NEWS

Yang Shi at the San Mateo office was invited to give a special lecture at the 2020 Siam Conference on Parallel Processing for Scientific Computing (PP20). The 3-day, bi-annual conference was held in Seattle, Washington this year. The conference is renowned for promoting parallel scientific computing and algorithms and as an essential forum for computer science, mathematics, and engineering communities.

Dr. Shi’s lecture was given at a mini-symposium titled, “High-Performance Tensor Computation and Applications.” She discussed, in detail, efficient tensor operations via sketching which builds upon her dissertation on tensor operations via compression and parallel computation. (Yang Shi joined Rakuten Americas shortly after completing her dissertation last spring at University of California, Irvine.)

One of the many benefits of attending a conference that directly involves one’s research is networking with like-minded academics and professionals in the field and—if you’re lucky—meeting those whose exemplary work has helped you find your own path to success. Dr. Shi was able to meet with Tamara Kolda, an applied mathematician and Distinguished Member of Technical Staff at Sandia National Laboratories.  Not only were her publications in the tensor algebra field essential readings during her course, but Kolda’s Matlab Tensortool box continues to be a useful software that she uses regularly in her ongoing research.

Finally, RIT is continuing to grow its presence in the US and many researchers are intrigued and interested in our activities and achievements. Dr. Shi shared innumerable conversations and moments with researchers and industry professionals to cultivate her identity and RIT. Many of the interactions encompassed her ongoing research to solve optimization problems in machine learning. This includes a cross-modal retrieval task which aims to retrieve relevant contents from different sources (text, images, etc.) and connect computer vision, natural language, and e-commerce.  She has also recently investigated efficient retrieval algorithms using hashing and quantization.

We would like to thank the Siam Conference on Parallel Processing for Scientific Computing, including SIAM Activity Group on Supercomputing as sponsor.