutational data show that our approach outperforms the top competitors in cancer subtype discovery. Furthermore, our approach can identify cancer subtypes that cannot be detected by other clustering algorithms in real cancer data.

The non-uniformity of the readout circuit response in the infrared focal plane array unit detector can result in fixed pattern noise with stripe, which seriously affects the quality of the infrared images. Considering the problems of existing non-uniformity correction, such as the loss of image detail and edge blurring, a multi-scale residual network with attention mechanism is proposed for single infrared image stripe noise removal. A multi-scale feature representation module is designed to decompose the original image into varying scales to obtain more image information. The product of the direction structure similarity parameter and the Gaussian weighted Mahalanobis distance is used as the similarity metric; a channel spatial attention mechanism based on similarity (CSAS) ensures the extraction of a more discriminative channel and spatial feature. The method is employed to eliminate the stripe noise in the vertical and horizontal directions, respectively, while preserving the edge texture information of the image. The experimental results show that the proposed method outperforms four state-of-the-art methods by a large margin in terms of the qualitative and quantitative assessments. One hundred infrared images with different simulated noise intensities are applied to verify the performance of our method, and the result shows that the average peak signal-to-noise ratio and average structural similarity of the corrected image exceed 40.08 dB and 0.98, respectively.

Palmprint recognition is one of the effective biometric technologies due to the advantages of convenience and safety. Recently, many deep learning-based methods are utilized in palmprint recognition and achieve satisfactory results. However, most of the existing learning methods are driven by the abundant labeled data. When the training data are insufficient, their performance drop sharply. In this work, we propose a novel and effective end-to-end algorithm for few-shot palmprint recognition, called Similarity Metric Hashing Network (SMHNet). SMHNet is designed to extract the features of palmprint images on both the structural and pixel levels. Specifically, an embedded structural similarity (SSIM) index block is constructed behind the last convolution layer to measure the structural similarity between query samples and support ones. A novel SSIM loss is designed with distance loss to train the entire model from scratch. Furthermore, a hashing block is added after the last fully connected (FC) layer to encode the features into hashing codes, which is convenient for large-scale feature storage and retrieval. Extensive experiments are conducted on three benchmark palmprint databases, and the results demonstrate that our model can achieve competitive accuracy compared with several state-of-the-art models. [ABSTRACT FROM AUTHOR]

Hypergraph matching has been attractive in the application of computer vision in recent years. The interference of external factors, such as squeezing, pulling, occlusion, and noise, results in the same target displaying different image characteristics under different influencing factors. After extracting the image feature point description, the traditional method directly measures the feature description using distance measurement methods such as Euclidean distance, cosine distance, and Manhattan distance, which lack a sufficient generalization ability and negatively impact the accuracy and effectiveness of matching. This paper proposes a metric-learning-based hypergraph matching (MLGM) approach that employs metric learning to express the similarity relationship between high-order image descriptors and learns a new metric function based on scene requirements and target characteristics. The experimental results show that our proposed method performs better than state-of-the-art algorithms on both synthetic and natural images.

Link prediction is an axial field in complex network analysis as it aims to infer new connections between nodes in a given network. Many applications of this task include: the reference system, the suggestion of friends in the social network and the prediction of interactions between proteins in biological networks. Several methods have been developed in the link prediction task, especially the similarity-based methods which are widely used due to their low complexity and good performance. In this paper, we propose a novel link prediction approach that combines the PageRank algorithm with local information-based methods to improve performance while retaining the advantage of low complexity of local methods. We conducted a series of experimental studies on eleven data sets where we compared our new combined methods with six well-known local methods. The results obtained show a significant gain in terms of performance in almost all data sets. In addition to this and to confirm the superiority of the proposed methods, another comparative study is performed, formed of nine local and global methods. According to the experimental results, our approach outperforms all other compared methods with linear complexity. • The analysis of networks evolution brings us to the problem of link prediction. • Similarity-based methods are used as a local method to predict new links. • New similarity measures between unconnected nodes that do not have shared neighbors. • Our method offers higher precision than some well-known local and global methods. • Our method offers linear computational complexity. [ABSTRACT FROM AUTHOR]

Although mutual information is a general method usually being used to measure the similarity of two images, the robustness is questionable due to the absence of spatial information. The purpose of this study is to develop a feasible sorting technique for 4D-CT. A novel sorting algorithm named mutual information and edge gradient (MIEG), which includes spatial information by combining mutual information with a term based on the edge gradient of the image, was proposed to sort sequential CT images. The edge of image was extracted by calculating the wavelet transform modulus maxima, and the gradient similarity coefficient of the edge image was calculated and used to multiply by mutual information to form the final similarity metric. This sorting technique was validated by comparing the 4D-CTs reconstructed using MIEG and Real-time Position Management system (Varian Medical Systems, Inc., Palo Alto, CA). Tumor motion trajectories derived from 4D-CTs were analyzed in three orthogonal directions. Their correlation coefficients (CC) and differences in tumor motion magnitude (Ds) were determined. In addition, Dice similarity coefficient (DSC) was used to measure how well the tumor volumes segmented from the two 4D datasets overlapped with each other. For all patients, the mean CC values were >0.95 in all directions. The mean Ds were

Automated change-point detection of EEG signals is becoming essential for the monitoring of health behaviors and health status in a wide range of clinical applications. This paper presents a structural time-series analysis to capture and characterize the dynamic behavior of EEG signals, and develops a method to detect the EEG change points. For a given EEG signal, the proposed method is operated as follows: 1) a sub-band pass filter is fist designed to capture those frequency components that can characterize the dynamic behavior of the data, and the so-called power spectrum is extracted as the EEG features; 2) together with a sliding-window technique, an automatic `segment-to-segment' analysis of EEG signal, is developed with a null hypothesis testing for decision making. In particular, the main challenge of the proposed method is to design an appropriate distance metric that is compatible with our considered data/problem. To achieve this end, we first collect a variety of metrics from other areas that would be potentially available for our problem, and then compare them for the considered EEG change point detection. Experiments are conducted on two different data sets. Results show the Bhattacharyya distance achieves the best detection result among all investigated metrics. Meanwhile, comparison with state-of-the-arts demonstrates the effectiveness of the method in real applications.

Searching genomic interval sets produced by sequencing methods has been widely and routinely performed; however, existing metrics for quantifying similarities among interval sets are inconsistent. Here we introduce Seqpare, a self-consistent and effective metric of similarity and tool for comparing sequences based on their interval sets. With this metric, the similarity of two interval sets is quantified by a single index, the ratio of their effective overlap over the union: an index of zero indicates unrelated interval sets, and an index of one means that the interval sets are identical. Analysis and tests confirm the effectiveness and self-consistency of the Seqpare metric.