Distance-based Approach Research Articles

Durability evaluation of retrofitted corroded reinforced concrete columns with FRP sheets in marine environmental conditions

The purpose of this paper is to investigate the effect of marine environmental condition on durability of RC-corroded columns strengthened with FRP sheets. In this study, three retrofit methods based on substrate repairs were evaluated. There are two common retrofit methods for the corroded RC columns, which are directly wrapping FRP (method 1) and wrapping FRP after replacing damaged concrete with repair mortar (method 2), respectively. Also, a durable retrofit method was developed in this study (method 3). In this method, after removing damaged concrete, wrapping was carried out, and then repair mortar was applied around GFRP sheet. After strengthening, specimens were stored in a marine simulator for 3000 and 9000h. In retrofit method 1, ultimate strength of retrofitted columns by one layer of GFRP and CFRP sheet decreased by 27.4% and 19.7%, respectively, after 9000h marine exposure. The decrease in ultimate strength of columns wrapped by one layer of GFRP and CFRP in retrofit method 2 was 25% and 18.3%, respectively. In retrofit method 3, 9000h marine exposure caused 8.8% and 2.2% increase in ultimate strength of strengthened columns with one and two layers of GFRP, respectively. The failure mode of strengthened columns by methods 2 and 3 was brittle. Finally, Distance Based Approach (DBA) was used in order to compare different retrofit methods. It was found that the performance of retrofitted columns in method 3 would be ideal in marine environmental conditions.

A knowledge-based outranking approach for multi-criteria decision-making with hesitant fuzzy linguistic term sets

The modeling and solving of multi-criteria decision-making (MCDM) problems under uncertainty is still a challenging topic. In real-life decision-making, using linguistic terms to represent experts’ judgments is suitable and straightforward since precise quantitative values may often be unavailable or the cost for their computation is too high. The introduction of hesitant fuzzy linguistic term sets (HFLTSs) was motivated by the limitations of prior linguistic fuzzy models and need for richer linguistic tools. However, since their introduction, comparing HFLTSs is still one of the major concerns of researchers in this area. The existing approaches in the literature commonly rely on (1) labels and intervals from the linguistic terms as the central elements of an envelope-based approach or (2) linguistic scale functions as the basis of a distance-based approach. The two approaches retain certain shortcomings resulting information distortion and loss which may inevitably degrade their credibility. In this paper, the authors are involved in the recent proposal of combining outranking approaches with HFLTSs in an MCDM context. After reviewing the existing approaches, an outranking method based on a novel knowledge-based paradigm for comparing HFLTSs is developed. Alternatively, the paradigm's foundations are the introduced concepts of fuzzy preference relations and profiles considering uncertainty degrees in decision makers’ assessments. The paradigm is then associated with a multi-criteria relational clustering (MCRC) algorithm that additionally extracts fuzzy preference relations between the resultant clusters. Last, an illustrative example is given to verify the appropriateness and efficacy of the developed approach and comparisons are made with other existing ones.

Design of character-based DNA barcode motif for species identification: A computational approach and its validation in fishes.

The DNA barcodes are generally interpreted using distance-based and character-based methods. The former uses clustering of comparable groups, based on the relative genetic distance, while the latter is based on the presence or absence of discrete nucleotide substitutions. The distance-based approach has a limitation in defining a universal species boundary across the taxa as the rate of mtDNA evolution is not constant throughout the taxa. However, character-based approach more accurately defines this using a unique set of nucleotide characters. The character-based analysis of full-length barcode has some inherent limitations, like sequencing of the full-length barcode, use of a sparse-data matrix and lack of a uniform diagnostic position for each group. A short continuous stretch of a fragment can be used to resolve the limitations. Here, we observe that a 154-bp fragment, from the transversion-rich domain of 1367 COI barcode sequences can successfully delimit species in the three most diverse orders of freshwater fishes. This fragment is used to design species-specific barcode motifs for 109 species by the character-based method, which successfully identifies the correct species using a pattern-matching program. The motifs also correctly identify geographically isolated population of the Cypriniformes species. Further, this region is validated as a species-specific mini-barcode for freshwater fishes by successful PCR amplification and sequencing of the motif (154bp) using the designed primers. We anticipate that use of such motifs will enhance the diagnostic power of DNA barcode, and the mini-barcode approach will greatly benefit the field-based system of rapid species identification.

Optimization of two-stage location–routing–inventory problem with time-windows in food distribution network

High transportation cost and low service quality are common weaknesses in different logistics networks, especially in food delivery. Due to its perishable features, the quality of the food will deteriorate during the while delivery process. In this paper, a two-stage location–routing–inventory problem with time windows (2S-LRITW) for food products is proposed. The first stage corresponds to a location–routing–inventory problem with time windows and the second stage is the transportation problem with vehicle capacity constraints. The problem is formulated as a mixed integer-programming model. Then a hybrid heuristic is proposed, in which distance-based clustering approach, mutation operation and the relocate exchange method are introduced to improve performance of algorithm. At last, the hybrid heuristic was tested using several cases, including some small instances and a real-life case. The results show that the distance-based clustering approach can efficiently improve the convergence speed. And the IACO and the relocate exchange method can enlarge the search space. As a result, the hybrid heuristic is suitable for solving practical larger scale problem. Furthermore, the results also indicate that customer sequence has great effect on the object due to the energy cost considered.

Scaling Online Games with NetConnectors

This article presents a peer-to-peer overlay for massively multiplayer online games with a focus on fast-paced action. More than other genres, action games like first-person shooters employ fast and dynamic game mechanics. In multiplayer environments, these properties have to be reflected by the underlying network structure. At the same time, the system should be able to support a huge amount of users in order to deliver a massive experience to the participating players. The capacity of current client/server systems limits the number of players in a game, preventing the desired massive experience. To provide both a scalable and a responsive system, we use a fully distributed peer-to-peer network with a dynamic connection scheme. By exploiting local interests in the virtual world, our system supports a huge number of users. Therefore, an area-of-interest mechanism is applied to the connection scheme. Users do not connect to all participating users, but they only establish connections to other users they are interested in. These neighbors are determined by the user's perception of the virtual world. Instead of using a purely distance-based approach, our system uses a more flexible neighbor-based approach that supports the use of multiple metrics to determine the set of interesting nodes for each user. A second kind of connection—so-called NetConnectors—utilizes the players' distribution in the virtual world to ensure overlay consistency. For the dissemination of messages, we use a publish/subscribe mechanism. This prevents inconsistencies introduced by unidirectional neighborhood relations that can occur with sender-oriented models. Further, the publish/subscribe mechanism models the users' interests more accurately. In addition to the regular sending mechanism, we implemented a Geocast algorithm that allows information distribution to arbitrary regions of the virtual world. While regular messages are always addressed to specific users, Geocasts cover certain geographical regions. Thus, Geocasts can be used to disseminate messages to all users that are located in the addressed region. Simulations show that our design performs well in terms of scalability. By keeping the amount of connections per user nearly constant, users do not get overloaded with too many connections. This also applies for crowded regions where the user density is much higher compared to an evenly populated virtual world. Another important aspect of fast-paced multiplayer games is the users' motion behavior. Different movement strategies are evaluated for their impact on network load and connection dynamics.

Parallel Distance-Based Instance Selection Algorithm for Feed-Forward Neural Network

AbstractInstance selection endeavors to decide which instances from the data set should be maintained for further use during the learning process. It can result in increased generalization of the learning model, shorter time of the learning process, or scaling up to large data sources. This paper presents a parallel distance-based instance selection approach for a feed-forward neural network (FFNN), which can utilize all available processing power to reduce the data set while obtaining similar levels of classification accuracy as when the original data set is used. The algorithm identifies the instances at the decision boundary between consecutive classes of data, which are essential for placing hyperplane decision surfaces, and retains these instances in the reduced data set (subset). Each identified instance, called a prototype, is one of the representatives of the decision boundary of its class that constitutes the shape or distribution model of the data set. No feature or dimension is sacrificed in the reduction process. Regarding reduction capability, the algorithm obtains approximately 85% reduction power on non-overlapping two-class synthetic data sets, 70% reduction power on highly overlapping two-class synthetic data sets, and 77% reduction power on multiclass real-world data sets. Regarding generalization, the reduced data sets obtain similar levels of classification accuracy as when the original data set is used on both FFNN and support vector machine. Regarding execution time requirement, the speedup of the parallel algorithm over the serial algorithm is proportional to the number of threads the processor can run concurrently.

Human Intention Inference Using Expectation-Maximization Algorithm With Online Model Learning

An algorithm called adaptive-neural-intention estimator (ANIE) is presented to infer the intent of a human operator’s arm movements based on the observations from a 3-D camera sensor (Microsoft Kinect). Intentions are modeled as the goal locations of reaching motions in 3-D space. Human arm’s nonlinear motion dynamics are modeled using an unknown nonlinear function with intentions represented as parameters. The unknown model is learned by using a neural network. Based on the learned model, an approximate expectation-maximization algorithm is developed to infer human intentions. Furthermore, an identifier-based online model learning algorithm is developed to adapt to the variations in the arm motion dynamics, the motion trajectory, the goal locations, and the initial conditions of different human subjects. The results of experiments conducted on data obtained from different users performing a variety of reaching motions are presented. The ANIE algorithm is compared with an unsupervised Gaussian mixture model algorithm and an Euclidean distance-based approach by using Cornell’s CAD-120 data set and data collected in the Robotics and Controls Laboratoy at UConn. The ANIE algorithm is compared with the inverse LQR and ATCRF algorithms using a labeling task carried out on the CAD-120 data set.

Robust Intra-Class Distance-Based Approach for Multimodal Biometric Game Theory-Based Rank-Level Fusion of Ear, Palmprint and Signature

Biometrics is a branch of science which recognizes the person by using its behavioral and physiological characteristics. Limitations of unimodal biometric systems lead the researchers to investigate a system based on fusion of some unimodal biometric systems. In this paper, an innovative verification framework based on game theory-based rank-level fusion of the high frequency information of ear images and combination of dynamic information of online signatures with the texture information of palmprint image is proposed. In this framework, Gabor filter is used to extract locally the texture features of the ear, palmprint, and signature images. Moreover, intra-class distances are used to distinguish the genuine samples from the imposters, and also investigate the proper threshold to separate these two classes from each other. To fuse the unimodal biometrics, we use the game theory to obtain optimal logistic regression weights of rank-level fusion. For this purpose, using a three-player game and a novel multi-biometric model, the rank-level fusion of the ear, palmprint, and signature with optimal weights is carried out. The proposed verification framework is evaluated on SVC2004, PolyU and UND databases. The obtained results show the effectiveness and efficiency of our framework in comparison with other presented biometric fusion algorithms in the literature.

Short Communication: HIV-1 Transmission Networks Across South Korea.

Molecular epidemiology can help clarify the properties and dynamics of HIV-1 transmission networks in both global and regional scales. We studied 143 HIV-1-infected individuals recruited from four medical centers of three cities in South Korea between April 2013 and May 2014. HIV-1 env V3 sequence data were generated (337-793 bp) and analyzed using a pairwise distance-based clustering approach to infer putative transmission networks. Participants whose viruses were ≤2.0% divergent according to Tamura-Nei 93 genetic distance were defined as clustering. We collected demographic, risk, and clinical data and analyzed these data in relation to clustering. Among 143 participants, we identified nine putative transmission clusters of different sizes (range 2-4 participants). The reported risk factor of participants were concordant in only one network involving two participants, that is, both individuals reported homosexual sex as their risk factor. The participants in the other eight networks did not report concordant risk factors, although they were phylogenetically linked. About half of the participants refused to report their risk factor. Overall, molecular epidemiology provides more information to understand local transmission networks and the risks associated with these networks.

A distance-based spectral clustering approach with applications to network community detection

Abstract Spectral clustering is an important unsupervised learning approach to many object partitioning and pattern analysis problems. In this paper, we present our work on a novel spectral clustering algorithm that groups a collection of objects using the spectrum of the pairwise distance matrix. If the points in a metric space can be associated with a well-defined distance, it is proven that the pairwise distance matrix is almost negative definite, and we show that the eigenvectors for its most significant negative eigenvalue can be used to approximate the solution to a quadratic binary partition problem. We define the quality measures for the one dimensional partitioning of the eigenvector entries, which are further applied to evaluate the partitioning results for the data points projected into the space spanned by the selected eigenvectors. Since the Lanczos iterative algorithm may be revised to find the eigenvalues efficiently in a distributed way, we adapt this algorithm to the network community detection problem using a decentralized multi-agent framework. The performance of the proposed approach is tested with different datasets, and the empirical experiments show that this approach is able to enhance the effectiveness of clustering.

Development of a data-based interval kicking program for preparation and rehabilitation purposes in professional football

ABSTRACTStudy Design: Cross-sectional Observational Cohort Study. Level of Evidence: 3.Objectives: Document kicking loads of professional soccer players in match play, and describe differences in velocity-based and distance-based classification. Describe a kicking programme suitable for physical preparation or return to play.Background: To determine reasonable kicking loads (volume and intensity) to plan for preparation and rehabilitation, normal match loads of kicking need to be described. It is not known if the kicking demands vary by playing position.Methods: Match analysis data for the entire 2012 season of the US Major League Soccer was examined using both a distance-based and velocity-based approach. Number and types of kicks were described for each playing position using both approaches.Results: Distance and velocity-based approaches were not seen to be equivalent, and a velocity-based approach was seen to be more reflective of the true kicking demands. The number and type of kicks made in each of the categories were seen to vary by playing position. On average, per game, goalkeepers made the fewest “slow” (0–6 m · s–1) kicks with 24.3 (9.1), left midfielders the most with 68.0 (32.2). Goalkeepers made the fewest “medium” (6–12 m · s–1) kicks with 9.1 (4.1) and central midfielders made the most with 36.8 (15.3). Attackers made the fewest “fast” (>12 m · s–1) kicks with 16.9 (10.1) while the left and right backs made the most with 30.8 (9.8).Conclusion: Kicking load varies by position and is best described using a velocity-based approach. These data can to inform physical preparation for kicking loads.

A Novel Satisfaction Function and Distance-Based Approach for Machining Performance Optimization During Electro-Discharge Machining on Super Alloy Inconel 718

The present work aims to determine an appropriate setting of process parameters (viz. gap voltage, peak discharge current, pulse-on time, duty factor and flushing pressure) for achieving optimal machining performance during electro-discharge machining of super alloy Inconel 718 by using copper tool electrode. Experiments have been performed based on \(L_{25}\) orthogonal array design of experiment by varying each of the aforesaid process parameter at five different levels. The machining performances have been evaluated in terms of material removal rate, electrode wear rate, surface roughness (\(R_{\mathrm{a}})\), surface crack density, white layer thickness, and micro-hardness of the EDMed Inconel 718 end product. In this paper, a novel optimization route (combining satisfaction function, distance measure approach in conjugation with Taguchi’s philosophy) has been introduced. Application feasibility of the aforementioned approach has been compared to that of obtained through exploration of principal component analysis and combined quality loss concept integrated with Taguchi method. Additionally, morphology of the EDMed work surface of Inconel 718 has been investigated through scanning electron microscopy.

Landscape genetic approaches to guide native plant restoration in the Mojave Desert.

Restoring dryland ecosystems is a global challenge due to synergistic drivers of disturbance coupled with unpredictable environmental conditions. Dryland plant species have evolved complex life-history strategies to cope with fluctuating resources and climatic extremes. Although rarely quantified, local adaptation is likely widespread among these species and potentially influences restoration outcomes. The common practice of reintroducing propagules to restore dryland ecosystems, often across large spatial scales, compels evaluation of adaptive divergence within these species. Such evaluations are critical to understanding the consequences of large-scale manipulation of gene flow and to predicting success of restoration efforts. However, genetic information for species of interest can be difficult and expensive to obtain through traditional common garden experiments. Recent advances in landscape genetics offer marker-based approaches for identifying environmental drivers of adaptive genetic variability in non-model species, but tools are still needed to link these approaches with practical aspects of ecological restoration. Here, we combine spatially explicit landscape genetics models with flexible visualization tools to demonstrate how cost-effective evaluations of adaptive genetic divergence can facilitate implementation of different seed sourcing strategies in ecological restoration. We apply these methods to Amplified Fragment Length Polymorphism (AFLP) markers genotyped in two Mojave Desert shrub species of high restoration importance: the long-lived, wind-pollinated gymnosperm Ephedra nevadensis, and the short-lived, insect-pollinated angiosperm Sphaeralcea ambigua. Mean annual temperature was identified as an important driver of adaptive genetic divergence for both species. Ephedra showed stronger adaptive divergence with respect to precipitation variability, while temperature variability and precipitation averages explained a larger fraction of adaptive divergence in Sphaeralcea. We describe multivariate statistical approaches for interpolating spatial patterns of adaptive divergence while accounting for potential bias due to neutral genetic structure. Through a spatial bootstrapping procedure, we also visualize patterns in the magnitude of model uncertainty. Finally, we introduce an interactive, distance-based mapping approach that explicitly links marker-based models of adaptive divergence with local or admixture seed sourcing strategies, promoting effective native plant restoration.

Manifold-ranking embedded order preserving hashing for image semantic retrieval

Due to the storage and computational efficiency of hashing technology, it has proven a valuable tool for large scale similarity search. In many cases, the large scale data in real-world lie near some (unknown) low-dimensional and non-linear manifold. Moreover, Manifold Ranking approach can preserve the global topological structure of the data set more effectively than Euclidean Distance-based Ranking approach, which fails to preserve the semantic relevance degree. However, most existing hashing methods ignore the global topological structure of the data set. The key issue is how to incorporate the global topological structure of data set into learning effective hashing function. In this paper, we propose a novel unsupervised hashing approach, namely Manifold-Ranking Embedded Order Preserving Hashing (MREOPH). A manifold ranking loss is introduced to solve the issue of global topological structure preserving. An order preserving loss is introduced to ensure the consistency between manifold ranking and hamming ranking. A hypercubic quantization loss is introduced to learn discrete binary codes. The information theoretic regularization term is taken into consideration for preserving desirable properties of hash codes. Finally, we integrate them in a joint optimization framework for minimizing the information loss in each processing. Experimental results on three datasets for semantic search clearly demonstrate the effectiveness of the proposed method.

Inferential Sensor for Estimation of the Concentration of Benzene in the Distillation Column Using TSK Fuzzy System Based on Modified Clustering Approach

An inferential sensor is a computer program used for inferring the process variables, which are very hard to measure from the available measurement data. Measurement noises can affect the quality of the data which can be improved by wavelet denoising method. The objective of this paper is to design an inferential sensor for estimation of Benzene concentration in a typical distillation column. Selection of the most relevant input variables for estimation can improve the performance of inferential sensor which is done by Principal Component Analysis (PCA) technique. In this paper an inferential sensor is proposed based on a novel modification of the nearest neighbor distance-based clustering for developing a Takagi-Sugeno-Kang (TSK) fuzzy model optimized by the Particle Swarm Optimization (PSO) algorithm. The proposed technique was compared against the conventional nearest neighbor distance-based clustering approach optimized by PSO. The simulation results confirm that the designed inferential sensor based on the proposed method is more accurate even for a noisy data set.

Statistical methods for damage detection applied to civil structures

Damage detection consists of monitoring the deviations of a current system from its reference state, characterized by some nominal property repeatable for every healthy state. Preferably, the damage detection is performed directly on vibration data, hereby avoiding modal identification of the structure. The practical aspect of using only the output measurements cause difficulties because of variations in ambient excitation due to variability in the environmental conditions, like sea, wind, and temperature. In this paper, a new Mahalanobis distance-based damage detection method is studied and compared to the well-known subspace-based damage detection algorithm in the context of two large case studies. Both methods are implemented in the modal analysis and structural health monitoring software ARTeMIS, in which the joint features of the methods are concluded in a control chart in an attempt to enhance the resolution of the damage detection. The damage indicators from both methods are evaluated based on the ambient vibration signals from numerical simulations on a novel offshore support structure and an experimental campaign with a full scale bridge. The results reveal that the performance of the two damage detection methods is similar, hereby implying merit of the new Mahalanobis distance-based approach, as it is less computational complex. The fusion of the damage indicators in the control chart provides the most accurate view on the progressively damaged systems.

Towards efficient mobile image-guided navigation through removal of outliers

A novel approach for positioning using smartphones and image processing techniques is developed. Using structure from motion, 3D reconstructions of given tracks are created and stored as sparse point clouds. Query images are matched later to these 3D models. High computational costs of image matching and limited storage require compressing point clouds without loss of positioning performance. In this work, localization is improved and memory and storage requirements are minimized. We assumed that the computational speed and, at the same time, storage requirements benefit from reducing the number of points with appropriate outlier detection. In particular, our hypothesis was that positioning accuracy is maintained while reducing outliers in a reconstructed model. To evaluate the hypothesis, three methods were compared: (i) density-based (Sotoodeh, International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences XXXVI-5, 2006), (ii) connectivity-based (Wang et al. Comput Graph Forum 32(5):207–10, 2013), and (iii) our distance-based approach. In tenfold cross-validation, applied to a pre-reconstructed reference 3D model, localization accuracy was measured. In each new model, the positions of test images were identified and compared to the according positions in the reference model. We observed that outlier removal has a positive impact on matching run-time and storage requirements, while there are no significant differences in the localization error within the methods. That confirmed our initial hypothesis and allows mobile application of image-based positioning.

Generalized sensitivity analysis study in basin and petroleum system modeling, case study on Piceance Basin, Colorado

Basin and petroleum system modeling is an interdisciplinary endeavor, and utilizes integrated data to study sedimentary basins. Numerical models are constructed and simulated to quantitatively model the geodynamic processes in sedimentary basins. Often, the modeling process covers large spatial and temporal intervals with many uncertain input parameters. These can be continuous parameters characterized by certain statistical distribution, spatially distributed variables, or discrete parameters such as geological scenarios. Identifying sensitivities from these complex input model parameters and recognizing key uncertainties are crucial and challenging for basin modeling development and applications. The major contribution of this work is to introduce and implement efficient and accurate sensitivity analysis approach for basin and petroleum system modeling discipline.We investigated two types of sensitivity analysis methods and compared their performance for identifying the impact of uncertain parameters on both spatial and temporal model responses. The first approach utilizes the variance-based Sobol indices to quantify parameter sensitivities, while the second approach is a distance-based sensitivity analysis which utilizes the distance between model responses to determine sensitivities of different parameters. The sensitivity analysis approaches are illustrated through a basin modeling example involving the processes of sediment compaction, source rock maturation and hydrocarbon generation in the Piceance Basin, Colorado, US. Monte Carlo samples of the input uncertainties related to physical properties of the source rock, thermal boundary conditions, and geological setting scenarios are generated. Multiple basin models constructed using these uncertain input parameters are simulated across the geological time span, and time-varying model response (hydrocarbon generation from Lower Cretaceous to present-day) and spatial model response (pressure and porosity distribution at present-day) are obtained. Sensitive parameters that impact these spatio-temporal model responses are then analyzed.Results show that the distance-based sensitivity analysis approach could achieve similar results with fewer model runs compared to the variance-based Sobol method. Model responses in spatial and temporal domain are impacted by different uncertain input parameters. Subtle relationship between input parameters and model response could also be identified. In particular an unexpected link between chemical kinetics and porosity versus depth behavior was uncovered. The knowledge obtained from sensitivity studies enhance the understanding of the complex geological processes and can benefit the modeling development and forecast capability. Though the sensitivity results are case-specific, the approach and workflow are generally applicable to other basins and earth sciences modeling studies.

An intelligent and improved density and distance-based clustering approach for industrial survey data classification

Engineering Asset Management (EAM) emphasizes on achieving sustainable business outcomes and competitive advantages by applying systematic and risk-based processes to decisions concerning an organization's physical assets. Nowadays, there is no specific method to evaluate performance of EAM and lack of benchmark to rank performance. To fill this gap, an improved density and distance-based clustering approach is proposed. The proposed approach is intelligent and efficient. It has largely simplified the current evaluating method so that the commitment in resources for manual data analyzing and performance ranking can be significantly reduced. Moreover, the proposed approach provides a basis on benchmarking for measuring and ranking the performance in Engineering Asset Management (EAM). Additionally, by using the intelligent approach, companies can avoid to pay expensive consultant fees for inviting external consultancy company to provide the necessary EAM auditing and performance benchmarking.

A New Approach to Dimensionality Reduction for Anomaly Detection in Data Traffic

The monitoring and management of high-volume feature-rich traffic in large networks offers significant challenges in storage, transmission, and computational costs. The predominant approach to reducing these costs is based on performing a linear mapping of the data to a low-dimensional subspace such that a certain large percentage of the variance in the data is preserved in the low-dimensional representation. This variance-based subspace approach to dimensionality reduction forces a fixed choice of the number of dimensions, is not responsive to real-time shifts in observed traffic patterns, and is vulnerable to normal traffic spoofing. Based on theoretical insights proved in this paper, we propose a new distance-based approach to dimensionality reduction motivated by the fact that the real-time structural differences between the covariance matrices of the observed and the normal traffic is more relevant to anomaly detection than the structure of the training data alone. Our approach, called the distance-based subspace method, allows a different number of reduced dimensions in different time windows and arrives at only the number of dimensions necessary for effective anomaly detection. We present centralized and distributed versions of our algorithm and, using simulation on real traffic traces, demonstrate the qualitative and quantitative advantages of the distance-based subspace approach.