• 전자공학회논문지A
• /
• 제31A권6호
• /
• pp.159-168
• /
• 1994

Stochastic computation is adopted to reduce the silicon area of the multipliers in implementing neural network in VLSI. In addition to this advantage, the stochastic computation has inherent random errors which is required for implementing Boltzmann machine. This random noise is useful for the simulated annealing which is employed to achieve the global minimum for the Boltzmann Machine. In this paper, we propose a method to implement the Boltzmann machine with stochastic computation and discuss the addition problem in stochastic computation and its simulated annealing in detail. According to this analysis Boltzmann machine using stochastic computation is suitable for the pattern recognition/completion problems. We have verified these results through the simulations for XOR, full adder and digit recognition problems, which are typical of the pattern recognition/completion problems.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1998년도 춘계학술대회 학술발표 논문집
• /
• pp.53-56
• /
• 1998

본 논문에서의 공학적인 체계성을 갖고 초기 연결 가중치 및 임계치를 결정해 주면서, 학습까지 가능한 신경망을 제안한다. 기존의 오류 역전파 신경망을 적용할 때 경험에 의하여 은닉층 노드수를 결정하거나 임의의 실수 값으로 초기 연결 가중치 및 임계값을 주었을 때 자주 발생하는 학습 마비 현상을 피할 수 있고, Bose가 제안된 Voronoi 공간 분류에 의한 신경망 구성에서 학습이 불가능하다는 제안적인 단점을 보안하였다. 초기 가중치는 Voronoi 공간 분류가 이루어져 있다고 할 때 Bose가 제안한 초기 가중치 결정법을 개선하여 사용하고, Bose의 경우 신경망 노드가 Step function을 이용하여 정보를 전달하였으나 본 연구에서는 학습이 가능한 함수인 Sigmoid function을 이용하였다. 제안된 새로운 신경망의 성능 및 효율성을 비교하기 위하여 선형분리가 불가능한 XOR문제를 실험한 결과, 기존의 학습 가능한 EBP에서 허용오차 0.05 수준일 때 80%정도 학습마비 현상이 발생하였던 심각한 문제점을 보완할 수 있었고, 또한 학습 속도면에서 8~9배 정도 빠른 성능을 나타내었다.

• 대한전기학회논문지
• /
• 제40권2호
• /
• pp.224-233
• /
• 1991

The multilayer expanson of single layer NN (Neural Network) was needed to solve the linear seperability problem as shown by the classic example using the XOR function. The EBP (Error Back Propagation ) learning rule is often used in multilayer Neural Networks, but it is not without its faults: 1)D.Rimmelhart expanded the Delta Rule but there is a problem in obtaining Ca from the linear combination of the Weight matrix N between the hidden layer and the output layer and H, wich is the result of another linear combination between the input pattern and the Weight matrix M between the input layer and the hidden layer. 2) Even if using the difference between Ca and Da to adjust the values of the Weight matrix N between the hidden layer and the output layer may be valid is correct, but using the same value to adjust the Weight matrixd M between the input layer and the hidden layer is wrong. Recognitron III was proposed to solve these faults. According to simulation results, since Recognitron III does not learn the three layer NN itself, but divides it into several single layer NNs and learns these with learning patterns, the learning time is 32.5 to 72.2 time faster than EBP NN one. The number of patterns learned in a EBP NN with n input and output cells and n+1 hidden cells are 2**n, but n in Recognitron III of the same size. [5] In the case of pattern generalization, however, EBP NN is less than Recognitron III.

• International Journal of Computer Science & Network Security
• /
• 제24권5호
• /
• pp.95-102
• /
• 2024

In distributed computing, accessible encryption strategy over Auditable data is a hot research field. Be that as it may, most existing system on encoded look and auditable over outsourced cloud information and disregard customized seek goal. Distributed storage space get to manage is imperative for the security of given information, where information security is executed just for the encoded content. It is a smaller amount secure in light of the fact that the Intruder has been endeavored to separate the scrambled records or Information. To determine this issue we have actualize (CBC) figure piece fastening. It is tied in with adding XOR each plaintext piece to the figure content square that was already delivered. We propose a novel heterogeneous structure to evaluate the issue of single-point execution bottleneck and give a more proficient access control plot with a reviewing component. In the interim, in our plan, a CA (Central Authority) is acquainted with create mystery keys for authenticity confirmed clients. Not at all like other multi specialist get to control plots, each of the experts in our plan deals with the entire trait set independently. Keywords: Cloud storage, Access control, Auditing, CBC.

• 한국정보통신학회논문지
• /
• 제20권8호
• /
• pp.1438-1443
• /
• 2016

본 논문에서는 양방향 중계 채널에서 중계기에 다중안테나가 존재할 경우 물리계층 네트워크 코딩 기법의 성능분석을 수행하였다. 2개의 소스 노드는 하나의 안테나를 가진다고 가정하고 중계기에는 다중안테나가 설치되었다고 가정한다. 본 논문에서는 소스 노드와 중계노드에 각각 하나의 안테나가 존재하는 기존 시스템을 중계기에 다중안테나가 존재하는 환경으로 확장한다. 특히, 중계기에서 2개의 소스 노드들로부터 도착한 패킷을 복호하는 과정에서 각 패캣을 복호한 후 네트워크 부호화를 수행하는 개별복호기술과 소스 노드들로부터 도착한 2개의 패킷을 동시에 고려하여 직접복호하는 기술을 비교분석한다. 다중안테나기반의 중계기를 사용할 경우 물리계층 네트워크 코딩의 비트오류성능이 중계기의 안테나의 개수가 증가함에 따라 향상되는 것을 확인하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권1호
• /
• pp.227-247
• /
• 2018

For intermittently connected wireless sensor networks deployed in hash environments, sensor nodes may fail due to internal or external reasons at any time. In the process of data collection and recovery, we need to speed up as much as possible so that all the sensory data can be restored by accessing as few survivors as possible. In this paper a novel redundant data storage algorithm based on minimum spanning tree and quasi-randomized matrix-QRNCDS is proposed. QRNCDS disseminates k source data packets to n sensor nodes in the network (n>k) according to the minimum spanning tree traversal mechanism. Every node stores only one encoded data packet in its storage which is the XOR result of the received source data packets in accordance with the quasi-randomized matrix theory. The algorithm adopts the minimum spanning tree traversal rule to reduce the complexity of the traversal message of the source packets. In order to solve the problem that some source packets cannot be restored if the random matrix is not full column rank, the semi-randomized network coding method is used in QRNCDS. Each source node only needs to store its own source data packet, and the storage nodes choose to receive or not. In the decoding phase, Gaussian Elimination and Belief Propagation are combined to improve the probability and efficiency of data decoding. As a result, part of the source data can be recovered in the case of semi-random matrix without full column rank. The simulation results show that QRNCDS has lower energy consumption, higher data collection efficiency, higher decoding efficiency, smaller data storage redundancy and larger network fault tolerance.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제6권9호
• /
• pp.2442-2454
• /
• 2012

To ensure the security of wireless sensor networks, it is important to have a robust key management scheme. In this paper, we propose a Quorum-based key management scheme. A specific sensor, called as key distribution server (KDS), generates a key matrix and establishes a quorum system from the key matrix. The quorum system is a set system of subsets that the intersection of any two subsets is non-empty. In our scheme, each sensor is assigned a subset of the quorum system as its pre-distributed keys. Whenever any two sensors need a shared key, they exchange their IDs, and then each sensor by itself finds a common key from its assigned subset. A shared key is then generated by the two sensors individually based on the common key. By our scheme, no key is needed to be refreshed as a sensor leaves the network. Upon a sensor joining the network, the KDS broadcasts a message containing the joining sensor ID. After receiving the broadcast message, each sensor updates the key which is in common with the new joining one. Only XOR and hash operations are required to be executed during key update process, and each sensor needs to update one key only. Furthermore, if multiple sensors would like to have a secure group communication, the KDS broadcasts a message containing the partial information of a group key, and then each sensor in the group by itself is able to restore the group key by using the secret sharing technique without cooperating with other sensors in the group.

• International Journal of Contents
• /
• 제10권1호
• /
• pp.62-67
• /
• 2014

Usually, cloud storage systems are developed based on DFS (Distributed File System) for scalability and reliability reasons. DFSs are designed to improve throughput than IO response time, and therefore, they are appropriate for batch processing jobs. Recently, cloud storage systems have been used for update intensive applications such as OLTP and so on. However, in DFSs, in-place update operations are not carefully considered. Therefore, when updates are frequent, I/O performance of DFSs are degraded significantly. DFSs with RAID techniques have been proposed to improve their performance and reliability. Their performance degradation caused by frequent update operations can be more significant. In this paper, we propose an in-place update method for DFS RAID exploiting a differential logging technique. The proposed method reduces the I/O costs, network traffic and XOR operation costs for RAID. We demonstrate the efficiency of our proposed in-place update method through various experiments.

• 전자공학회논문지C
• /
• 제35C권2호
• /
• pp.56-64
• /
• 1998

Multi-layered perceptrons that are a nonlinear neural network model, have been widely used for various applications mainly thanks to good function approximation capability for nonlinear fuctions. However, for digital hardware implementation of the multi-layere perceptrons, the quantization scheme using "look-up tables (LUTs)" is commonly employed to handle nonlinear signmoid activation functions in the neworks, and thus requires large amount of storage to prevent unacceptable quantization errors. This paper is concerned with a new effective methodology for digital hardware implementation of multi-layered perceptrons, and proposes a "piecewise affine approximation" method in which input domain is divided into (small number of) sub-intervals and nonlinear sigmoid function is linearly approximated within each sub-interval. Using the proposed method, we develop an expression and an error backpropagation type learning algorithm for a multi-layered perceptron, and compare the performance with the quantization method through Monte Carlo simulations on XOR problems. Simulation results show that, in terms of learning convergece, the proposed method with a small number of sub-intervals significantly outperforms the quantization method with a very large storage requirement. We expect from these results that the proposed method can be utilized in digital system implementation to significantly reduce the storage requirement, quantization error, and learning time of the quantization method.quantization method.

• 한국컴퓨터정보학회논문지
• /
• 제10권1호
• /
• pp.27-34
• /
• 2005

현재의 침입차단, 침입탐지 등의 보안강화시스템은 공격자에 대하여 해당 트래픽만을 차단하는 수동적인 방어 시스템으로 실제 공격자에 대한 능동적인 대응이 부족하여 재공격 및 우회공격에 취약하다. 또한, 현재의 역추적 기술은 수작업을 통한 로그정보 수집 및 추적으로 인해 많은 시간과 인력이 필요하여 능동적 대응이 불가능하다. 본 논문에서는 현재의 인터넷 환경에 적용 가능하며 재공격 및 우회공격에 대응하기 위하여 IP헤더에 마크를 삽입하여 추적하는 TCP 연결 역추적 기법을 제안한다. 제안된 기법은 기존네트워크 구성요소의 수정이 불필요하고, 응답패킷에 XOR 연산 기법을 적용하여 마킹되는 정보의 양과 자원의 오버헤드를 줄일 수 있다.