Unsupervised Dynamic Sensor for Real-Time Segmentation of Big Data
Tackling the challenge of Big Data processing is focused today on two levels:
Increase computing power of processors and cloud platforms
Improving parallel processing algorithms for optimal utilization of the above hardware
The proposed novel solution tackles those limitations by reducing problem’ dimensions. To reduce the dimensions of unsupervised segmentation problem, it uses “dynamic samples” from within the population that are stored in small memory buffers. Samples-buffers whose population varies over time remain small in size, and thus enable real-time processing (i.e. unsupervised segmentation) of big data via “standard” hardware and algorithms.