Nearest-neighbor Method Research Articles

Optimizing the Number of Clusters for Billion-Scale Quantization-Based Nearest Neighbor Search

Optimizing the Number of Clusters for Billion-Scale Quantization-Based Nearest Neighbor Search

An efficient infinity norm minimization algorithm for under-determined inverse problems

The problem of solving under-determined systems of linear equations with minimum peak magnitude (ℓ∞ norm) has numerous applications in signal processing. These include Peak-to-Average Power Ratio (PAPR) reduction in MIMO-OFDM systems, vector quantization, approximate nearest neighbor search, optimal control in robotics, and power grid optimization. Several methods have been proposed to address this problem, but they often face limitations in computational speed or representation accuracy. Some methods also impose constraints on the frame matrix, such as restrictions on the type of its entries or its aspect ratio. In this paper, we present the Fast Iterative Peak Shrinkage Algorithm (FIPSA), which iterates over feasible solutions to consistently reduce peak magnitude and provably converge to near-optimal solutions. Our experimental results, conducted across various frame matrix types and aspect ratios, demonstrate that FIPSA consistently achieves near-minimal ℓ∞ norm values. In addition, it operates 1.3 to 7.3 times faster than previous methods, while maintaining an average representation error of 10−15. Notably, these advancements are achieved without imposing any constraints on the frame matrix.

DPC clustering algorithm based on K-nearest neighbors and kernel density estimation

Clustering by fast search and find of density peaks (DPC) is a density-based clustering algorithm. This algorithm does not require iteration and too many parameter settings, but it cannot identify cluster centers with small cluster density because the local structure of the data is not considered when calculating the local density. To address the above problems, a DPC clustering algorithm (KKDPC) based on K-reciprocal Neighbors (KN) and Kernel Density Estimation (KDE) is proposed. First, the number of reciprocal neighbors and local kernel density of data points are obtained by nearest neighbor and kernel density estimation methods ; second, the number of K reciprocal neighbors is added to the local kernel density to obtain a new local density; finally, the relative distance is obtained according to the local density of the data points, and the cluster center is selected and the non-center point is allocated by constructing a decision graph. Experiments are conducted using artificial and real data sets, and compared with seven clustering algorithms: DPC, DBSCAN, K-means, FCM, DPC-KNN, DPC-NN, and DPC-FWSN. The performance of the KKDPC clustering algorithm is verified by calculating the adjusted mutual information (AMI), adjusted Rand index (ARI), and normalized mutual information (NMI). The experimental results show that this method can more accurately identify the cluster center and effectively improve the clustering accuracy.

A novel ranked k-nearest neighbors algorithm for missing data imputation

Missing data is a common problem in many domains that rely on data analysis. The k Nearest Neighbors imputation method has been widely used to address this issue, but it has limitations in accurately imputing missing values, especially for datasets with small pairwise correlations and small values of k. In this study, we proposed a method, Ranked k Nearest Neighbors imputation that uses a similar approach to k Nearest Neighbor, but utilizing the concept of Ranked set sampling to select the most relevant neighbors for imputation. Our results show that the proposed method outperforms the standard k nearest neighbor method in terms of imputation accuracy both in case of Missing Completely at Random and Missing at Random mechanism, as demonstrated by consistently lower MSIE and MAIE values across all datasets. This suggests that the proposed method is a promising alternative for imputing missing values in datasets with small pairwise correlations and small values of k. Thus, the proposed Ranked k Nearest Neighbor method has important implications for data imputation in various domains and can contribute to the development of more efficient and accurate imputation methods without adding any computational complexity to an algorithm.

Plantar Load System Analysis Using FSR Sensors and Interpolation Methods

The foot is considered a wonder of biological engineering due to its structure, formed by bones, ligaments, and tendons that collaborate to ensure stability and mobility. A key area often examined by medical professionals in patients with diabetic feet is the plantar surface, due to the risk of ulcer development. If left untreated, these ulcers can lead to severe complications, including amputation of the toe, foot, or even the limb. Interpolation methods are used to find areas with overloads in a system of sensor maps that are based on capacitive, load cells, or force-sensitive resistors (FSRs). This manuscript presents the assessment of linear, nearest neighbors, and bicubic methods in comparison with ground truth to calculate the root mean square error (RMSE) in two assessments using a dataset of eight healthy subjects, four men and four women, with an average age of 25 years, height of 1.63 m, and weight of 72 kg with shoe sizes from 7.3 USA using FSR map with 48 sensors. Additionally, this paper describes the conditioning circuit development to implement a plantar surface system that enables interpolating loads on the plantar surface. The proposed system’s results show that the first assessment indicates an RMSE of 0.089, 0.126, and 0.089 for linear, nearest neighbor, and bicubic methods, while the second assessment shows a mean RMSE for linear, nearest neighbor, and bicubic methods of 0.114, 0.159, and 0.112.

Influence of High-Altitude Residential History on Optimal HbA1c Cutoff for Detecting Abnormal Glucose Metabolism.

Aims: To explore the influence of recent high-altitude residential history on the optimal cutoff of glycosylated hemoglobin (HbA1c) for detecting abnormal glucose metabolism. Methods: The study included 505 self-reported healthy Han participants of age 18-65 years, recruited in Chengdu and categorized based on recent (within 3 months) high-altitude (>2,500 m) residential history. The 1999 WHO criteria was used as the gold standard for defining prediabetes and diabetes. HbA1c test performance was assessed using receiver operating characteristic curve, with the optimal cutoff determined by Maximum Youden index. Propensity score matching with 0.02 calipers and nearest neighbor method was used to balance confounding factors between groups. Results: Of the participants, 238 (47.13%) were populations with recent high-altitude residential history (HA group), and 267 (52.87%) were low-altitude dwellers (LA group). The HA group had slightly higher HbA1c levels (p > 0.05) and higher erythrocyte and hemoglobin levels (p < 0.05), compared to the LA group. Weak correlations between prediabetes and HbA1c levels were observed in the HA group (rs = 0.21, p < 0.05) and the LA group (rs = 0.07, p = 0.25). The optimal cutoff for the detection of diabetes was 6.5% (area under the curve [AUC] 0.94) in the HA group and 5.9% (AUC 0.97) in the LA group, which remained unchanged after adjustment for confounders. Conclusions: The optimal cutoff of HbA1c for the detection of diabetes in populations with recent history of living at high altitude was higher than that in general populations living at low altitude, and the diagnostic value of HbA1c for prediabetes was also inadequate.

Application of Machine Learning Methods to Analyze the Dependence of Vibration Acceleration Signals on the Tightening Forces of Bolted Connections

The article is devoted to the application of machine learning methods for the analysis of vibration acceleration signals in bolted joints that occur under impact. The relevance of the work is due to the need to ensure the reliability and operability of bolted joints under loads. One of the promising areas of non-destructive testing is the analysis of the characteristics of vibration acceleration signals that change with changes in the state of structures, for example, with an increase in the tightening torque. This allows for the timely detection of possible defects and the prevention of emergency situations, ensuring the safety and durability of structures. The vibration acceleration signals under impact action on a bolted joint were analyzed. After calculating the signal characteristics, data sets were formed, including the values of tightening torques and the corresponding calculated signal characteristics, a search for a correlation between the tightening torque of bolted joints and the obtained set of vibration signal parameters was carried out. The frequency spectrum of signals, the Higuchi fractal dimension, detrended fluctuations, spectral power density and position of signal peaks were calculated. Machine learning models such as decision trees, the nearest neighbor method, the k-means method and neural networks were used for the analysis. For these methods, an optimal set of parameters correlating with the tightening torque was searched for. The main results show that machine learning methods are effective in classifying signals and finding correlations with stress state parameters. They allow us to detect the relationship between a set of signal characteristics and tightening torque, which opens up opportunities for more accurate and reliable monitoring of the condition of bolted connections. This should help to increase their operational reliability and durability, as well as reduce the likelihood of failures and accidents. The use of such methods can improve the quality of monitoring and diagnostics of bolted connections.

A Novel Energy-Aware Path Planning by Autonomous Underwater Vehicle in Underwater Wireless Sensor Networks

Wireless sensor networks can monitor the environment to detect anomalies and reduce the risk of maritime traffic. Energy is necessary for low-power conditions where wireless sensor networks are used. Ensuring the lifespan of energy constraints and providing continuous environmental observation, data collecting, and communication requires management. Battery replacement and energy consumption issues can be resolved with path planning and energy-efficient autonomous underwater vehicle charging for sensor nodes. The nearest neighbour technique is used in this study to solve the energy-aware path planning problem of an autonomous underwater vehicle. Path planning simulations show that the nearest neighbour strategy converges faster and produces a better result than the genetic algorithm. We develop robust and energy-efficient path-planning algorithms that efficiently acquire sensor data while consuming less energy, allowing the monitoring system to respond to anomalies more rapidly. Increased sensor connectivity lowers energy usage and increases network longevity. This study also considers the situation when it is recommended to avoid taking direct travel paths between particular node pairs for a variety of reasons. This recommendation is considered in this study. We present a strategy based on a modified Nearest Neighbour-based Approach from the Nearest Neighbour method to address this more challenging scenario. The direct pathways between such nodes are constrained within the context of this technique. The modified version of Nearest Neighbor-based approach performs well even in that particular situation.

Efficient Distance Queries on Non-point Data

Distance queries, including distance-range queries, k -nearest neighbors search, and distance joins, are very popular in spatial databases. However, they have been studied mainly for point data. Inspired by a recent approach on indexing non-point objects for rectangular range queries, we propose a secondary partitioning approach for space-partitioning indices, which is appropriate for distance queries. Our approach classifies the contents of each primary partition into 16 secondary partitions, taking into consideration the begin and end values of objects with respect to the spatial extent of the primary partition. Based on this, we define algorithms for three popular spatial query types, that avoid duplicate results and skip unnecessary computations. We compare our approach to the previous secondary partitioning method and to state-of-the-art data-partitioning indexing and find that it has a significant performance advantage.

Active Learning for Deep Visual Tracking.

Convolutional neural networks (CNNs) have been successfully applied to the single target tracking task in recent years. Generally, training a deep CNN model requires numerous labeled training samples, and the number and quality of these samples directly affect the representational capability of the trained model. However, this approach is restrictive in practice, because manually labeling such a large number of training samples is time-consuming and prohibitively expensive. In this article, we propose an active learning method for deep visual tracking, which selects and annotates the unlabeled samples to train the deep CNN model. Under the guidance of active learning, the tracker based on the trained deep CNN model can achieve competitive tracking performance while reducing the labeling cost. More specifically, to ensure the diversity of selected samples, we propose an active learning method based on multiframe collaboration to select those training samples that should be and need to be annotated. Meanwhile, considering the representativeness of these selected samples, we adopt a nearest-neighbor discrimination method based on the average nearest-neighbor distance to screen isolated samples and low-quality samples. Therefore, the training samples' subset selected based on our method requires only a given budget to maintain the diversity and representativeness of the entire sample set. Furthermore, we adopt a Tversky loss to improve the bounding box estimation of our tracker, which can ensure that the tracker achieves more accurate target states. Extensive experimental results confirm that our active-learning-based tracker (ALT) achieves competitive tracking accuracy and speed compared with state-of-the-art trackers on the seven most challenging evaluation benchmarks. Project website: https://sites.google.com/view/altrack/.

In-memory search with learning to hash based on resistive memory for recommendation acceleration

Similarity search is essential in current artificial intelligence applications and widely utilized in various fields, such as recommender systems. However, the exponential growth of data poses significant challenges in search time and energy consumption on traditional digital hardware. Here, we propose a software-hardware co-optimization to address these challenges. On the software side, we employ a learning-to-hash method for vector encoding and achieve an approximate nearest neighbor search by calculating Hamming distance, thereby reducing computational complexity. On the hardware side, we leverage the resistance random-access memory crossbar array to implement the hash encoding process and the content-addressable memory with an in-memory computing paradigm to lower the energy consumption during searches. Simulations on the MovieLens dataset demonstrate that the implementation achieves comparable accuracy to software and reduces energy consumption by 30-fold compared to traditional digital systems. These results provide insight into the development of energy-efficient in-memory search systems for edge computing.

Settlement Patterns of the Population Use the Average Nearest Neighbor Method

This study utilizes the Average Nearest Neighbor (ANN) method to analyze settlement patterns based on population density data across 15 regencies and cities in Lampung Province, Indonesia. ANN is a spatial statistical technique used to determine spatial distribution patterns by calculating the average distance between points, the number of points, and the area of the region. The analysis involved computing the ANN index and z-scores for each regency and city. The results, as shown in the analysis, indicate that the ANN index values are below one, ranging from 0.114 to 0.769, and the z-scores are negative, ranging from -146.719 to -13.235. These findings suggest that the settlement patterns across all studied areas in Lampung Province exhibit a clustered pattern. This implies that densely populated areas do not significantly influence the settlement patterns of less densely populated regions. The study concludes that the population density pattern in Lampung Province is predominantly clustered, reflecting a tendency for settlements to be concentrated in specific areas rather than dispersed.

A continuous-discontinuous coupling computational method for multi-material mixtures

Considering the important role of particle-reinforced composites (PRCs), and to accurately assess the mechanical properties of PRC with arbitrarily complex internal structures, this paper proposes a new continuous-discontinuous coupled computational method, i.e., the multi-material integrated analysis (MIA) method, by combining the advantages of the discontinuous deformation analysis (DDA) method, the numerical manifold method (NMM), and the meshless method, for the continuous-discontinuous dual state of this kind of materials when they are subjected to forces. The core idea of this method is by establishing a single physical cover (PC) on each particle and defining a unified and piecewise approximation function over it to achieve displacement coordination between the particles and the matrix. In this model, the particles can maintain the contact and separation characteristics like the blocks, while the matrix is able to achieve the bonding function by overlapping different covers. Two numerical integration schemes, namely single-point Gaussian integration and Monte Carlo integration, are proposed to address the integration issues in stiffness matrix calculations. Meanwhile by considering the similarity between NMM and meshless method interpolation function construction, this paper combines circular coverage and 0th-order moving least squares to construct the interpolation function to form the displacement approximation function for the whole solution space. Moreover, the method of lattice retrieval (also Nearest Neighbor Search, NNS) is invoked for particle contact determination to improve the efficiency of contact and coverage relationship determination. The validity of the model is verified by taking asphalt mixture, a typical PRC, as a demonstrative case study. The results show that the MIA method solves the one-sided problem that the existing algorithms usually only consider the continuous or discontinuous properties of PRC, and provides a new method for comprehensively analyzing the micromechanical response of PRC and solving its large deformation problem.

HIERARCHICAL NEURAL NETWORKS IN PREDICTING THE PROPERTIES OF OIL AND GAS RESERVOIRS BASED ON WELL AND SEISMIC DATA

This paper describes a technique for hierarchical neural networks based on the nearest neighbor method with preliminary clustering of the original training dataset and construction of a search cluster decision tree. This method is a promising alternative to neural network technologies with deep learning and has quite a few advantages: high learning rate, identification of objects with a low degree of similarity, and the ability to generalize and retrain. As shown by testing the hierarchical neural network method on real data from the West Siberian oil and gas province, predicting the oil saturation in the Vikulov suite interval is much faster and more efficient than inversion approaches to quantitative interpretation of seismic data while achieving fairly similar geological results. This characterizes the proposed method of hierarchical neural networks as an effective tool for the quantitative interpretation of seismic data to solve geological problems.

Identification of Effective Factors and Prediction of Ischemic Heart Disease Using Machine Learning Methods and Data from the Yazd Health Study (YaHS) "

Introduction: Ischemic heart diseases are one of the most common diseases that cause high mortality worldwide. This article has identified various factors affecting heart disease and identified susceptible people using various machine learning methods. Methods: The current research was conducted on the Yazd Health Study (YaHS) database. YaHS was conducted on adults aged 20-70 years who were residents of Yazd Greater Area and collected information on the health and various diseases of nearly 10,000 people in the form of a questionnaire with 300 different questions. In this research, by using the correlation of questions with heart disease, the most important factors of heart disease have been identified. By using the identified factors and machine learning algorithms, susceptible people with heart disease have been identified. Results: The results of the evaluations have shown that factors such as age, family history of heart disease, blood pressure, diabetes, blood cholesterol, stress, anxiety, depression, quality of life, quality of sleep, physical activity, smoking, and diet have an effect on heart disease. Likewise, among the different machine learning methods, the nearest neighbor method, the deep neural network method, and the multi-layer perceptron method with a recall criterion of 99.94%, 99.88%, and 99.11% have performed the best in the identifying sick people, respectively. Conclusion: According to the findings of the research, it can be understood that by controlling factors such as blood pressure, diabetes, blood cholesterol, stress, anxiety, and depression, changing factors such as quality of life, sleep status, physical activity, and eating patterns of people and quitting smoking, it is possible to move towards improving the health of society. On the other hand, the identification of people prone to heart disease using machine learning methods is done faster and at a lower cost than the traditional methods that are done by referring to medical centers and doctors and performing various tests.

Using artificial intelligence algorithms in the investigation of mixed layer depth seasonal changes in the Barents Sea

The study aims at discovering features of seasonal changes in the mixed layer depth (MLD) of the Barents Sea in 1993–2020.Charts of the distribution of the MLD in the Barents Sea in 1993–2020 provided by the Copernicus Marine Service were used as the material of the study.Methods of the study: cluster analysis, machine learning, neuronal networks, the nearest neighbor method (kNN).Results. Classification of the data sets of the MLD distribution according to their seasonal features was carried out based on the modelling using AI algorithms and machine learning. It was concluded that winter is specified by two classes (increased/decreased values of the layer thickness). The third class includes spring and autumn when distributions of the MLD are close to one another, and the fourth class comprises summer (June-September) when the MLD grows very slowly.Practical relevance. The results will contribute to a better understanding of the hydrophysical processes of the Barents Sea and can further be used as series of independent variables to study the Barents Sea ecosystem and to estimate a stock and a catch forecast of commercial aquatic organisms.

Declustering characteristics of the North China Plain seismic belt and its effect on probabilistic seismic hazard analysis

The accurate and impartial identification of background seismic activity is an important foundation for earthquake model construction and probabilistic seismic hazard analysis (PSHA). We use the Gardner and Knopoff, Reasenberg, nearest-neighbor and stochastic declustering methods to decluster in the North China Plain seismic belt, analyzing the effect of declustering. We optimize the input parameters of the algorithms through Kolmogorov-Smirnov Poisson distribution tests (e.g., the clustering threshold of the nearest-neighbor method is set to 0.6). Four algorithms demonstrate good declustering effects, removing the impact of the strong seismicity in Tangshan while retaining similar trends to all events. The occurrence of major earthquake events can significantly impact the declustering characteristics in terms of time, space and magnitude. There is a difference in the overall hazard between the four declustering algorithms and the Fifth Generation Seismic Ground Motion Parameters Zonation Map of China. The difference in seismic hazard curves is mainly influenced by the annual average occurrence rate (background rates) and the Gutenberg-Richter b value, with the effect of the b value being more pronounced. The analysis of the effect of algorithms and their impact on PSHA can provide reference for earthquake risk assessment, engineering seismic design and disaster research.

Spatial-Aware and Load Balancing Distributed Data Partitioning Strategies for Content-Based Multimedia Retrieval.

Content-Based Multimedia Retrieval (CBMR) has become very popular in several applications, driven by the growing routine use of multimedia data. Since the datasets used in real-world applications are very large and descriptor's dimensionality is high, querying is an expensive, albeit important functionality. Further, exact search is prohibitive in most cases, motivating the use of Approximate Nearest Neighbour Search (ANNS) algorithms, trading accuracy for performance. These have been mainly developed targeting a sequential execution in a single node. However, the large and increasing datasets used and the high query loads submitted to those systems typically surpass the memory and computing resources available in a single node. This motivated the development of parallel distributed memory ANNS solutions to meet the computing capabilities required by those applications. A common problem that must be handled when using distributed memory systems is data partitioning and its impact on load imbalance. Several data partitioning approaches have already been proposed, including elaborated spatial-aware strategies. However, little effort has been put into carefully analyzing the performance of those strategies at scale. Here, we evaluated the commonly used data partitioning strategies in ANNS and identified their limitations to propose a novel class of partitioning algorithms that can minimize load imbalance while improving data locality to attain high performance on the distributed memory search. Experimentally, we found that our proposed algorithms (SABBS and SABBSR) improved search performance by up to 1.64× compared to the best previous solution. In a distributed memory weak scaling evaluation, with up to 12 billion 128-dimensional descriptors and 60 compute nodes, the gains were maintained as the system scaled with our novel approaches. These results demonstrate the efficiency of our new algorithms for billion-scale ANNS and the importance of considering not only data locality but also data and load imbalance in the data partitioning.

Revealing alterations in heart rate fluctuations during the progression of Chagas disease.

The heart rate variability (HRV) continually evolves throughout life, reflecting modifications in the architecture of the sinoatrial node (SAN) and in the regulation of heart rate by the autonomic nervous system (ANS). Both can be considerably affected by Chagas disease, causing important changes in the complex nature of HRV. We aim to evaluate the ability of an index based on the false nearest neighbors method (FN10) to reflect these changes during disease progression. We perform a retrospective, descriptive, and cross-sectional study analyzing HRV time series of participants with Chagas disease. We determine the dependence of FN10 on age and sex in a healthy population, and then evaluate FN10 in individuals with Chagas disease. In the healthy population, FN10 has a scaling behavior with age, which is independent of sex. In Chagas disease, some individuals show FN10 values significantly above those seen in the healthy population. We relate the findings to the pathophysiological mechanisms that determine the progression of the disease. The results indicate that FN10 may be a candidate prognostic biomarker for heart disease.

Development and Evaluation of the DMIND Questionnaire: Preparing for AI Integration into an Effective Depression Screening Tool

Objective: Thailand’s mental health crisis is exacerbated by high demand and a shortage of mental health professionals. The research objective was to develop and validate the Detection and Monitoring Intelligence Network for Depression (DMIND) questionnaire, designed to be culturally relevant and easily administered in clinical settings. Crafted with expert input, items specifically conducive to artificial intelligence (AI) analysis were selected to facilitate the future development of an AI-assisted depression scoring model. This approach underscores the tool’s dual utility in both human-led and technology-enhanced diagnostics. Materials and Methods: We enrolled 81 participants from psychiatric and tertiary care hospitals in Bangkok. Participants were assessed using the DMIND questionnaire, followed by the Hamilton Depression Rating Scale (HDRS-17). Statistical analyses included the content validity index (CVI), Cronbach’s alpha, Pearson’s correlation coefficient, Cohen’s kappa, and receiver operating characteristic (ROC) analysis. The Liu method, Youden index, and nearest neighbor method were used to determine the optimal cut-off point Results: The DMIND questionnaire showed strong validity, with an item-level CVI (I-CVI) and scale-level CVI (S-CVI) exceeding 1.0, indicating strong consensus on its relevance and utility. The tool also demonstrated high internal consistency (Cronbach’s alpha = 0.96). ROC analysis showed an AUC of 0.88, indicating high accuracy in depression screening. An optimal cut-off score of 11.5 was identified, balancing predictive value and sensitivity. Conclusion: The DMIND questionnaire represents a significant advancement in innovative mental health diagnostics, addressing unmet clinical needs by providing accurate and efficient assessments capable of AI integration for further enhancing mental health service delivery in Thailand.