Long training times are a deterrent to experimentation while building machine learning models. Intel® AI Builders member RocketML is on a mission to solve this problem. RocketML is developing a distributed machine learning platform tuned for higher order optimization methods. Their solution of full-batch limited-memory BFGS on distributed clusters is built with a view to enabling machine learning training on “any” data size simply by increasing the number of nodes in the cluster.
Banks, financial institutions and advertising technology companies not only deal with increasing data volume but also have to respond rapidly to changing data for risk management, price optimization and other factors. The RocketML platform on Intel® Xeon® Scalable clusters has the potential to significantly improve training times and lower cost for these customers. In the immunology segment where phoneme/genome data sets are growing at an astronomical rate, RocketML may allow researchers to make new discoveries regarding previously intractable problems. Lastly, RocketML can be an ideal platform for video analytics and subsequent machine learning. Video content analysis—the capability to analyze spatial and temporal events in videos—is increasingly important in various domains including healthcare, entertainment, transportation and home automation. RocketML plans to target all these segments with its distributed machine learning platform.
Come see RocketML at the One Intel Station at SC18 from November 11-14, 2018 in Dallas, Texas.