Dinh-Mao Bui

On curating multimodal sensory data for personalized health and wellness services

In recent years, the focus of healthcare and wellness technologies has shown a significant shift towards personal vital signs devices. The technology has evolved from smartphone-based wellness applications to fitness bands and smartwatches. The novelty of these devices is the accumulation of activity data as their users go about their daily life routine. However, these implementations are device specific and lack the ability to incorporate multimodal data sources. Data accumulated in their usage does not offer rich contextual information that is adequate for providing a holistic view of a user’s lifelog. As a result, making decisions and generating recommendations based on this data are single dimensional. In this paper, we present our Data Curation Framework (DCF) which is device independent and accumulates a user’s sensory data from multimodal data sources in real time. DCF curates the context of this accumulated data over the user’s lifelog. DCF provides rule-based anomaly detection over this context-rich lifelog in real time. To provide computation and persistence over the large volume of sensory data, DCF utilizes the distributed and ubiquitous environment of the cloud platform. DCF has been evaluated for its performance, correctness, ability to detect complex anomalies, and management support for a large volume of sensory data.

Energy efficiency for cloud computing system based on predictive optimization

In recent years, power consumption has become one of the hottest research trends in computer science and industry. Most of the reasons are related to the operational budget and the environmental issues. In this paper, we would like to propose an energy-efficient solution for orchestrating the resource in cloud computing. In nature, the proposed approach firstly predicts the resource utilization of the upcoming period based on the Gaussian process regression method. Subsequently, the convex optimization technique is engaged to compute an appropriate quantity of physical servers for each monitoring window. This quantity of interest is calculated to ensure that a minimum number of servers can still provide an acceptable quality of service. Finally, a corresponding migrating instruction is issued to stack the virtual machines and turn off the idle physical servers to achieve the objective of energy savings. In order to evaluate the proposed method, we conduct the experiments using synthetic data from 29-day period of Google traces and real workload from the Montage open-source toolkit. Through the evaluation, we show that the proposed approach can achieve a significant result in reducing the energy consumption as well as maintaining the system performance. Balance between energy efficiency and quality of service in the cloud computing.Apply prediction technique to enhance the usefulness of monitoring statistics.Design optimal energy efficiency architecture to orchestrate the cloud system.After VM consolidation, adaptively turn-off idle physical machines to save energy.

Adaptive Replication Management in HDFS Based on Supervised Learning

The number of applications based on Apache Hadoop is dramatically increasing due to the robustness and dynamic features of this system. At the heart of Apache Hadoop, the Hadoop Distributed File System (HDFS) provides the reliability and high availability for computation by applying a static replication by default. However, because of the characteristics of parallel operations on the application layer, the access rate for each data file in HDFS is completely different. Consequently, maintaining the same replication mechanism for every data file leads to detrimental effects on the performance. By rigorously considering the drawbacks of the HDFS replication, this paper proposes an approach to dynamically replicate the data file based on the predictive analysis. With the help of probability theory, the utilization of each data file can be predicted to create a corresponding replication strategy. Eventually, the popular files can be subsequently replicated according to their own access potentials. For the remaining low potential files, an erasure code is applied to maintain the reliability. Hence, our approach simultaneously improves the availability while keeping the reliability in comparison to the default scheme. Furthermore, the complexity reduction is applied to enhance the effectiveness of the prediction when dealing with Big Data.

Gaussian process for predicting CPU utilization and its application to energy efficiency

For the past ten years, Gaussian process has become increasingly popular for modeling numerous inferences and reasoning solutions due to the robustness and dynamic features. Particularly concerning regression and classification data, the combination of Gaussian process and Bayesian learning is considered to be one of the most appropriate supervised learning approaches in terms of accuracy and tractability. However, due to the high complexity in computation and data storage, Gaussian process performs poorly when processing large input dataset. Because of the limitation, this method is ill-equipped to deal with the large-scale system that requires reasonable precision and fast reaction rate. To improve the drawback, our research focuses on a comprehensive analysis of Gaussian process performance issues, highlighting ways to drastically reduce the complexity of hyper-parameter learning and training phases, which could be applicable in predicting the CPU utilization in the demonstrated application. In fact, the purpose of this application is to save the energy by distributively engaging the Gaussian process regression to monitor and predict the status of each computing node. Subsequently, a migration mechanism is applied to migrate the system-level processes between multi-core and turn off the idle one in order to reduce the power consumption while still maintaining the overall performance.

PAM-based flexible generative topic model for 3D interactive activity recognition

Interactive activity recognition from the RGB videos still remains a challenge, therefore some existing approaches paid the attention to RGB-Depth video process to avoid problems relating to mutual occlusion and redundant human pose and to improve accuracy of skeleton extraction. From the single action to complex interaction activity, it is necessary an efficient model to describe the relationship of body components between multi-human objects. In this research, the authors proposed a hierarchical model based on the Pachinko Allocation Model for interaction recognition. Concretely, the joint features comprising joint distant and joint motion are calculated from the skeleton position and then support to topic modeling. The probabilistic models describing the flexible relationship between features - poselets - activities are generated by this model. Finally, the Binary Tree of Support Vector Machine is applied for classification. Compared with existing state-of-the-arts, the proposed method outperforms in overall classification accuracy (8-21% approximately) with the SBU Kinect Interaction Dataset.

Traffic Behavior Recognition Using the Pachinko Allocation Model.

CCTV-based behavior recognition systems have gained considerable attention in recent years in the transportation surveillance domain for identifying unusual patterns, such as traffic jams, accidents, dangerous driving and other abnormal behaviors. In this paper, a novel approach for traffic behavior modeling is presented for video-based road surveillance. The proposed system combines the pachinko allocation model (PAM) and support vector machine (SVM) for a hierarchical representation and identification of traffic behavior. A background subtraction technique using Gaussian mixture models (GMMs) and an object tracking mechanism based on Kalman filters are utilized to firstly construct the object trajectories. Then, the sparse features comprising the locations and directions of the moving objects are modeled by PAM into traffic topics, namely activities and behaviors. As a key innovation, PAM captures not only the correlation among the activities, but also among the behaviors based on the arbitrary directed acyclic graph (DAG). The SVM classifier is then utilized on top to train and recognize the traffic activity and behavior. The proposed model shows more flexibility and greater expressive power than the commonly-used latent Dirichlet allocation (LDA) approach, leading to a higher recognition accuracy in the behavior classification.

Energy savings in processor based on prediction technique

Green computing has become one of the hottest trends in recent years. In this research area, the major purpose is to reduce the energy consumption as well as the CO2 emission. Obviously, this topic has been the important issue in the field of electronic and computer engineering. In fact, energy factor might be considered to be a significant cost when running any computing system. Basically, energy savings can be obtained in many parts of the system including memory, peripheral devices, hard disk drive and processor. In processor or CPU level, there is a number of solutions to handle the power consumption. However, most of them based on reactive model which engages the thresholds. Obviously, these techniques are not accuracy and limited to save the power. In this research, a proactive solution based on prediction technique is proposed. Firstly, the utilization of each core of processor is anticipated by using Gaussian process regression. Subsequently, a migration mechanism can be used to migrate the system-level processes between these cores. Finally, the idle cores can be turned off to save the power while still maintaining an acceptable performance.

Evaluating Large-Scale Biomedical Ontology Matching Over Parallel Platforms

ABSTRACT Biomedical systems have been using ontology matching as a primary technique for heterogeneity resolution. However, the natural intricacy and vastness of biomedical data have compelled biomedical ontologies to become large-scale and complex; consequently, biomedical ontology matching has become a computationally intensive task. Our parallel heterogeneity resolution system, i.e., SPHeRe, is built to cater the performance needs of ontology matching by exploiting the parallelism-enabled multicore nature of todays desktop PC and cloud infrastructure. In this paper, we present the execution and evaluation results of SPHeRe over large-scale biomedical ontologies. We evaluate our system by integrating it with the interoperability engine of a clinical decision support system (CDSS), which generates matching requests for large-scale NCI, FMA, and SNOMED-CT biomedical ontologies. Results demonstrate that our methodology provides an impressive performance speedup of 4.8 and 9.5 times over a quad-core desktop PC and a four virtual machine (VM) cloud platform, respectively.

Optimizing Power Consumption in Cloud Computing based on Optimization and Predictive Analysis

Due to the budget and the environmental issues, achieving energy efficiency gradually receives a lot of attentions these days. In our previous research, a prediction technique has been developed to improve the monitoring statistics. In this research, by adopting the predictive monitoring information, our new proposal can perform the optimization to solve the energy issue of cloud computing. Actually, the optimization technique, which is convex optimization, is coupled with the proposed prediction method to produce a near-optimal set of hosting physical machines. After that, a corresponding migrating instruction can be created eventually. Based on this instruction, the cloud orchestrator can suitably relocate virtual machines to a designed subset of infrastructure. Subsequently, the idle physical servers can be turned off in an appropriate manner to save the power as well as maintain the system performance. For the purpose of evaluation, an experiment is conducted based on 29-day period of Google traces. By utilizing this evaluation, the proposed approach shows the potential to significantly reduce the power consumption without affecting the quality of services.

Early fault detection in IaaS cloud computing based on fuzzy logic and prediction technique

Availability is one of the most important requirements in production system. Keeping a persistent level of high availability in the Infrastructure-as-a-Service (IaaS) cloud computing is a challenge due to the complexity of service providing. By definition, the availability can be maintained by coupling with the fault tolerance approaches. Recently, many fault tolerance methods have been developed, but few of them adequately consider the fault detection aspect, which is critical to issue the appropriate recovery actions just in time. In this paper, based on a rigorous analysis on the nature of failures, we would like to introduce a method to early identify the faults occurring in the IaaS system. By engaging fuzzy logic algorithm and prediction technique, the proposed approach can provide better performance in terms of accuracy and reaction rate, which subsequently enhances the system reliability.