This paper introduces a novel network planning problem called Multi-stage Edge Server Deployment (M-ESD). The problem calls for a solution that (i) adds fixed edge servers to an existing Multi-access Edge Computing (MEC) network incrementally over multiple stages, e.g., in years, and (ii) optimizes the offloading of tasks to installed servers. More specifically, when upgrading a network, at each stage, the problem involves the following constraints: (i) budget (in $), (ii) server deployment cost (in $) and cost depreciation rate (in %), (iii) number of tasks and their increase rate (in %), and (iv) server storage capacity . The goal of M-ESD is to ensure the resulting network maximizes the average number of tasks that meet their delay requirement. This paper presents a Mixed Integer Linear Programming (MILP) model and a heuristic approach called M-ESD/H to solve the M-ESD problem. Simulation results on small networks show that M-ESD/H produces results that are within 13.6% of the optimal MILP solution. Further, it significantly reduces runtime and produces results in less than 0.1 s as compared to MILP, which failed to produce results in some networks after running for over 48 h. For large networks, M-ESD/H is compared against two versions of M-ESD that consider arbitrary budget allocation and/or edge server placement, i.e., M-ESD/A1 and M-ESD/A2. The results show that M-ESD/H outperforms both M-ESD/A1 and M-ESD/A2 across various options with varying numbers of stages, budget allocation, and tasks.

Visual scene understanding mainly depends on pixel-wise classification obtained from a deep convolutional neural network. However, existing semantic segmentation models often face difficulties in real-time applications due to their large network architecture. Although there are real-time semantic segmentation models available, their shallow backbone can degrade the performance considerably. This paper introduces SDBNetV2, a lightweight semantic segmentation model designed to improve real-time performance without increasing computational costs. A key contribution is a novel Short-term Dense Bottleneck (SDB) module in the encoder, which provides varied field-of-views to capture different geometrical objects in a complex scene. Additionally, we propose dense feature refinement and improved semantic aggregation modules at the decoder end to enhance contextualization and object localization. We evaluate the proposed model’s performance on several indoor and outdoor datasets in structured and unstructured environments. The results show that SDBNetV2 achieves superior segmentation performance over other real-time models with less than 2 million parameters.

Deep neural networks (DNNs) are currently being deployed as machine learning technology in a wide range of important real-world applications. DNNs consist of a huge number of parameters that require millions of floating-point operations (FLOPs) to be executed both in learning and prediction modes. A more effective method is to implement DNNs in a cloud computing system equipped with centralized servers and data storage sub-systems with high-speed and high-performance computing capabilities. This paper presents an up-to-date survey on current state-of-the-art deployed DNNs for cloud computing. Various DNN complexities associated with different architectures are presented and discussed alongside the necessities of using cloud computing. We also present an extensive overview of different cloud computing platforms for the deployment of DNNs and discuss them in detail. Moreover, DNN applications already deployed in cloud computing systems are reviewed to demonstrate the advantages of using cloud computing for DNNs. The paper emphasizes the challenges of deploying DNNs in cloud computing systems and provides guidance on enhancing current and new deployments.

Recent studies show a significant growth in semantic segmentation . However, many semantic segmentation models still have a large number of parameters, making them unsuitable for resource-constrained embedded devices. To address this issue, we propose an efficient Shared Feature Reuse Segmentation (SFRSeg) model containing several novelties: a new yet effective shared-branch multiple sub-encoders design, a context mining module and a semantic aggregating module for better context granularity . In particular, our shared-branch approach improves the entire feature hierarchy by sharing the spatial and context knowledge in both shallow and deep branches. After every shared point in each sub-encoder, a proposed cascading context mining (CCM) module is deployed to filter out the noisy spatial details from the feature maps and provides a diverse size of receptive fields for capturing the latent context between multi-scale geometric shapes in the scene. To overcome the gradient vanishing issue at the early stage, we reduce the number of layers in the first sub-encoder and employ a unique multiple sub-encoders design which reprocesses the rich global feature maps through multiple sub-encoders for better feature refinement. Later, the rich semantic features generated by the efficient sub-encoders at different levels are fused by the proposed Hybrid Path Attention Semantic Aggregation (HPA-SA) module that effectively reduces the semantic gap between feature maps at different levels and alleviate the well-known boundary degeneration effect. To make it computationally efficient for resource-constrained embedded devices, a series of lightweight methods such as a lightweight encoder, a squeeze-and-excitation design, separable convolution filters , channel reduction (CR) are carefully exploited. With an exceptional performance on Cityscapes (70.6% test mIoU) and CamVid (74.7% test mIoU) data sets, the proposed model is shown to be superior over existing light real-time semantic segmentation models whilst having only 1.6 million parameters.

This paper considers the problem of upgrading a legacy network into a Software Defined Network (SDN) over multiple stages and maximizing energy saving (ES) in the resulting upgraded network or hybrid SDN. In each stage, an operator needs to select and replace legacy switches with SDN switches and seek to switch off as many cables as possible over each link. This paper addresses the said problem where it considers (i) the available budget at each stage, (ii) maximum path delays, (iii) maximum link utilization, (iv) per-stage increase (decrease) in traffic size (upgrade cost), and (v) each non SDN switch must comply with the Open Shortest Path First (OSPF)-Equal Cost Multi-Path (ECMP) protocol. It outlines a Mixed Integer Program (MIP) and a heuristic algorithm called M-GMSU. The results show that (i) MIP and M-GMSU achieve ES of up to 71.93%, (ii) using a larger budget and/or number of stages increases ES, and (iii) the ES of M-GMSU is within 3.55% away from the optimal ES computed by MIP.

A major challenge in stream clustering is the evolution in the statistical properties of the underlying data. As clustering is inherently unsupervised, selecting suitable parameter values is often difficult. Clustering algorithms with sensitive parameters are often not robust to such changes, leadi ng to poor clustering outputs. Algorithms using K-NN graphs face this problem, as they have a sensitive K-connectivity parameter which prohibits them from adapting to stream concept evolution. We address this by controlling the excess of the skewness of edge length distributions in the underlying K-NN graph by introducing novel skewness excess concept. We demonstrate the asymptotic linear dependency of skewness excess against the graph connectivity and propose the novel RobustRepStream algorithm, which extends the RepStream algorithm, and provides improved robustness against stream evolution. By automatically controlling the skewness excess, the user no longer needs to specify the K-connectivity parameter, and RobustRepStream can adjust the graph connectivity locally in order to achieve performance close to when the optimal K value is known. We demonstrate that RobustRepStream’s skewness threshold parameter is insensitive and universal across all data sets. We comprehensively evaluate RobustRepStream on real-world benchmark data sets against previous stream clustering algorithms, and demonstrate that it provides better clustering performance.

Rotary kilns are important industrial plants to process material. In this article, we address the problem of detecting material rings formed in alumina rotary kilns, which leads to high waste and substantial economic loss. This necessitates the use of artificial intelligence to help kiln operators detect material rings promptly. We describe a recently developed system that extracts useful features and detects the presence of material rings efficiently and accurately in real time. Our major contribution is a novel feature extraction method based on geometric transformation and peak localization exploiting the domain knowledge. We also contribute a large data set covering many thousand labeled frames for the evaluation of ring detection. On this data set, we demonstrate that our novel feature extraction method outperforms other alternatives in the literature. Our method is also demonstrated as favorable over deep learning approaches. Our system produces an overall accuracy of nearly 96%, which is acceptable for deployment.

We propose an effective subspace selection scheme as a post-processing step to improve results obtained by sparse subspace clustering (SSC). Our method starts by the computation of stable subspaces using a novel random sampling scheme. Thus constructed preliminary subspaces are used to identify the initially incorrectly clustered data points and then to reassign them to more suitable clusters based on their goodness-of-fit to the preliminary model. To improve the robustness of the algorithm, we use a dominant nearest subspace classification scheme that controls the level of sensitivity against reassignment. We demonstrate that our algorithm is convergent and superior to the direct application of a generic alternative such as principal component analysis. On several popular datasets for motion segmentation and face clustering pervasively used in the sparse subspace clustering literature the proposed method is shown to reduce greatly the incidence of clustering errors while introducing negligible disturbance to the data points already correctly clustered.

Our aim is to estimate the perspective-effected geometric distortion of a scene from a video feed. In contrast to most related previous work, in this task we are constrained to use low-level spatiotemporally local motion features only. This particular challenge arises in many semiautomatic surveillance systems that alert a human operator to potential abnormalities in the scene. Low-level spatiotemporally local motion features are sparse (and thus require comparatively little storage space) and sufficiently powerful in the context of video abnormality detection to reduce the need for human intervention by more than 100-fold. This paper introduces three significant contributions. First, we describe a dense algorithm for perspective estimation, which uses motion features to estimate the perspective distortion at each image locus and then polls all such local estimates to arrive at the globally best estimate. Second, we also present an alternative coarse algorithm that subdivides the image frame into blocks and uses motion features to derive block-specific motion characteristics and constrain the relationships between these characteristics, with the perspective estimate emerging as a result of a global optimization scheme. Third, we report the results of an evaluation using nine large sets acquired using existing closed-circuit television cameras, not installed specifically for the purposes of this paper. Our findings demonstrate that both proposed methods are successful, their accuracy matching that of human labeling using complete visual data (by the constraints of the setup unavailable to our algorithms).

The need to estimate a particular quantile of a distribution is an important problem which frequently arises in many computer vision and signal processing applications. For example, our work was motivated by the requirements of many semi-automatic surveillance analytics systems which detect abnormalities in close-circuit television (CCTV) footage using statistical models of low-level motion features. In this paper we specifically address the problem of estimating the running quantile of a data stream with non-stationary stochasticity when the memory for storing observations is limited. We make several major contributions: (i) we derive an important theoretical result which shows that the change in the quantile of a stream is constrained regardless of the stochastic properties of data, (ii) we describe a set of high-level design goals for an effective estimation algorithm that emerge as a consequence of our theoretical findings, (iii) we introduce a novel algorithm which implements the aforementioned design goals by retaining a sample of data values in a manner adaptive to changes in the distribution of data and progressively narrowing down its focus in the periods of quasi-stationary stochasticity, and (iv) we present a comprehensive evaluation of the proposed algorithm and compare it with the existing methods in the literature on both synthetic data sets and three large ‘real-world’ streams acquired in the course of operation of an existing commercial surveillance system. Our findings convincingly demonstrate that the proposed method is highly successful and vastly outperforms the existing alternatives, especially when the target quantile is high valued and the available buffer capacity severely limited.

Many vision problems deal with high-dimensional data, such as motion segmentation and face clustering. However, these high-dimensional data usually lie in a low-dimensional structure. Sparse representation is a powerful principle for solving a number of clustering problems with high-dimensional data. This principle is motivated from an ideal modeling of data points according to linear algebra theory. However, real data in computer vision are unlikely to follow the ideal model perfectly. In this paper, we exploit the mixed norm regularization for sparse subspace clustering. This regularization term is a convex combination of the ℓ1norm, which promotes sparsity at the individual level and the block norm ℓ2/1which promotes group sparsity. Combining these powerful regularization terms will provide a more accurate modeling, subsequently leading to a better solution for the affinity matrix used in sparse subspace clustering. This could help us achieve better performance on motion segmentation and face clustering problems. This formulation also caters for different types of data corruptions. We derive a provably convergent algorithm based on the alternating direction method of multipliers (ADMM) framework, which is computationally efficient, to solve the formulation. We demonstrate that this formulation outperforms other state-of-arts on both motion segmentation and face clustering.

The need to estimate a particular quantile of a distribution is an important problem that frequently arises in many computer vision and signal processing applications. For example, our work was motivated by the requirements of many semiautomatic surveillance analytics systems that detect abnormalities in close-circuit television footage using statistical models of low-level motion features. In this paper, we specifically address the problem of estimating the running quantile of a data stream when the memory for storing observations is limited. We make the following several major contributions: 1) we highlight the limitations of approaches previously described in the literature that make them unsuitable for nonstationary streams; 2) we describe a novel principle for the utilization of the available storage space; 3) we introduce two novel algorithms that exploit the proposed principle in different ways; and 4) we present a comprehensive evaluation and analysis of the proposed algorithms and the existing methods in the literature on both synthetic data sets and three large real-world streams acquired in the course of operation of an existing commercial surveillance system. Our findings convincingly demonstrate that both of the proposed methods are highly successful and vastly outperform the existing alternatives. We show that the better of the two algorithms (data-aligned histogram) exhibits far superior performance in comparison with the previously described methods, achieving more than 10 times lower estimate errors on real-world data, even when its available working memory is an order of magnitude smaller.

Dynamically changing background (dynamic background) still presents a great challenge to many motion-based video surveillance systems. In the context of event detection, it is a major source of false alarms. There is a strong need from the security industry either to detect and suppress these false alarms, or dampen the effects of background changes, so as to increase the sensitivity to meaningful events of interest. In this paper, we restrict our focus to one of the most common causes of dynamic background changes: 1) that of swaying tree branches and 2) their shadows under windy conditions. Considering the ultimate goal in a video analytics pipeline, we formulate a new dynamic background detection problem as a signal processing alternative to the previously described but unreliable computer vision-based approaches. Within this new framework, we directly reduce the number of false alarms by testing if the detected events are due to characteristic background motions. In addition, we introduce a new data set suitable for the evaluation of dynamic background detection. It consists of real-world events detected by a commercial surveillance system from two static surveillance cameras. The research question we address is whether dynamic background can be detected reliably and efficiently using simple motion features and in the presence of similar but meaningful events, such as loitering. Inspired by the tree aerodynamics theory, we propose a novel method named local variation persistence (LVP), that captures the key characteristics of swaying motions. The method is posed as a convex optimization problem, whose variable is the local variation. We derive a computationally efficient algorithm for solving the optimization problem, the solution of which is then used to form a powerful detection statistic. On our newly collected data set, we demonstrate that the proposed LVP achieves excellent detection results and outperforms the best alternative adapted from existing art in the dynamic background literature.

We propose in this paper a novel sparse subspace clustering method that regularizes sparse subspace representation by exploiting the structural sharing between tasks and data points via group sparse coding. We derive simple, provably convergent, and computationally efficient algorithms for solving the proposed group formulations. We demonstrate the advantage of the framework on three challenging benchmark datasets ranging from medical record data to image and text clustering and show that they consistently outperforms rival methods.

Compressed sensing (CS) is an important theory for sub-Nyquist sampling and recovery of compressible data. Recently, it has been extended to cope with the case where corruption to the CS data is modeled as impulsive noise. The new formulation, termed as robust CS, combines robust statistics and CS into a single framework to suppress outliers in the CS recovery. To solve the newly formulated robust CS problem, a scheme that iteratively solves a number of CS problems--the solutions from which provably converge to the true robust CS solution--is suggested. This scheme is, however, rather inefficient as it has to use existing CS solvers as a proxy. To overcome limitations with the original robust CS algorithm, we propose in this paper more computationally efficient algorithms by following latest advances in large-scale convex optimization for nonsmooth regularization. Furthermore, we also extend the robust CS formulation to various settings, including additional affine constraints, l1-norm loss function, mix-norm regularization, and multitasking, so as to further improve robust CS and derive simple but effective algorithms to solve these extensions. We demonstrate that the new algorithms provide much better computational advantage over the original robust CS method on the original robust CS formulation, and effectively solve more sophisticated extensions where the original methods simply cannot. We demonstrate the usefulness of the extensions on several imaging tasks.

We present a novel approach to improving subspace clustering by exploiting the spatial constraints. The new method encourages the sparse solution to be consistent with the spatial geometry of the tracked points, by embedding weights into the sparse formulation. By doing so, we are able to correct sparse representations in a principled manner without introducing much additional computational cost. We discuss alternative ways to treat the missing and corrupted data using the latest theory in robust lasso regression and suggest numerical algorithms so solve the proposed formulation. The experiments on the benchmark Johns Hopkins 155 dataset demonstrate that exploiting spatial constraints significantly improves motion segmentation.

We present a novel method for document clustering using sparse representation of documents in conjunction with spectral clustering. An ℓ1-norm optimization formulation is posed to learn the sparse representation of each document, allowing us to characterize the affinity between documents by considering the overall information instead of traditional pair wise similarities. This document affinity is encoded through a graph on which spectral clustering is performed. The decomposition into multiple subspaces allows documents to be part of a sub-group that shares a smaller set of similar vocabulary, thus allowing for cleaner clusters. Extensive experimental evaluations on two real-world datasets from Reuters-21578 and 20Newsgroup corpora show that our proposed method consistently outperforms state-of-the-art algorithms. Significantly, the performance improvement over other methods is prominent for this datasets.

Compressed sensing (CS) is a new information sampling theory for acquiring sparse or compressible data with much fewer measurements than those otherwise required by the Nyquist/Shannon counterpart. This is particularly important for some imaging applications such as magnetic resonance imaging or in astronomy. However, in the existing CS formulation, the use of the l(2) norm on the residuals is not particularly efficient when the noise is impulsive. This could lead to an increase in the upper bound of the recovery error. To address this problem, we consider a robust formulation for CS to suppress outliers in the residuals. We propose an iterative algorithm for solving the robust CS problem that exploits the power of existing CS solvers. We also show that the upper bound on the recovery error in the case of non-Gaussian noise is reduced and then demonstrate the efficacy of the method through numerical studies.

We identify and formulate a novel problem: cross channel anomaly detection from multiple data channels. Cross channel anomalies are common amongst the individual channel anomalies, and are often portent of significant events. Using spectral approaches, we propose a two-stage detection method: anomaly detection at a single-channel level, followed by the detection of cross-channel anomalies from the amalgamation of single channel anomalies. Our mathematical analysis shows that our method is likely to reduce the false alarm rate. We demonstrate our method in two applications: document understanding with multiple text corpora, and detection of repeated anomalies in video surveillance. The experimental results consistently demonstrate the superior performance of our method compared with related state-of-art methods, including the one-class SVM and principal component pursuit. In addition, our framework can be deployed in a decentralized manner, lending itself for large scale data stream analysis.

We extend a recent Sparse Representation-based Classification (SRC) algorithm for face recognition to work on 2D images directly, aiming to reduce the computational complexity whilst still maintaining performance. Our contributions include: (1) a new 2D extension of SRC algorithm; (2) an incremental computing procedure which can reduce the eigen decomposition expense of each 2D-SRC for sequential input data; and (3) extensive numerical studies to validate the proposed methods.

This work addresses the issue of Bayesian robustness in the multivariate normal model when the prior covariance matrix is not completely specified, but rather is described in terms of positive semi-definite bounds. This occurs in situations where, for example, the only prior information available is the bound on the diagonal of the covariance matrix derived from some physical constraints, and that the covariance matrix is positive semi-definite, but otherwise arbitrary. Under the conditional Gamma-minimax principle, previous work by DasGupta and Studden shows that an analytically exact solution is readily available for a special case where the bound difference is a scaled identity. The goal in this work is to consider this problem for general positive definite matrices. The contribution in this paper is a theoretical study of the geometry of the minimax problem. Extension of previous results to a more general case is shown and a practical algorithm that relies on semi-definite programming and the convexity of the minimax formulation is derived. Although the algorithm is numerically exact for up to the bivariate case, its exactness for other cases remains open. Numerical studies demonstrate the accuracy of the proposed algorithm and the robustness of the minimax solution relative to standard and recently proposed methods.

This paper addresses a major challenge in data mining applications where the full information about the underlying processes, such as sensor networks or large online database, cannot be practically obtained due to physical limitations such as low bandwidth or memory, storage, or computing power. Motivated by the recent theory on direct information sampling called compressed sensing (CS), we propose a framework for detecting anomalies from these large-scale data mining applications where the full information is not practically possible to obtain. Exploiting the fact that the intrinsic dimension of the data in these applications are typically small relative to the raw dimension and the fact that compressed sensing is capable of capturing most information with few measurements, our work show that spectral methods that used for volume anomaly detection can be directly applied to the CS data with guarantee on performance. Our theoretical contributions are supported by extensive experimental results on large datasets which show satisfactory performance.

Learning a robust projection with a small number of training samples is still a challenging problem in face recognition, especially when the unseen faces have extreme variation in pose, illumination, and facial expression. To address this problem, we propose a framework formulated under statistical learning theory that facilitates robust learning of a discriminative projection. Dimensionality reduction using the projection matrix is combined with a linear classifier in the regularized framework of lasso regression. The projection matrix in conjunction with the classifier parameters are then found by solving an optimization problem over the Stiefel manifold. The experimental results on standard face databases suggest that the proposed method outperforms some recent regularized techniques when the number of training samples is small.

We propose a joint representation and classification framework that achieves the dual goal of finding the most discriminative sparse overcomplete encoding and optimal classifier parameters. Formulating an optimization problem that combines the objective function of the classification with the representation error of both labeled and unlabeled data, constrained by sparsity, we propose an algorithm that alternates between solving for subsets of parameters, whilst preserving the sparsity. The method is then evaluated over two important classification problems in computer vision: object categorization of natural images using the Caltech 101 database and face recognition using the Extended Yale B face database. The results show that the proposed method is competitive against other recently proposed sparse overcomplete counterparts and considerably outperforms many recently proposed face recognition techniques when the number training samples is small.

We propose a sequential M-estimation algorithm as an alternative to sequential least squares. Being an approximation of the exact M-estimator, the proposed technique is robust to nonGaussian processes and outperforms sequential least squares. Simulation results demonstrate the power of the proposed sequential M-estimator.

We consider the problem of joint detection and decoding for CDMA systems that employ forward error control (FEC). The main problems encountered in designing this type of receiver include high complexity and a lack of robustness in impulsive noise and contamination in the prior model. We propose that the extrinsic information may not need to be absolutely reliable as we consider impulsive interference. By allowing the judgement of confidence in using this extrinsic information, the proposed M-estimation based soft-in soft-out (SISO) detector exhibits robust performance for a low complexity.