Multiple Performance Metrics Research Articles

An Approach for Mitigation of Beam Blockage in mmWave-Based Indoor Networks

The rapidly evolving scenario around cellular networks, Internet of Things (IoT), and the associated services ecosystem has generated a wide range of expectations for the future networks. However, the unavailability of sufficient bandwidth and severe interference prohibit the existing commercially used radio-frequency (RF) bands from supporting some of these services. In this respect, the millimeter wave (mmWave)-based systems are seen as promising alternatives for supporting various high data rate enhanced mobile broadband (eMBB) services. Nevertheless, the adverse wave propagation characteristics in these bands and the resulting media access control (MAC) layer problems of beam blockage restrict these systems from operating in a reliable manner. Therefore, in this article, we propose a MAC layer-based approach of control delegation to tackle the problem of beam blockage. The mechanism uses policy-based relinquishment of access point (AP) control from the default AP to other AP capable member nodes to improve node visibility and alleviate the blockage scenario without requiring any additional network infrastructure. To account for the dynamic behavior of the network, a discrete-time Markov decision process (MDP)-based model is proposed. It uses a dynamic optimization methodology to accommodate multiple reconfigurable performance metrics, which can then be adjusted to cater to different performance criteria. Using computer simulations, we evaluate the effectiveness of our proposed solution for some alternate static and dynamic control policies. Our simulation results show that our proposed solution can significantly reduce the blockage in a network while satisfying a variety of performance objectives over different classes of indoor IoT scenarios.

Novel Application of Predictive Modeling: A Tailored Approach to Promoting HCC Surveillance in Patients With Cirrhosis

There has been increased interest in interventions to promote hepatocellular carcinoma (HCC) surveillance given low utilization and high proportions of late stage detection. Accurate prediction of patients likely versus unlikely to respond to interventions could allow a cost-effective approach to outreach and facilitate targeting more intensive interventions to likely non-responders. We conducted a secondary analysis of a randomized clinical trial evaluating a mailed outreach strategy to promote HCC surveillance among 1200 cirrhosis patients at a safety-net health system between December 2014 and March 2017. We developed regularized logistic regression (RLR) and gradient boosting machine (GBM) algorithm models to predict surveillance completion during each of the 3 screening rounds in a training set (n = 960). Model performance was assessed using multiple performance metrics in an independent test set (n = 240). Among 1200 patients, surveillance was completed in 41-47% of patients over the three rounds. The RLR and GBM models demonstrated good discriminatory accuracy, with area under receiver operating characteristic (AUROC) curves of 0.67 and 0.66 respectively in the first surveillance round and improved to 0.77 by the third surveillance round after incorporating prior screening behavior as a feature. Additional performance characteristics including the Brier score, Hosmer-Lemeshow test and reliability diagrams were also evaluated. The most important variables for the predictive model were prior screening completion status and past primary care contact. Predictive models can help stratify patients' likelihood to respond to surveillance outreach invitations, facilitating tailored strategies to maximize effectiveness and cost-effectiveness of HCC surveillance population health programs.

Cost-Aware Dynamic SFC Mapping and Scheduling in SDN/NFV-Enabled Space–Air–Ground-Integrated Networks for Internet of Vehicles

Space–air–ground-integrated networks (SAGINs) are deemed as a promising solution to support multifarious Internet of Vehicles (IoV) services with diversified Quality-of-Service (QoS) requirements in future communication networks. Network function virtualization (NFV) and software-defined networking (SDN) are two complementary and promising technologies to reduce the function provisioning cost and coordinate the heterogeneous physical resources in SAGIN. In this article, we investigate the online dynamic virtual network function (VNF) mapping and scheduling in SAGIN, considering the dynamicity of IoV services. The VNF live migration, VNF reinstantiation, and VNF rescheduling are enabled to increase the service acceptance ratio and service provider’s profits. Considering the heterogeneity of space, air, and ground nodes, we first model the migration cost and additional delay incurred by VNF live migration and reinstantiation. We then formulate the dynamic VNF mapping and scheduling jointly as a mixed-integer linear programming (MILP) problem with specified cost and delay models. We propose two Tabu search (TS)-based algorithms, i.e., TS-based VNF remapping and rescheduling (TS-MAPSCH) algorithm and TS-based pure VNF rescheduling (TS-PSCH) algorithm, to obtain suboptimal solutions to the MILP problem efficiently. Simulation results show that the proposed solution is very close to the optimum and that the proposed dynamic algorithms outperform existing works with respect to multiple performance metrics, including the service provider’s profit, service acceptance ratio, and QoS satisfaction level.

CNN architecture comparison for radio galaxy classification

ABSTRACT The morphological classification of radio sources is important to gain a full understanding of galaxy evolution processes and their relation with local environmental properties. Furthermore, the complex nature of the problem, its appeal for citizen scientists, and the large data rates generated by existing and upcoming radio telescopes combine to make the morphological classification of radio sources an ideal test case for the application of machine learning techniques. One approach that has shown great promise recently is convolutional neural networks (CNNs). Literature, however, lacks two major things when it comes to CNNs and radio galaxy morphological classification. First, a proper analysis of whether overfitting occurs when training CNNs to perform radio galaxy morphological classification using a small curated training set is needed. Secondly, a good comparative study regarding the practical applicability of the CNN architectures in literature is required. Both of these shortcomings are addressed in this paper. Multiple performance metrics are used for the latter comparative study, such as inference time, model complexity, computational complexity, and mean per class accuracy. As part of this study, we also investigate the effect that receptive field, stride length, and coverage have on recognition performance. For the sake of completeness, we also investigate the recognition performance gains that we can obtain by employing classification ensembles. A ranking system based upon recognition and computational performance is proposed. MCRGNet, Radio Galaxy Zoo, and ConvXpress (novel classifier) are the architectures that best balance computational requirements with recognition performance.

The MVTec Anomaly Detection Dataset: A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection

The detection of anomalous structures in natural image data is of utmost importance for numerous tasks in the field of computer vision. The development of methods for unsupervised anomaly detection requires data on which to train and evaluate new approaches and ideas. We introduce the MVTec anomaly detection dataset containing 5354 high-resolution color images of different object and texture categories. It contains normal, i.e., defect-free images intended for training and images with anomalies intended for testing. The anomalies manifest themselves in the form of over 70 different types of defects such as scratches, dents, contaminations, and various structural changes. In addition, we provide pixel-precise ground truth annotations for all anomalies. We conduct a thorough evaluation of current state-of-the-art unsupervised anomaly detection methods based on deep architectures such as convolutional autoencoders, generative adversarial networks, and feature descriptors using pretrained convolutional neural networks, as well as classical computer vision methods. We highlight the advantages and disadvantages of multiple performance metrics as well as threshold estimation techniques. This benchmark indicates that methods that leverage descriptors of pretrained networks outperform all other approaches and deep-learning-based generative models show considerable room for improvement.

Evaluating and Comparing Ontology Alignment Systems: An Mcdm Approach

Ontology alignment is vital in Semantic Web technologies with numerous applications in diverse disciplines. Due to diversity and abundance of ontology alignment systems, a proper evaluation can portray the evolution of ontology alignment and depicts the efficiency of a system for a particular domain. Evaluation can help system designers recognize the strength and shortcomings of their systems, and aid application developers to select a proper alignment system. This article presents a new evaluation and comparison methodology based on multiple performance met-rics that accommodates experts' preferences via a multi-criteria decision-making (MCDM) method, i.e., Bayesian best-worst method (BWM). First, the importance of a performance metric for a specific task/application is determined according to experts' preferences. The alignment systems are then evaluated based on proposed expert-based collective performance (ECP) that takes into account multiple metrics as well as their calibrated importance. For comparison , the alignment systems are ranked based on a probabilistic scheme, where it includes the extent to which one alignment system is preferred over another. The proposed methodology is applied to six tracks from ontology alignment evaluation initiative (OAEI), where the importance of performance metrics are calibrated by designing a survey and eliciting the preferences of ontology alignment experts. Accordingly, the participating alignment systems in the OAEI 2018 are evaluated and ranked. While the proposed methodology is applied to six OAEI tracks to demonstrate its applicability, it can also be applied to any benchmark or application of ontology alignment.

Machine Learning Based PV Power Generation Forecasting in Alice Springs

The generation volatility of photovoltaics (PVs) has created several control and operation challenges for grid operators. For a secure and reliable day or hour-ahead electricity dispatch, the grid operators need the visibility of their synchronous and asynchronous generators' capacity. It helps them to manage the spinning reserve, inertia and frequency response during any contingency events. This study attempts to provide a machine learning-based PV power generation forecasting for both the short and long-term. The study has chosen Alice Springs, one of the geographically solar energy-rich areas in Australia, and considered various environmental parameters. Different machine learning algorithms, including Linear Regression, Polynomial Regression, Decision Tree Regression, Support Vector Regression, Random Forest Regression, Long Short-Term Memory, and Multilayer Perceptron Regression, are considered in the study. Various comparative performance analysis is conducted for both normal and uncertain cases and found that Random Forest Regression performed better for our dataset. The impact of data normalization on forecasting performance is also analyzed using multiple performance metrics. The study may help the grid operators to choose an appropriate PV power forecasting algorithm and plan the time-ahead generation volatility.

Performance assessment of contemporary energy-optimized office buildings under the impact of climate change

Climate change can significantly impact the total energy consumption and indoor environment of buildings. This study investigates a potential adaptation and mitigation pathway by implementing a multi-objective optimization strategy to minimize the impact. A baseline building model was carefully established. A sampling-based approach and a multi-objective optimization method were combined to enlarge the vision of this study. Performance of a baseline model, baseline variants, and optimized models under different climate change scenarios and time frames were compared. The results reveal that the present building model will experience a 7.2%–12.3% increase in total energy consumption, along with a significantly longer overheating period in the future. Optimal models show considerable superiority in energy savings and overheating prevention compared to the baseline model and its variants; but under future climate conditions, none of them can help maintain both energy consumption and overheating at the present level of the baseline building. This study recommends that simultaneous evaluation of multiple performance metrics helps make a more accurate and comprehensive assessment of climate change's impact on buildings.

Comparison of 11 automated PET segmentation methods in lymphoma

Segmentation of lymphoma lesions in FDG PET/CT images is critical in both assessing individual lesions and quantifying patient disease burden. Simple thresholding methods remain common despite the large heterogeneity in lymphoma lesion location, size, and contrast. Here, we assess 11 automated PET segmentation methods for their use in two scenarios: individual lesion segmentation and patient-level disease quantification in lymphoma. Lesions on 18F-FDG PET/CT scans of 90 lymphoma patients were contoured by a nuclear medicine physician. Thresholding, active contours, clustering, adaptive region-growing, and convolutional neural network (CNN) methods were implemented on all physician-identified lesions. Lesion-level segmentation was evaluated using multiple segmentation performance metrics (Dice, Hausdorff Distance). Patient-level quantification of total disease burden (SUVtotal) and metabolic tumor volume (MTV) was assessed using Spearman’s correlation coefficients between the segmentation output and physician contours. Lesion segmentation and patient quantification performance was compared to inter-physician agreement in a subset of 20 patients segmented by a second nuclear medicine physician. In total, 1223 lesions with median tumor-to-background ratio of 4.0 and volume of 1.8 cm3, were evaluated. When assessed for lesion segmentation, a 3D CNN, DeepMedic, achieved the highest performance across all evaluation metrics. DeepMedic, clustering methods, and an iterative threshold method had lesion-level segmentation performance comparable to the degree of inter-physician agreement. For patient-level SUVtotal and MTV quantification, all methods except 40% and 50% SUVmax and adaptive region-growing achieved a performance that was similar the agreement of the two physicians. Multiple methods, including a 3D CNN, clustering, and an iterative threshold method, achieved both good lesion-level segmentation and patient-level quantification performance in a population of 90 lymphoma patients. These methods are thus recommended over thresholding methods such as 40% and 50% SUVmax, which were consistently found to be significantly outside the limits defined by inter-physician agreement.

An empirical characterization of fair machine learning for clinical risk prediction

The use of machine learning to guide clinical decision making has the potential to worsen existing health disparities. Several recent works frame the problem as that of algorithmic fairness, a framework that has attracted considerable attention and criticism. However, the appropriateness of this framework is unclear due to both ethical as well as technical considerations, the latter of which include trade-offs between measures of fairness and model performance that are not well-understood for predictive models of clinical outcomes. To inform the ongoing debate, we conduct an empirical study to characterize the impact of penalizing group fairness violations on an array of measures of model performance and group fairness. We repeat the analysis across multiple observational healthcare databases, clinical outcomes, and sensitive attributes. We find that procedures that penalize differences between the distributions of predictions across groups induce nearly-universal degradation of multiple performance metrics within groups. On examining the secondary impact of these procedures, we observe heterogeneity of the effect of these procedures on measures of fairness in calibration and ranking across experimental conditions. Beyond the reported trade-offs, we emphasize that analyses of algorithmic fairness in healthcare lack the contextual grounding and causal awareness necessary to reason about the mechanisms that lead to health disparities, as well as about the potential of algorithmic fairness methods to counteract those mechanisms. In light of these limitations, we encourage researchers building predictive models for clinical use to step outside the algorithmic fairness frame and engage critically with the broader sociotechnical context surrounding the use of machine learning in healthcare.

Innovative deep learning models for EEG-based vigilance detection

Electroencephalography (EEG) is one of the most signals used for studying and demonstrating the electrical activity of the brain due to the absence of side effects, its noninvasive nature and its well temporal resolution. Indeed, it provides real-time information, so it can be easily suitable for predicting drivers’ vigilance states. The classification of these states through this signal requires sophisticated approaches in order to achieve the best prediction performance. Furthermore, deep learning (DL) approaches have shown a good performance in learning the high-level features of the EEG signal and in resolving classification issues. In this paper, we will predict individuals’ states of vigilance based on the study of their brain activity by analyzing EEG signals using DL architectures. In fact, we propose two types of networks: (i) a 1D-UNet model, which is composed only of deep one-dimensional convolutional neural network (1D-CNN) layers and (ii) 1D-UNet-long short-term memory (1D-UNet-LSTM) that combines the proposed 1D-UNet architecture with the LSTM recurrent model. The experimental results reveal that the suggested models can stabilize the training model, well recognize the subject vigilance states and compete with the state of art on multiple performance metrics. The per-class average of precision and recall can be, respectively, up to 86% with 1D-UNet and 85% with 1D-UNet-LSTM, hence the effectiveness of the proposed methods. In order to complete our virtual prototyping and to get a real evaluation of our alert equipment, these proposed DL models are implemented also on a Raspberry Pi3 device allowing measuring the execution time necessary for predicting the state vigilance in real time.

Few-Shot Learning and Self-Training for eNodeB Log Analysis for Service-Level Assurance in LTE Networks

With the increasing network topology complexity and continuous evolution of the new wireless technology, it is challenging to address the network service outage with traditional methods. In the long-term evolution (LTE) networks, a large number of base stations called eNodeBs are deployed to cover the entire service areas spanning various kinds of geographical regions. Each eNodeB generates a large number of key performance indicators (KPIs). Hundreds of thousands of eNodeBs are typically deployed to cover a nation-wide service area. Operators need to handle hundreds of millions of KPIs to cover the areas. It is impractical to handle manually such a huge amount of KPI data, and automation of data processing is therefore desired. To improve network operation efficiency, a suitable machine learning technique is used to learn and classify individual eNodeBs into different states based on multiple performance metrics during a specific time window. However, an issue with supervised learning requires a large amount of labeled dataset, which takes costly human-labor and time to annotate data. To mitigate the cost and time issues, we propose a method based on few-shot learning that uses Prototypical Networks algorithm to complement the eNodeB states analysis. Using a dataset from a live LTE network that consists of thousand of eNodeB, our experiment results show that the proposed technique provides high performance while using a low number of labeled data.

Spatio-temporal analysis of land use and land cover change: a systematic model inter-comparison driven by integrated modelling techniques

ABSTRACT Currently, land use and land cover change (LULCC) is of utmost concern for global environmental change and sustainability. Among the portfolio of techniques, modelling is considered as the best approach to explore LULC dynamics. We have performed a model inter-comparison exercise (including hybrid and non-hybrid models) using multiple performance metrics to identify the best modelling approach for the subsequent projection of future LULC. The methodology is demonstrated on the Subarnarekha basin of Eastern India by utilizing the LANDSAT imagery of 1989, 1994, 2006, and 2011. Before LULC modelling, the cross-tabulation and trend-surface analyses were performed to identify dominant land transitions in post-classification maps. Temporal mapping results over 1989–2011 exhibited a drastic decrease in the area under dense forest (25.7% to 19.0%), a substantial increase in the area under scrubland (21.0% to 26.1%) and a nominal reduction in the coverage of the agricultural land (51.2% to 49.0%). Four integrated models namely Multilayer perceptron-Markov Model (MLP-MC), Logistic Regression-Markov Model (LR-MC), and two hybrid models, i.e. Multilayer perceptron-Cellular automata-Markov model (MLP-CA-MC) and Logistic Regression-Cellular automata-Markov model (LR-CA-MC) were tested for their suitability for predicting future LULC for the basin. Based on the multiple model validation techniques, the MLP-MC model performed the best. MLP-MC model subsequently used a non-stationary relationship between selected explanatory variables and LULC to predict the future LULC for 2020 and 2030. The MLP-MC model projected that relative to the level of 2011, agricultural land, dense forest, and barren land may decrease by 8.3, 28.2 and 23.5%, respectively, and scrubland, built-up area, and water bodies may increase by 22.5, 87.3 and 13.3%, respectively, by 2030. Our findings contradict the prevalent view regarding the nationwide intensification of agriculture over the Indian subcontinent but are consistent with the national decreasing trend in the dense forest. The study provides a transferable methodology for the systematic comparison of LULC models (including hybrid and non-hybrid) against multiple performance metrics. The outcomes of the study may help land-use planners, environmentalist, and policymakers in framing better policies and management .recommendations.

Light intensity and seed density differentially affect the establishment, survival, and biomass of an exotic invader and three species of native competitors

Understanding how invasive and native plant species respond to their environment is essential for explaining and managing plant invasions in changing ecosystems. Light limitation and density-dependent interactions affect plant performance, but few studies have experimentally evaluated the effects of these factors on invasive and native neighbors simultaneously. Our objective was to test the hypothesis that light limitation and density-dependent interference are critical factors that influence the performance of the invasive species Bromus madritensis ssp. rubens (“B. rubens”) and three native species that commonly compete with B. rubens in North American drylands. In a full-factorial experiment, we examined the effects of (a) light limitation, (b) increasing B. rubens seed density, and (c) increasing native seed density on the establishment, survival, and biomass of each study species over a 12-week growing period in a controlled greenhouse. Light limitation reduced at least one performance metric (establishment, survival, or biomass) for all species. Increasing B. rubens seed density had no intraspecific effect on its own performance but had variable and species-specific negative effects on natives. Increasing native seed density reduced B. rubens establishment, survival, and biomass but had variable and species-specific effects on the natives themselves. These findings suggest that light intensity and density-dependent interactions can strongly but differentially influence the establishment, survival, and biomass of an important subset of invasive and native species in North American drylands. Evaluating multiple performance metrics at the species level is essential for understanding how invasive and native neighbors respond to environmental variation in changing dryland ecosystems.

On the efficiency of K-means clustering

This paper presents a thorough evaluation of the existing methods that accelerate Lloyd's algorithm for fast k -means clustering. To do so, we analyze the pruning mechanisms of existing methods, and summarize their common pipeline into a unified evaluation framework UniK. UniK embraces a class of well-known methods and enables a fine-grained performance breakdown. Within UniK, we thoroughly evaluate the pros and cons of existing methods using multiple performance metrics on a number of datasets. Furthermore, we derive an optimized algorithm over UniK, which effectively hybridizes multiple existing methods for more aggressive pruning. To take this further, we investigate whether the most efficient method for a given clustering task can be automatically selected by machine learning, to benefit practitioners and researchers.

Queueing Versus Surge Pricing Mechanism: Efficiency, Equity, and Consumer Welfare

At the end of 2017, the leading ride-hailing platform DiDi Express replaced the commonly used surge pricing mechanism with a queueing mechanism, called the virtual queueing mechanism, in a number of cities across China to mitigate public relations pressure. To explore the benefits of the virtual queueing mechanism over the surge pricing mechanism in terms of operational efficiency, equity, and consumer welfare, we build a model of a data-driven state-dependent open queueing network, which we show can be approximated by an M/M/s+M queueing model with balking and reneging. We derive closed-form expressions for multiple performance metrics, including response rate, demand satisfaction rate, gross merchandise value (GMV), consumer surplus, and degree of inequality. Our theoretical and numerical analyses show that: (1) the virtual queueing mechanism outperforms the surge pricing mechanism in terms of consumer surplus, but is not as advantageous to generate GMV; (2) with respect to the response rate and demand satisfaction rate, the virtual queueing mechanism performs increasingly better than the surge pricing mechanism as demand increases; (3) the virtual queueing mechanism contributes to a more equitable ridesharing system, in the sense that consumer surplus is distributed across different classes of riders in a more balanced way. Using real data, we conduct a case study in Beijing, which verifies all the aforementioned results. The case study further finds that the response and demand satisfaction rates are significantly correlated with driver idleness.

Modeling and Analysis of Dynamic Charging for EVs: A Stochastic Geometry Approach

With the increasing demand for greener and more energy efficient transportation solutions, electric vehicles (EVs) have emerged to be the future of transportation across the globe. However, currently, one of the biggest bottlenecks of EVs is the battery. Small batteries limit the EVs driving range, while big batteries are expensive and not environmentally friendly. One potential solution to this challenge is the deployment of charging roads, i.e., dynamic wireless charging systems installed under the roads that enable EVs to be charged while driving. In this paper, we use tools from stochastic geometry to establish a framework that enables evaluating the performance of charging roads deployment in metropolitan cities. We first present the course of actions that a driver should take when driving from a random source to a random destination in order to maximize dynamic charging during the trip. Next, we analyze the distribution of the distance to the nearest charging road. This distribution is vital for studying multiple performance metrics such as the trip efficiency, which we define as the fraction of the total trip spent on charging roads. Next, we derive the probability that a given trip passes through at least one charging road. The derived probability distributions can be used to assist urban planners and policy makers in designing the deployment plans of dynamic wireless charging systems. In addition, they can also be used by drivers and automobile manufacturers in choosing the best driving routes given the road conditions and level of energy of EV battery.

A Node Deployment Model with Variable Transmission Distance for Wireless Sensor Networks

The deployment of network nodes is essential to ensure the wireless sensor network's regular operation and affects the multiple network performance metrics, such as connectivity, coverage, lifetime, and cost. This paper focuses on the problem of minimizing network costs while meeting network requirements, and proposes a corona-based deployment method by using the variable transmission distance sensor. Based on the analysis of node energy consumption and network cost, an optimization model to minimize Cost Per Unit Area is given. The transmission distances and initial energy of the sensors are obtained by solving the model. The optimization model is improved to ensure the energy consumption balance of nodes in the same corona. Based on these parameters, the process of network node deployment is given. Deploying the network through this method will greatly reduce network costs.

Balancing services from built and natural assets via river basin trade-off analysis

Built water infrastructure impacts the balance of services provided by a river and its flow regime. Impacts on both commercial and subsistence activities should be considered in water management decision-making. Various methods used to define mandatory minimum environmental releases do not account for the inherent and often complex trade-offs and synergies which must be considered in selecting a balance of ecosystem and engineered services. This paper demonstrates the value and use of optimised many-objective trade-off analysis for managing resource-systems providing diverse and sometimes competing services. Using Kenya’s Tana River basin as a demonstration it shows controlled releases from multi-reservoir systems can be optimised using multiple performance metrics, representing individual provisioning ecosystem and engineered services at different locations and relating to different time periods. This enables better understanding the interactions between natural and built assets, and selecting river basin interventions that appropriately trade-off their services. Our demonstration shows prioritising Kenya’s statutory minimum environmental ‘reserve’ flows degrades flood-related provisioning services. Low overall flow regime alteration correlates negatively with consistency of hydropower generation, but positively with other provisioning services.

Deep clustering by maximizing mutual information in variational auto-encoder

Unsupervised clustering, which is extensively employed in deep learning and computer vision as a fundamental technique, has attracted much attention in recent years. Deep embedding clustering often uses auto-encoders to learn representations for clustering. However, auto-encoders tend to corrupt the learning representations when simultaneously learning embedded representations and performing clustering. In this paper, we propose a Deep Clustering via Variational Auto-Encoder (DC-VAE) of mutual information maximization. First, we formulate the deep clustering problem as learning soft cluster assignments within the framework of variational auto-encoder. Second, we impose mutual information maximization on the observed data and the representations to prevent soft cluster assignments from distorting learning representations. Third, we derive a new generalization evidence lower bound objects related to several previous models and introduce parameters to balance learning informative representations and clustering. It is shown that the proposed model can significantly boost the performance of clustering by learning effective and reliable representations for downstream machine learning tasks. Through experimental results on several datasets, we demonstrate that the proposed model is competitive with existing state-of-the-arts on multiple performance metrics.