A team of ECE Florida graduate students has emerged as one of two finalists in the first ever Dell EMC AI challenge, a competition organized by Dell with the aim of supporting researchers in solving technical or business problems using AI technologies such as machine learning and deep learning.
The UF team consists of ECE graduate students Shreyas Gaadikere, Chao Jiang, Nirali Patel, and Akash Vasishta. The team worked under the mentorship of postdoctoral associate Dr. Dave Ojika and the supervision of Dr. Herman Lam, Site Director of the NSF Center for Space, High-Performance, and Resilient Computing (SHREC).
The UF team proposed to use heterogeneous computing (CPU+GPU+FPGA) to enable and accelerate machine learning for applications in high-energy physics. The proof-of-concept system will use GPUs to perform the training of convolutional neural network (CNN) models on CERN Large Hadron Collider datasets for physics particle identification – a pattern recognition problem that the team aims to tackle using deep-learning algorithms. The CNN models are then inputted into an inference engine that is equipped with FPGAs for acceleration.
Entry submissions for the challenge were reviewed by Dell based on technology innovation, socio-economic impact, and potential for commercialization. Of these submissions, two finalists were identified: the University of Florida and Cognitive Scale, a company providing an AI-based patient scheduling system. The winning team will be announced Nov. 12 at the SC18 Supercomputing Conference in Dallas, TX.
The finalists are provided exclusive remote access to a Dell EMC HPC and AI Innovation Lab cluster, powered by a Dell EMC PowerEdge C4140, to prove out their ideas. Thereafter, one overall winner will receive 200K core-hour access to a Dell EMC HPC and AI Innovation Lab cluster, a guest blog on the Dell EMC website, a spotlight in Dell EMC’s booth at the 2018 Supercomputing Conference, and a solution feature at a relevant industry conference or event.
Congratulations to the team! Stay tuned to ECE News and social media for updates.