Average Cosine Similarity Research Articles

Підхід до виявлення аномалій в потоках тектових даних

Збільшення потоків інформації веде за собою необхідність розробки спеціалізованих інтелектуальних засобів та методів для обробки величезної кількості даних. Так популярність соціальних мереж, різного роду месенджерів вимагає створення спеціалізованих засобів для обробки потоку текстових повідомлень. Дане дослідження присвячене дослідженню та розробці методу виявлення аномальних елементів в потоках текстових даних. Особлива увага присвячена підтримці україномовних текстів.

A Study on the Synthetic ECG Generation for User Recognition

심전도 신호는 시간 및 환경 변화에 따라 측정되는 시계열 데이터로 매번 등록 데이터와 동일한 크기의 비교 데이터를 취득해야 하는 문제점이 발생한다. 본 논문에서는 신호 크기 부적합 문제를 해결하기 위해 가상 생체신호 생성을 위한 보조 분류기 기반 적대적 생성 신경망(Auxiliary Classifier Generative Adversarial Networks)의 네트워크 모델을 제안한다. 생성된 가상생체신호의 유사성을 확인하기 위해 코사인 각도와 교차 상관관계를 이용하였다. 실험 결과, 코사인 유사도 측정 결과로 평균 유사도는 0.991의 결과를 나타냈으며, 교차 상관관계를 이용한 유클리디언 거리 기반 유사성 측정 결과는 평균 0.25 유사도 결과를 나타냈다. 이는 등록 데이터와 실험 데이터간의 크기가 일치하지 않더라도 가상 생체신호 생성을 통해 신호 크기 부적합 문제를 해결함을 확인하였다.Because the ECG signals are time-series data acquired as time elapses, it is important to obtain comparative data the same in size as the enrolled data every time. This paper suggests a network model of GAN (Generative Adversarial Networks) based on an auxiliary classifier to generate synthetic ECG signals which may address the different data size issues. The Cosine similarity and Cross-correlation are used to examine the similarity of synthetic ECG signals. The analysis shows that the Average Cosine similarity was 0.991 and the Average Euclidean distance similarity based on cross-correlation was 0.25: such results indicate that data size difference issue can be resolved while the generated synthetic ECG signals, similar to real ECG signals, can create synthetic data even when the registered data are not the same as the comparative data in size.

Pulmonary nodule segmentation with CT sample synthesis using adversarial networks.

Segmentation of pulmonary nodules is critical for the analysis of nodules and lung cancer diagnosis. We present a novel framework of segmentation for various types of nodules using convolutional neural networks (CNNs). The proposed framework is composed of two major parts. The first part is to increase the variety of samples and build a more balanced dataset. A conditional generative adversarial network (cGAN) is employed to produce synthetic CT images. Semantic labels are generated to impart spatial contextual knowledge to the network. Nine attribute scoring labels are combined as well to preserve nodule features. To refine the realism of synthesized samples, reconstruction error loss is introduced into cGAN. The second part is to train a nodule segmentation network on the extended dataset. We build a three-dimensional (3D) CNN model that exploits heterogeneous maps including edge maps and local binary pattern maps. The incorporation of these maps informs the model of texture patterns and boundary information of nodules, which assists high-level feature learning for segmentation. Residual unit, which learns to reduce residual error, is adopted to accelerate training and improve accuracy. Validation on LIDC-IDRI dataset demonstrates that the generated samples are realistic. The mean squared error and average cosine similarity between real and synthesized samples are and 0.9534, respectively. The Dice coefficient, positive predicted value, sensitivity, and accuracy are, respectively, 0.8483, 0.8895, 0.8511, and 0.9904 for the segmentation results. The proposed 3D CNN segmentation framework, based on the use of synthesized samples and multiple maps with residual learning, achieves more accurate nodule segmentation compared to existing state-of-the-art methods. The proposed CT image synthesis method can not only output samples close to real images but also allow for stochastic variation in image diversity.

Feasibility of Diagnosing Both Severity and Features of Diabetic Retinopathy in Fundus Photography

Diabetic retinopathy (DR) diagnosis methods in the literature are usually criticized as being limit in diagnosing DR-related features or being lack of interpretability. To deal with these issues, this paper investigates the feasibility of diagnosing both DR severity levels and the presence of DR-related features in a two-step procedure. Specifically, this paper first analyzes the quality of annotations in DR grading by measuring inter-grader variability. Cosine similarity is considered to evaluate the inter-grader variability of the presence of DR-related features, and quadratic weighted Cohen's kappa is employed to assess the inter-grader variability of DR severity levels. Next, different annotation methods as follows are compared to DR severity prediction performance using logistic regression: 1) single annotations by single grader (SASG); 2) single annotations from multiple graders (SAMG); 3) multiple annotations by voting (MAV); and 4) double annotations with adjudication of disagreement (DAAD). Based on the comparison results, the feasibility of diagnosing both DR severity and features is investigated. In the experiments, 1589 fundus images graded by three retinal specialists and four general ophthalmologists are considered. The results demonstrate that retinal specialists are more consistent than general ophthalmologists in grading both the presence of DR-related features and DR severity. The SASG and MAV should be avoided if possible while the DAAD is the good option when prediction performance is the highest priority and the SAMG is especially beneficial when both prediction performance and grading costs are considered. The upper limit performance of DR severity prediction gets accuracy 95.6% and kappa 0.962. When DR-related feature prediction achieves average cosine similarity 0.823, it is potential to get accuracy 91.2% and kappa 0.905 for DR severity prediction in real applications. These results together suggest the potential of diagnosis of both DR severity and the presence of DR-related features in a two-step procedure.

A text-Image feature mapping algorithm based on transfer learning

Abstract The traditional uniform distribution algorithm does not filter the image data when extracting the approximate features of text-image data under the event, so the similarity between the image data and the text is low, which leads to low accuracy of the algorithm. This paper proposes a text-image feature mapping algorithm based on transfer learning. The existing data is filtered by ‘clustering technology’ to obtain similar data with the target data. The significant text features are calculated through the latent Dirichlet allocation (LDA) model and information gain based on Gibbs sampling. Bag of visual word (BOVW) model and Naive Bayesian method are used to model image data. With the help of the text-image co-occurrence data in the same event, the text feature distribution is mapped to the image feature space, and the feature distribution of image data under the same event is approximated. Experimental results show that the proposed algorithm can obtain the feature distribution of image data under different events, and the average cosine similarity is as high as 92%, the average dispersion is as low as 0.06%, and the accuracy of the algorithm is high.

Repeatability of electromyography recordings and muscle synergies during gait among children with cerebral palsy

BackgroundClinical gait analysis is commonly used in the evaluation and treatment of children with cerebral palsy (CP). While the repeatability of kinematic and kinetic measures of gait has previously been evaluated, the repeatability of electromyography (EMG) recordings or measures calculated from EMG data, such as muscle synergies, remains unclear for this population. Research questionAre EMG recordings and muscle synergies from clinical gait analysis repeatable between visits for children with CP? MethodsWe recruited 20 children with bilateral CP who had been referred for clinical gait analysis. The children completed two visits less than six weeks apart with EMG data collected bilaterally from five muscles (rectus femoris, medial hamstrings, vastus lateralis, anterior tibialis, and medial gastrocnemius). Variance ratio and cosine similarity were used to evaluate repeatability of EMG waveforms between visits. Nonnegative matrix factorization was used to calculate synergies from EMG data at each visit to compare synergy weights and activations. Results & significanceThe inter-visit variance ratios of EMG data for children with CP were similar to previously reported results for typically-developing children and unimpaired adults (range: 0.39 for vastus lateralis to 0.66 for rectus femoris). The average cosine similarity of the EMG waveforms between visits was greater than 0.9 for all muscles, while synergy weights and activations also had high similarity – greater than 0.8 and 0.9 between visits, respectively. These results demonstrate that EMG repeatability between visits during clinical gait analysis for children with CP is similar to unimpaired individuals. These results provide a baseline for evaluating whether observed changes in EMG recordings between visits reflect real changes in muscle activity or are within the range of inter-visit variability.

Predicting Mission Alignment and Preventing Mission Drift: Do Revenue Sources Matter?

Activities of nonprofit organizations do not always align with their missions, a managerial problem termed as “mission drift.” Mission drift is difficult to operationalize and quantify; thus, as a critical issue, only a few conceptual pieces or empirical case studies have explored this topic. This paper develops innovative measures to operationalize “mission alignment” using data science methodology, and examines the impact of revenue sources on mission alignment. By using the cosine similarity of text between a mission statement and program description, four measures of mission alignment are devised: the sum cosine similarity, average cosine similarity, weighted sum cosine similarity, and weighted average cosine similarity. Text analysis indicates that a majority of the programs evidence educational purposes, and for-profit business plays an important role in foundations’ projects and funding. The regression analysis shows that personal donation and service revenue can increase mission alignment, while organizational donation and membership dues decrease mission alignment.

A Novel Bio-Inspired Algorithm Based on Social Spiders for Improving Performance and Efficiency of Data Clustering

Abstract Since the last decade, the collective intelligent behavior of groups of animals, birds or insects have attracted the attention of researchers. Swarm intelligence is the branch of artificial intelligence that deals with the implementation of intelligent systems by taking inspiration from the collective behavior of social insects and other societies of animals. Many meta-heuristic algorithms based on aggregative conduct of swarms through complex interactions with no supervision have been used to solve complex optimization problems. Data clustering organizes data into groups called clusters, such that each cluster has similar data. It also produces clusters that could be disjoint. Accuracy and efficiency are the important measures in data clustering. Several recent studies describe bio-inspired systems as information processing systems capable of some cognitive ability. However, existing popular bio-inspired algorithms for data clustering ignored good balance between exploration and exploitation for producing better clustering results. In this article, we propose a bio-inspired algorithm, namely social spider optimization (SSO), for clustering that maintains a good balance between exploration and exploitation using female and male spiders, respectively. We compare results of the proposed algorithm SSO with K means and other nature-inspired algorithms such as particle swarm optimization (PSO), ant colony optimization (ACO) and improved bee colony optimization (IBCO). We find it to be more robust as it produces better clustering results. Although SSO solves the problem of getting stuck in the local optimum, it needs to be modified for locating the best solution in the proximity of the generated global solution. Hence, we hybridize SSO with K means, which produces good results in local searches. We compare proposed hybrid algorithms SSO+K means (SSOKC), integrated SSOKC (ISSOKC), and interleaved SSOKC (ILSSOKC) with K means+PSO (KPSO), K means+genetic algorithm (KGA), K means+artificial bee colony (KABC) and interleaved K means+IBCO (IKIBCO) and find better clustering results. We use sum of intra-cluster distances (SICD), average cosine similarity, accuracy and inter-cluster distance to measure and validate the performance and efficiency of the proposed clustering techniques.

Intra-Subject Consistency during Locomotion: Similarity in Shared and Subject-Specific Muscle Synergies.

Human locomotion is a complex motor task. Previous research hypothesized that muscle synergies reflect the modular control of muscle groups operated by the Central Nervous System (CNS). Despite the high stride-to-stride variability characterizing human gait, most studies analyze only a few strides. This may be limiting, because the intra-subject variability of motor output is neglected. This gap could be filled by recording and analyzing many gait cycles during a single walking task. In this way, it can be investigated if CNS recruits the same muscle synergies consistently or if different strategies are adopted during the locomotion task. The aim of this work is to investigate the intra-subject consistency of muscle synergies during overground walking. Twelve young healthy volunteers were instructed to walk for 5 min at their natural pace. On the average, 181 ± 10 gait cycles were analyzed for each subject. Surface electromyography was recorded from 12 muscles of the dominant lower limb and the trunk. Gait cycles were grouped into subgroups containing 10 gait cycles each. The consistency of the muscle synergies extracted during the gait trial was assessed by measuring cosine similarity (CS) of muscle weights vectors, and zero-lag cross-correlation (CC) of activation signals. The average intra-subject CS and CC were 0.94 ± 0.10 and 0.96 ± 0.06, respectively. We found five synergies shared by all the subjects: high consistency values were found for these synergies (CS = 0.96 ± 0.05, CC = 0.97 ± 0.03). In addition, we found 10 subject-specific synergies. These synergies were less consistent (CS = 0.80 ± 0.20, CC = 0.89 ± 0.14). In conclusion, our results demonstrated that shared muscle synergies were highly consistent during walking. Subject-specific muscle synergies were also consistent, although to a lesser extent.

Predicting Mission Alignment and Preventing Mission Drift: How Revenue Sources Matter?

Activities of nonprofit organizations do not always align with their missions, a managerial problem termed as “mission drift.” Mission drift is difficult to operationalize and quantify thus, as a critical issue, only a few conceptual pieces or empirical case studies explored this topic. This paper develops innovative measures to operationalize “mission alignment” using data science methodology and examines how different revenue sources influence the mission alignment of Chinese foundations. Specifically, four measures of mission alignment are devised based on the cosine similarity of text between mission statement and program description (i.e., the sum cosine similarity, average cosine similarity, weighted sum cosine similarity, and weighted average cosine similarity). The text analysis of programs indicates that the majority of the foundations’ projects have educational purposes and for-profit businesses play a significant role in foundations’ projects and funding. The regression analysis shows that personal donation and service revenue can increase mission alignment while organizational donation and membership dues decrease mission alignment. The results suggest validity of the mission alignment measures.

Building bridges across electronic health record systems through inferred phenotypic topics

ObjectiveData in electronic health records (EHRs) is being increasingly leveraged for secondary uses, ranging from biomedical association studies to comparative effectiveness. To perform studies at scale and transfer knowledge from one institution to another in a meaningful way, we need to harmonize the phenotypes in such systems. Traditionally, this has been accomplished through expert specification of phenotypes via standardized terminologies, such as billing codes. However, this approach may be biased by the experience and expectations of the experts, as well as the vocabulary used to describe such patients. The goal of this work is to develop a data-driven strategy to (1) infer phenotypic topics within patient populations and (2) assess the degree to which such topics facilitate a mapping across populations in disparate healthcare systems. MethodsWe adapt a generative topic modeling strategy, based on latent Dirichlet allocation, to infer phenotypic topics. We utilize a variance analysis to assess the projection of a patient population from one healthcare system onto the topics learned from another system. The consistency of learned phenotypic topics was evaluated using (1) the similarity of topics, (2) the stability of a patient population across topics, and (3) the transferability of a topic across sites. We evaluated our approaches using four months of inpatient data from two geographically distinct healthcare systems: (1) Northwestern Memorial Hospital (NMH) and (2) Vanderbilt University Medical Center (VUMC). ResultsThe method learned 25 phenotypic topics from each healthcare system. The average cosine similarity between matched topics across the two sites was 0.39, a remarkably high value given the very high dimensionality of the feature space. The average stability of VUMC and NMH patients across the topics of two sites was 0.988 and 0.812, respectively, as measured by the Pearson correlation coefficient. Also the VUMC and NMH topics have smaller variance of characterizing patient population of two sites than standard clinical terminologies (e.g., ICD9), suggesting they may be more reliably transferred across hospital systems. ConclusionsPhenotypic topics learned from EHR data can be more stable and transferable than billing codes for characterizing the general status of a patient population. This suggests that EHR-based research may be able to leverage such phenotypic topics as variables when pooling patient populations in predictive models.

Matching Medical Websites to Medical Guidelines through Clinical Vocabularies in the Context of Website Quality Assessment

Modern technology offers a wide array of possibilities to publish almost any content freely on the Internet. Because of the importance and delicacy of medical information, the quality of such texts provided to general public seems to be a serious issue nowadays. Unfortunately, the only feasible way to approve the adequacy of the medical information content is human verification. Best practices in medicine are systematically captured by medical guidelines (MGL), which are provided by renowned medical societies and based on results of Evidence-Based Medicine (EBM). We propose a simple approach exploiting MGL content as a benchmark for the assessment of a content quality in medical web sites (WS). It is based on the idea that the information content or at least the scope of a medical text is reflected in the domain terminology used. We discuss a possible use of this approach in semiautomatic human-based quality verification and various aspects related to its application. Concept candidates discovered in a MGL and in the tested web pages are matched to UMLS, yielding sets of used medical terms and corresponding concepts. Several aggregation techniques for MGLs were proposed and tested. The two sets are analyzed for overall similarity at term and concept level. The method was applied on a selected medical topic employing relevant MGL and 100 WS. All the analyzed web pages fell into five distinct categories (corresponding to the target audience). Aggregations for the MGLs were proposed and tested. The average cosine similarity to MGL across all tested WS reached 0.69 whereas the average similarity calculated per each category varied up to 7,6% against the overall number. The research done is the first step towards automated evaluation of a medical web page content on the basis of MGLs as the quality standard. We describe further tasks which would improve the outputs of comparison and the possibility of its common application.