Search

Recent deep learning models succeed in achieving high accuracy and fast inference time, but they require high-performance computing resources because they have a large number of parameters. However, not all systems have high-performance hardware. Sometimes, a deep learning model needs to be run on edge devices such as IoT devices or smartphones. On edge devices, however, limited computing resources are available and the amount of computation must be reduced to launch the deep learning models. Pruning is one of the well-known approaches for deriving light-weight models by eliminating weights, channels or filters. In this work, we propose “zero-keep filter pruning” for energy-efficient deep neural networks. The proposed method maximizes the number of zero elements in filters by replacing small values with zero and pruning the filter that has the lowest number of zeros. In the conventional approach, the filters that have the highest number of zeros are generally pruned. As a result, through this zero-keep filter pruning, we can have the filters that have many zeros in a model. We compared the results of the proposed method with the random filter pruning and proved that our method shows better performance with many fewer non-zero elements with a marginal drop in accuracy. Finally, we discuss a possible multiplier architecture, zero-skip multiplier circuit, which skips the multiplications with zero to accelerate and reduce energy consumption.

Recent deep learning models succeed to achieve high accuracy and fast inference time, but they require high-performance computing resources because of a large number of parameters. However, not all systems have high-performance hardware. Sometimes, deep learning model needs to be run on edge devices such as IoT devices or smartphones. The edge devices have limited performance and energy consumption. On these devices, the amount of computation must be reduced. Pruning is one of the well-known approaches to solve this problem. In this work, we propose "zero-keep filter pruning" for an energy-efficient deep neural network. The proposed method maximizes the number of zero elements in filters by replacing small values with zero and pruning the filter that has the lowest number of zeros. In the conventional approach, the filters that have the highest number of zeros are generally pruned. As a result, through this zero-keep filter pruning, we can have the filters that have many zeros in a model. We compared the results of the proposed method with the random filter pruning and proved that our method shows better performance with much fewer non-zero elements with marginal accuracy drop. We also compare the number of remained filters with random and proposed pruning methods after pruning. Finally, we discuss a possible multiplier architecture, zero-skip multiplier circuit, which skips the multiplications with zero to accelerate and reduce energy consumption.

Case based reasoning is a knowledge discovery technique that uses similar past problems to solve current new problems. It has been applied to many tasks, including the prediction of temporal variables as well as learning techniques such as neural networks, genetic algorithms, decision trees, etc. This paper presents a geometric criterion for selecting similar cases that serve as an exemplar for the target. The proposed technique, called geometric Case Based Reasoning, uses a shape distance method that uses the number of sign changes of features for the target case, especially when extracting nearest neighbors. Thus, this method overcomes the limitation of conventional case-based reasoning in that it uses Euclidean distance and does not consider how nearest neighbors are similar to the target case in terms of changes between previous and current features in a time series. These concepts are investigated against the backdrop of a practical application involving the prediction of a stock market index. The results show that the proposed technique is significantly better than the random walk model at p &lt, 0.01. However, it was not significantly better than the conventional CBR model in the hit rate measure and did not surpass the conventional CBR in the mean absolute percentage error.

We investigate the operational stability of bottom-gate/top-contact-structured indium-oxide (In2O3) thin-film transistors (TFTs) in atmospheric air and under vacuum. Based on the thermogravimetric analysis of the In2O3 precursor solution, we utilize a thermal annealing process at 400 °C for 40 min to prepare the In2O3 films. The results of X-ray photoemission spectroscopy and field-emission scanning electron microscopy show that the electron is the majority carrier in the In2O3 semiconductor film prepared by a spin-coating method and that the film has a polycrystalline morphology with grain boundaries. The fabricated In2O3 TFTs operate in an n-type enhancement mode. When constant drain and gate voltages are applied, these TFTs in atmospheric air exhibit a more acute decay in the drain currents with time compared to that observed under vacuum. In the positive gate-bias stress experiments, a decrease in the field-effect mobility and a positive shift in the threshold voltage are invariably observed both in atmospheric air and under vacuum, but such characteristic variations are also found to be more pronounced for the atmospheric-air case. These results are explained in terms of the electron-trapping phenomenon at the grain boundaries in the In2O3 semiconductor, as well as the electrostatic interactions between electrons and polar water molecules.

In digital pathology, image similarity algorithms are used to find cancer in tissue cells from medical images. However, it is very difficult to apply image similarity algorithms used in general purpose system. Because in the medical field, accuracy and reliability must be perfect when looking for cancer cells by using image similarity techniques to pathology images. To cope with this problem, this paper proposes an efficient similar image search algorithm for digital pathology by applying leveling and tiling scheme on OpenSlide format. Furthermore, we apply image sync method to extract feature key points during image similarity processing. In the experiment, to prove the efficiency of the proposed system, we conduct several experiments including algorithm performance, algorithm accuracy and computation time. The experiments result shows that the proposed system efficiently retrieves similar cell images from pathology images.

Augmented Reality (AR)–based video telephony service can allow mobile users a better user experience (UX) since it allows participants to place and transmit augmented objects on video frames to a peer. However, there are quite a few AR-based mobile video communication models today, yet the existing models are limited and insufficient in supporting technical service such as real-time object detection, dynamic data selection, and discrimination between local data augmentation and remote data augmentation. This paper presents an enhanced AR–based mobile video telephony scheme, in which the object of interest can be dynamically combined with a video frame through real-time object detection, and users can immediately share their experience with their friend during a video call. In order to evaluate the effectiveness and feasibility of the proposed scheme, an application has been implemented on the mobile system and the computational time has been measured. Experimental results show that the proposed system can give customers better UX with small increase of computational time.

Image display time is defined as the time taken to load, decode, and render images on a screen. When users attempt to open images stored on a mobile phone, they can sometimes feel that it takes a lot of time to display the images. This paper analyzes the various causes of the delay and presents ideas to reduce the time and improve the customer’s experience. To evaluate these ideas, an enhanced image viewer application was implemented and evaluated quantitatively and qualitatively by comparing it to the existing viewer. Experimental results show that the proposed solutions can reduce image display time by more than 10 % for high-pixel images. In addition, it can reduce thumbnail construction times effectively.

So far, the data anonymization approaches based on k-anonymity and l-diversity has contributed much to privacy protection from record and attributes linkage attacks. However, the existing solutions are not efficient when applied to multimedia Big Data anonymization. This paper analyzes this problem in detail in terms of the processing time, memory space, and usability, and presents two schemes to overcome such inefficiency. The first one is to reduce the processing time and space by minimizing the temporary buffer usage during anonymization process. The second is to construct an early taxonomy during the database design. The idea behind this approach is that database designers should take preliminary actions for anonymization during the early stage of a database design to alleviate the burden placed on data publishers. To evaluate the effectiveness and feasibility of these schemes, specific application tools based on the proposed approaches were implemented and experiments were conducted.

In this paper, we design and implement a firmware for portable PCR devices that is controlled by a smart phone. The firmware has the host-local structure in which the firmware receives operation commands from the smart phone and sends operation results to the smart phone through Bluetooth communication. The firmware is designed to accommodate unstable wireless communication of Bluetooth. We implement a low-cost small PCR device with the proposed firmware on microchip PIC18F4550, and verify that the implemented PCR device significantly reduces volume size of existing commercial PCR devices with a similar performance.

In this paper, we present an enhanced storage system that supports Byte-index chunking algorithm. The storage system aims to provide efficient data deduplication with high performance and to be performed in rapid time. We describe the overall procedure of Byteindex chunking based storage system including read/write procedure and how the system works. The key idea of Byte-index chunking is to adapt fixed-size block chunk scheme which are distributed to “Index-table” by chunk’s both side boundary values. We have found that Byte-index chunking in storage system provides high performance compared with other chunking schemes. Experiments result shows that the storage system with Byte-index chunking compresses overall data with high deduplication capability and reduce the speed of file processing.

In this paper, we propose techniques that reduce the MAC overhead in IEEE 802.11-based wireless networks. As the channel bit rate increases, MAC overhead becomes a major source of throughput degradation. This is due to the fact that the MAC overhead is rate-independent; it is fixed regardless of the data rate. Rate-independent overhead can be reduced by dividing a high-rate wideband channel into multiple low-rate narrow channels, and send data on these narrow channels simultaneously using multiple or compound radios. In this environment with multiple narrow channels, we propose two techniques that further reduce MAC overhead and thus improve resource efficiency. First, channel contention is done only on one of the narrow channels, called the contention channel. The other channels are accessed based on the status of the contention channel, without contention. This technique significantly improves network throughput, but has a problem in which if a collision occurs on the contention channel, collisions can also occur on other channels. To avoid this phenomenon, we propose the second technique in which a node observing collision notifies the senders that cause collision, so that they do not access other channels simultaneously. Extensive simulation results show that the proposed techniques improve network throughput by a considerable amount.

These days, Windows Operating System is widely used and it is dominant than other commercial operating systems. Besides, most of the commercial software vendors provide a number of software mainly for Windows OS. Therefore, software similarity analysis systems for Windows OS are highly useful for software plagiarism detection and copyright protection. However, mostly the software similarity analysis systems and techniques work on source codes, not executables, even though the majority of commercial software barely makes public its source codes. In this paper, we propose a novel scheme for software plagiarism detection for Windows executables. Our technique exploits both extended basic blocks' opcode sequences and imported API functions information in executable files. We implemented our scheme and measured its performance through several kinds of experiments using code samples and different versions of commercial software.

Background/Objectives: Recently, storage systems and backup systems are popularly used and the number of duplicated data is increased drastically. To minimize data storage size and efficient use of network bandwidth, we proposed deduplication systems and file similarity measurement schemes with GPGPU scheme. The GPGPUs are applied to file similarity measurement for computation speedup. Methods/Statistical Analysis: To cope with the problem accompanying the parallelization of the measurement, we compare two implementations with shared memory and preprocessing. In addition, we propose an alternative to Rabin fingerprinting algorithm to lessen the computational burden of the algorithm to the GPUs. We compare the performance of the systems in time elapsed for several files. Findings: First, we found through experiments that the preprocessing was slightly faster than the shared memory scheme for the overlapped region of consecutive data segments which were assigned to different cores. This region should be shared by two cores for fingerprinting. By adapting GPGPU parallelization with the preprocessing technique for file similarity measurement, the proposed system outperformed the systems with a multi-core CPU. Also, it gets faster for the bigger file. In addition, we made the system three times faster by adapting an alternative to Rabin fingerprinting algorithm. It eliminates the computational burden of the algorithm and provides comparable results to the system with the latter. Improvements: The procedure will be beneficial to de-duplication system in determining file similarity and finding duplicated regions of two files. We achieved speedup in the measurement of file similarity by parallelization on GP-GPUs with two methods for overlaps of consecutive data segments and an alternative fingerprinting algorithm.

In traditional target-based data deduplication system, all of the files should be chunked and compared for reducing duplicated data blocks. One of the critical problem of this system arises as the number of files are increasing. The system suffers from computational delay for calculating hash value and processing metadata for handling each file. To overcome this problem, in this paper, we propose a novel data deduplication system using logical partition of storage system. The system applies data deduplication scheme to each logical partition not each file. Experiment result shows that the proposed system is more efficient compared with traditional deduplication scheme where the logical partition is full of files by 50% in terms of deduplication capacity and processing time.

Virtualization allows multiple commodity operating systems to share on a single physical machine. Resource allocation among virtual machines is a key to determine virtual machine performance. To satisfy time-sensitive task on a domain, hypervisor needs to observe the resource requirements and allocates proper amount of CPU resources in a timely manner. In this paper, we propose a realtime credit scheduler for latency sensitive application on virtual machines. The key idea is to register a time event in the Xen hypervisor. Experiment result shows that the proposed scheme is superior to Credit scheduler.

In this paper, we design and analyze an asynchronous pipelined FIFO called a micropipeline with the awareness of “place & route” (P&R) on an FPGA device. We use a commercially available 65nm Virtex-5 devices and design a high-speed implementation of the asynchronous four-phase micropipeline with considering its layout on the device. The layout of our design is modified manually to meet timing constraints and to accelerate the speed of circuits. The asynchronous FIFO implemented on the Virtex-5 device shows 452MHz throughput and 648ps per-stage latency at the simulation under the worst case operating condition and around 472MHz throughput is observed at the actual measurement on a real working chip at room temperature.

We used several machine learning techniques to predict the susceptibility to chronic hepatitis from single nucleotide polymorphism (SNP) data. These are integrated with several feature selection algorithms to identify a set of SNPs relevant to the disease. In addition, we apply a backtracking technique to a couple of feature selection algorithms, forward selection and backward elimination, and show that it is beneficial to find the better solutions by experiment. The experimental results show that the decision rule is able to distinguish chronic hepatitis from normal with a maximum accuracy of 73.20%, whereas the accuracy of the support vector machine is 67.53% and that of the decision tree is 72.68%. It is also shown that the decision tree and decision rule are potential tools to predict susceptibility to chronic hepatitis from SNP data.

For user convenience, we implement a user-friendly GUI program that plays a calibration tool for the coefficient variables of Steinhart-Hart equation. The proposed tool provides automatic modification of the coefficient variables used for the resistance-temperature measurement of NTC thermistor. In temperature measurement evaluation of PCR devices, the proposed tool yields 0.1C precision.

General purpose operating systems are Increasingly being used for serving time-sensitive applications. These applications require soft real-time characteristics from the kernel and from other system-level services. In this paper, we explore various operating systems techniques needed to support time-sensitive applications and describe the design of MUSMA(Multi-level Scheduler for Multimedia Application). MUSMA is a framework that combination of user-level top scheduler and kernel-level bottom scheduler. We develope MUSMA in linux environment and it`s performance is evaluated. Experiment result shows that it is possible to satisfy the constraints of multimedia in a general purpose operating system without significantly compromising the performance of non-realtime applications.

The addressing mode for data is performed by the software in yon Neumann-concept(SISD) computer a priori without hardware design of an address counter for operands. Therefore, in the addressing mode for the vector the corresponding variables as much as the number of the elements should be specified and used also in the software method. This is because not for operand but only for an instructions, quasi PC(program counter) is designed in hardware physically. A vector has a characteristic of a structural dimension. In this paper we propose to design a hardware unit physically external to the CPU for addressing only the elements of a vector unit with the structure and dimension. Because of the high speed performance for a vector processing it should be designed in the SIMD pipeline mechanics. The proposed mechanics is evaluated through a simulation. Our result shows to performance enhancement over CRAY architecture under the same hardware consideration(processing unit).

Long Term Evolution (LTE) networks that are composed of macrocells and femtocells can provide an efficient solution to not only extend coverage of macrocells but also deal with the growth of traffic within macrocells. LTE is now being considered to be a vital connectivity solution for the success of the Internet of Things (IoT) because it can provide broadband connectivity to the growing number of sensing and monitoring devices and even the wireless sensor networks (WSNs). However, it is still challenging to properly allocate radio frequency resources in the handover procedure between a macrocell and a femtocell. In this paper, we propose a new handover algorithm that increases the efficient utilization of a radio frequency resource and thereby maximizes the capacity of the overall LTE network, including the femtocells within network. The handover decision criteria take into account the strength of the received signal, the radio resource reuse, and the overall capacity of the network throughput. The performance of the proposed algorithm is verified through a simulation, and the simulation results indicate that the proposed handover algorithm improves the reusability of the cell bandwidth and increases the overall capacity of the network.

We design a compact PCR device based on the host-local structure. The PCR device receives operation commands from the smart phone and sends operation results to the smart phone through Bluetooth communication. The implemented PCR device significantly reduces volume size of existing commercial PCR devices with a similar performance.

This paper investigates subscription pricing, pay-per-use pricing, and two-part tariff from perspectives of a provider’s profit, consumer surplus, and social welfare, respectively. In particular, this paper extends previous studies on pricing schemes to two-part tariff in addition to subscription pricing and pay-per-use pricing. This paper shows that two-part tariff is more efficient than the other two pricing from a provider’s, consumers’ and social point of view.

In virtual machine environments, it is difficult to allocate CPU resource in a timely manner when lots of domains are in BOOST priority. In this paper, we present a virtual machine scheduling scheme based on realtime priority and efficiently deals with multi BOOST problem. We evaluate our prototype in terms of latency over diverse workloads. Our experiment result shows that the proposed realtime priority scheme effectively allocates CPU resources for low latency guest domain over varying workloads.

To support real time task in a virtual machine, hypervisor have to monitor system resource and provide exact resource for real time task. There are several monitor tools running on virtual machine, however, it has limitations on supporting detailed information for monitoring real time workload, performance degradation and difficult for controlling user interface. In this work, we proposed a resource monitoring and feedback tool for Xen virtual machine called XMM. XMM provides accurate resource monitoring in a hypervisor level without incurring much overhead. We also provide a GUI based control tool adjusting system resources based on user intervention, which can be useful in a multimedia system where user preference is important. XMM shows lots of information reflecting system behavior and it can give scheduling feedback for handling real time task. We believe XMM can be a useful tool for Xen virtual machine for various purposes.

Virtualization is a way of maximizing physical resources by allowing multiple domains run on a single physical computer. Virtualization technology provides efficient use of computer systems and reduces cost managing hardware, space and energy. Although virtualization provides various advantages, it struggles to support realtime application. In this paper, we propose a soft realtime scheduling framework for Xen Hypervisor. The key ideas of this paper are twofold. First, we adapt a compositional realtime scheduling framework that can guarantee realtime task to meet its deadline. Second, we consider a mixed realtime environment that allows realtime and non-realtime tasks to coexist.

A file synchronization program has to exchange lots of metadata information between a client and a server. As data have been growing rapidly in file synchronization systems, the metadata exchanging overhead is more critical now than before. In this paper, we propose a practical approach for minimizing metadata exchange overhead on the file synchronization system using variable-length chunking. The key idea is to minimize metadata by adapting file similarity information and generating only one hash key for several metadata. In the experiment, traditional file synchronization program sends 4.3 Mbyte over 500 MByte data file. However, the proposed algorithm only sends 37 Kbyte metadata.

In virtual machine environments, it is difficult to allocate CPU resource in a timely manner when lots of domains are in BOOST priority. In this paper, we present a virtual machine scheduling scheme based on VCPU shaping and efficiently deals with multi BOOST problem. We evaluate our prototype in terms of latency over diverse workloads. Our experiment result shows that the proposed realtime priority scheme effectively allocates CPU resources for low latency guest domain over varying workloads.

In this paper, we design and analyze two asynchronous circuits: 1) a simple micropipeline and 2) an asynchronous MIPS processor. We use a commercially available 65 nm Virtex-5 FPGA device for those two designs. The asynchronous FIFO implemented on the Virtex-5 device shows 452 MHz throughput at the simulation under the worst case operating condition. The micropipeline is extended to incorporate a conventional 5-stage pipelined MIPS datapath. Our asynchronous MIPS processor works successfully on the Virtex-5 FPGA device without a clock source, crystal oscillator, and 53.2 MHz throughput is measured.

Traditional file synchronization programs have difficulties in handling lots of metadata information when the file is transferred to the server. In this paper, we propose a practical approach for minimizing metadata exchanging overhead on the file synchronization system. The key idea of this paper is to adapt hash key compression technique that computes new hash key with several hash keys.

TAR archive format does not support efficient sub-file deletion and update. Therefore it has drawbacks for utilizing disk storage when it used as a backup tool. To address this problem, we propose an extended TAR file format called DTAR (Deduplication Tape Archive) which provides file-level deduplication and compression. The key idea of our work is to adapt user-level file system which supports efficient block insertion/deletion for DTAR format using block relocation. Experiment results show that the proposed system can reduce data storage space and computation time extremely.

In a virtual machine, there is much requirement for supporting realtime characteristic. To support realtime task in a virtual machine, we have to monitor and control resources including CPU, memory, I/O device. Recently several monitor tools are released running on virtual machines including Xenmon. However, it has limitations on supporting detailed information for monitoring realtime workloads. Furthermore, frequent access to trace buffer severely degrades the overall system performance. In this paper, we propose a light-weight monitor tools for Xen virtual machine, which provides workloads monitoring with high accuracy, and users can monitor lots of system information with low overhead. We also provide GUI tools for user friendly purpose.

Tar archive format does not support file deduplication scheme, therefore it has drawbacks for utilizing disk storage system efficiently. In this paper, we propose an extended TAR file format called DTAR (Deduplication Tape Archive) which provides file-level deduplication. Key idea of our work is to provide block aligned compressed file that can speed up file insertion and deletion for archiving. Furthermore, DTAR shows performance enhancement using file similarity scheme for distributing files into several storage nodes. Experiment results show that the proposed system can reduce data storage space efficiently and diminish network data traffic compared to general file transfer system.

In this paper, we propose Real-time Xen (RTXen) scheduling framework that supports time sensitive workloads on Xen hypervisor. Our primary goal is to provide real-time guarantee for real-time components that are running on the hypervisor. To accomplish this goal, first, we modified Xen credit scheduler to support real-time scheduling for real-time components. Second, we evaluated multi-core CPU capacity with/without hyper-thread capabilities. To guarantee timely execution of real-time task, we have to measure CPU resources for exact CPU capacity analysis. Our experiments demonstrate that the proposed system can support real-time workloads efficiently and it also shows fair share execution of non-real-time workloads.

The implementation approaches of data deduplication system divide into several modes including SBA(source-based approach), ILA(in-line approach) and PPA(post-process approach). Currently, most commercial systems are implemented and operated in an ILA and PPA approach, and some researchers have focused on the SBA approach. As data deduplication systems are widely used, to choose an appropriate mode considering operation environment becomes more and more important than ever. Because the overhead of each mode and resource usage wasn't fully studied, in some operating environments, the deduplication mode can lead to inefficiency and poor performance. In this study, we propose a data deduplication system supporting multi-mode. The proposed system can be operated in a mode that a user specifies during system operation, therefore, this system can be dynamically adjusted under consideration of system characteristics. In this paper, we operate the proposed system with the SBA, ILA and PPA mode, respectively, and we present the measurement results with a comparative analysis of the mode-specific performance and overhead.

In a mobile device, it is important to minimize battery consumption when we operate an application. In this paper, we propose an energy efficient file transfer scheme for mobile device by exploiting deduplication scheme. Generally, a mobile phone needs more power for WiFi work rather than CPU work, therefore we have to minimize WiFi work. We can minimize the number of packets by adapting deduplication which consumes lots of CPU power. Deduplication work consumes additional CPU power during data hashing, however overall energy consumption can be diminished. Experiment result shows that the proposed scheme significantly reduces battery consumption if a file has over 5% SRP(Space Reduction Percentage).

In virtual machine environments, it is difficult to allocate CPU resource to virtual machine efficiently because virtual machine lacks knowledge of each domains workload. Especially, realtime tasks in guest domain have to finish before their deadline, however, virtual machine scheduler is not aware of guest-level tasks and how much resources guest domain requires. In this paper, we present a virtual machine scheduling framework based on feedback mechanism. The proposed mechanism exploits various scheduling information from each domain. Xen scheduler controls the CPU allocation by increasing or decreasing CPU slices. We evaluate our prototype in terms of realtime task performance over diverse workload. Our experiment result shows that feedback mechanism effectively allocates CPU resources for guest domain in varying workloads.

In this paper we proposes a clustering backup system that exploits file fingerprint mechanism for multi-level deduplication of redundant data. Our approach differs from the traditional file server system. First, we avoid the data redundancy by block-level fingerprint. The proposed approach enables the efficient use of the storage capacity and network bandwidth without the transmission of the duplicate data block. Second, we applied clustering technology because data transfer and I/O time is reduced a fraction of a percent for each node. In this paper, we made several experiments to verify performance of our proposed method. Experiments result shows that storage space is used efficiently and the performance is noticeably improved.

The key issues in ubiquitous sensor networks for context awareness is location-tracking. The main goal of location-tracking is to find the position of a moving object. In this paper, we describes the design and implementation of ICLS (Intelligent Cricket-based Location System). Especially, a more accurate location-tracking system using a sensor fusing mechanism is proposed herein. The benefits of sensor fusion are that it can obtain more accurate, complete, or dependable results. ICLS employs an ltrasonic sensor for ranging by time-of-flight measurements and uses a magnetic sensor to detect obstacles. To yield a more accurate navigation, a novel pass finding algorithm is also proposed in this paper. The results of the experiment that was conducted in this study show that a mobile robot moves towards a destination with a high degree of accuracy.

A machine learning technique, decision tree, is used to predict the susceptibility to two liver diseases, chronic hepatitis and cirrhosis, from single nucleotide polymorphism(SNP) data. Also, it is used to identify a set of SNPs relevant to those diseases. The experimental results show that a decision tree is able to distinguish chronic hepatitis from normal with accuracy of 69.59% and cirrhosis from normal with accuracy of 76.72% and the C4.5 decision rule is with accuracy of 69.59% for chronic hepatitis and 79.31% for cirrhosis. The experimental results show that decision tree is a potential tool to predict the susceptibility to chronic hepatitis and cirrhosis from SNP data.

The embedded mobile phone based on L4 microkernel are known to be difficult to predict their behavioral correctness since L4 microkernel communicates with various components for their execution. That means its correctness is not easy to guarantee by using traditional design and testing methods. In this paper, we propose a formal approach to prove the correctness of L4 based mobile platforms.