View All Items
(1 - 2 of 2)
- Title
- Applied Advanced Error Control Coding for General Purpose Representation and Association Machine Systems.
- Creator
-
Dai, Bowen, Wei, Lei, Lin, Mingjie, Rahnavard, Nazanin, Turgut, Damla, Sun, Qiyu, University of Central Florida
- Abstract / Description
-
General-Purpose Representation and Association Machine (GPRAM) is proposed to be focusing on computations in terms of variation and flexibility, rather than precision and speed. GPRAM system has a vague representation and has no predefined tasks. With several important lessons learned from error control coding, neuroscience and human visual system, we investigate several types of error control codes, including Hamming code and Low-Density Parity Check (LDPC) codes, and extend them to...
Show moreGeneral-Purpose Representation and Association Machine (GPRAM) is proposed to be focusing on computations in terms of variation and flexibility, rather than precision and speed. GPRAM system has a vague representation and has no predefined tasks. With several important lessons learned from error control coding, neuroscience and human visual system, we investigate several types of error control codes, including Hamming code and Low-Density Parity Check (LDPC) codes, and extend them to different directions.While in error control codes, solely XOR logic gate is used to connect different nodes. Inspired by bio-systems and Turbo codes, we suggest and study non-linear codes with expanded operations, such as codes including AND and OR gates which raises the problem of prior-probabilities mismatching. Prior discussions about critical challenges in designing codes and iterative decoding for non-equiprobable symbols may pave the way for a more comprehensive understanding of bio-signal processing. The limitation of XOR operation in iterative decoding with non-equiprobable symbols is described and can be potentially resolved by applying quasi-XOR operation and intermediate transformation layer. Constructing codes for non-equiprobable symbols with the former approach cannot satisfyingly perform with regarding to error correction capability. Probabilistic messages for sum-product algorithm using XOR, AND, and OR operations with non-equiprobable symbols are further computed. The primary motivation for the constructing codes is to establish the GPRAM system rather than to conduct error control coding per se. The GPRAM system is fundamentally developed by applying various operations with substantial over-complete basis. This system is capable of continuously achieving better and simpler approximations for complex tasks.The approaches of decoding LDPC codes with non-equiprobable binary symbols are discussed due to the aforementioned prior-probabilities mismatching problem. The traditional Tanner graph should be modified because of the distinction of message passing to information bits and to parity check bits from check nodes. In other words, the message passing along two directions are identical in conventional Tanner graph, while the message along the forward direction and backward direction are different in our case. A method of optimizing signal constellation is described, which is able to maximize the channel mutual information.A simple Image Processing Unit (IPU) structure is proposed for GPRAM system, to which images are inputted. The IPU consists of a randomly constructed LDPC code, an iterative decoder, a switch, and scaling and decision device. The quality of input images has been severely deteriorated for the purpose of mimicking visual information variability (VIV) experienced in human visual systems. The IPU is capable of (a) reliably recognizing digits from images of which quality is extremely inadequate; (b) achieving similar hyper-acuity performance comparing to human visual system; and (c) significantly improving the recognition rate with applying randomly constructed LDPC code, which is not specifically optimized for the tasks.
Show less - Date Issued
- 2016
- Identifier
- CFE0006449, ucf:51413
- Format
- Document (PDF)
- PURL
- http://purl.flvc.org/ucf/fd/CFE0006449
- Title
- Improving the performance of data-intensive computing on Cloud platforms.
- Creator
-
Dai, Wei, Bassiouni, Mostafa, Zou, Changchun, Wang, Jun, Lin, Mingjie, Bai, Yuanli, University of Central Florida
- Abstract / Description
-
Big Data such as Terabyte and Petabyte datasets are rapidly becoming the new norm for various organizations across a wide range of industries. The widespread data-intensive computing needs have inspired innovations in parallel and distributed computing, which has been the effective way to tackle massive computing workload for decades. One significant example is MapReduce, which is a programming model for expressing distributed computations on huge datasets, and an execution framework for data...
Show moreBig Data such as Terabyte and Petabyte datasets are rapidly becoming the new norm for various organizations across a wide range of industries. The widespread data-intensive computing needs have inspired innovations in parallel and distributed computing, which has been the effective way to tackle massive computing workload for decades. One significant example is MapReduce, which is a programming model for expressing distributed computations on huge datasets, and an execution framework for data-intensive computing on commodity clusters as well. Since it was originally proposed by Google, MapReduce has become the most popular technology for data-intensive computing. While Google owns its proprietary implementation of MapReduce, an open source implementation called Hadoop has gained wide adoption in the rest of the world. The combination of Hadoop and Cloud platforms has made data-intensive computing much more accessible and affordable than ever before.This dissertation addresses the performance issue of data-intensive computing on Cloud platforms from three different aspects: task assignment, replica placement, and straggler identification. Both task assignment and replica placement are subjects closely related to load balancing, which is one of the key issues that can significantly affect the performance of parallel and distributed applications. While task assignment schemes strive to balance data processing load among cluster nodes to achieve minimum job completion time, replica placement policies aim to assign block replicas to cluster nodes according to their processing capabilities to exploit data locality to the maximum extent. Straggler identification is also one of the crucial issues data-intensive computing has to deal with, as the overall performance of parallel and distributed applications is often determined by the node with the lowest performance. The results of extensive evaluation tests confirm that the schemes/policies proposed in this dissertation can improve the performance of data-intensive applications running on Cloud platforms.
Show less - Date Issued
- 2017
- Identifier
- CFE0006731, ucf:51896
- Format
- Document (PDF)
- PURL
- http://purl.flvc.org/ucf/fd/CFE0006731
