Design and Evaluation of Classification Algorithms on CUDA Enabled GPU

Abstract

Recent expansion in GPUs(Graphics Processing Units) have empowered high performance computing for general purpose applications. Due to GPU's magnificent capacity, it has started working as co-processor of CPU to get high performance. CUDA programming model developed by Nvidia provides APIs to manipulate power of GPUs. Classification is a data mining technique which is used to assign items in order to get target class. To perfectly predict the target class for every case of data is main goal of classification. Classification techniques are widely used in many fields and has important applications in many domains. There is large size of images captured by satellites. Processing this images takes huge amount of time. So, some useful information is needed then we need to wait for huge amount of time as processing of these images consume a lot of time. Hence, we need parallelism to get information from images faster. I propose classification algorithm implemented parallely using CUDA for the satellite images and have compared the performance of the algorithm with and without using CUDA. The main aim of my research is to detect the vegetation layer from the satellite images with the use of GPU.