Robust classification techniques for connection pattern analysis with adaptive decision boundaries using CUDA

Social networking become an essential part of life in both personal as well as professional relationships. We, therefore implement the pattern analysis of connections in a social media network. For more effective and faster connection suggestions on social networking websites we need to analyze connected networks on the basis of clustering and adaptive decision boundary techniques. The connection suggestion and classification based on minimum computation time in social networks has become an area of major interest. These algorithms occupy a lot of host machine resources to execute the bunch of nested threads that result in the overall speed reduction. Therefore we have seen the GPU is become an essential part of standard internet browsers to speed up the applications. If we want to run a simple decision boundary classifier application in real-time, either a very high speed processor along with bulk of free memory is required or some other parallel computing techniques should be used. In this paper, we are trying to take benefit from NVIDIA's GPU to build a robust classification technique for pattern recognition with adaptive decision boundaries using CUDA to ensure high speed connection suggestions and better network connection analysis.