Rajasekar Karthik is a research scientist in Oak Ridge National Laboratory. He joined the Geographic Information Science and Technology (GIST) group at ORNL in June 2011. His areas of expertise include proficiency with a vast array of programming languages, concepts and technologies especially web-based software design and development, database programming, full life-cycle software development process, and search engine technologies.
With latest generation EO systems, Remote Sensing data is being generated in very large volumes across multiple formats, around the clock, such as LiDAR, SAR, and Hyperspectral (in global level and high resolution).\r\nAdvances in data acquisition techniques in Remote Sensing have resulted in introduction of multitude of operating payload having low GSD, higher revisit frequency, capable of operating round the clock and under any weather conditions. The datasets have become rich in higher spatial and spectral resolution, complex in structures and metadata, and diverse in applications areas.\r\nFor example, NASA EOSDIS, had 8292 unique data products, summing up to 9.1 Petabytes (PB) and growing at 6.4 Terabytes (TB) daily during the period from Oct 1, 2013 to Sept 30, 2014. These datasets were used by about 2 million users with an average end user distribution volume of 27.9 Terabyte each day .\r\nSuch recent trends in data-driven analysis necessities the need for Cyber-infrastructures capable of high-performance, scalable, or real-time computing that can efficiently handle “Big Data” workloads .\r\nIn this poster, our focus will be on “Data Storage”, and technologies being developed and used in other areas that can be applied in Remote Sensing Cyber-infrastructures.
Donking Kialanda Nsidiovova has graduated in Mathematic and Computer Science from University of Kinshasa. Currently, he is working as an assistant professor on data analysis at the same university and also in the Central Bank of Congo as an Responsible of IT Security and Planning on IT Department.
Neural network allow, from input information, determine wich cluster it belongs. The network must learn to produce such result. We use supervised learning in a network with three layers: input, hidden and output. MNN algorithm implemented by Microsoft SQL Server 2012 is used to assess our network. A low learning coefficient leads slow convergence and the high may produce oscillations. rnDiscriminant analysis separates clusters with linear combinations descriptors. Decision depends on rules to produce a minimum error in future. To estimate the performance of discriminant functions, we have used test sample and bootstrap validation in SPAD 5.0. rn