Using Blue Waters for Hardware Acceleration of Deep Learning for Big Data Image Analytics

Tao Xie, University of Illinois at Urbana-Champaign

Usage Details

The project aims at using the Blue Waters platform for hardware acceleration of deep learning for big data image analytics, to enable near real-time learning (feature extraction, modeling construction) and prediction for big data analytics, using image data as a design driver. This proposal is to leverage the Blue Waters platform to accelerate machine learning algorithms by several orders of magnitude than the present state-of-the-art, to achieve real-time performance for extracting meaningful information from images. The proposed effort requires a cross-layer holistic solutions with two faculty members each with distinguished areas of expertise, covering two main technical areas: (1) software engineering for machine learning algorithms (Tao Xie from Computer Science department at Illinois), (2) computer architecture (Yuan Xie from ECE department at UCSB).