This sprint was productive and gave us so much fun because we made progress.
Orbit v2 was the main objective of this sprint. Orbit v1 proved its performance, but it has a drawback, too small trade frequency. To overcome this problem, we have started orbit v2 since last Aug.
We have conducted thousands of tests with many ideas to pave the way. We have to admit that it was not a pleasant journey; most of the experiments did not pass the verification process. Few ideas could survive and are eligible to go to the next stage. The stringent evaluation process picked one idea, and we have planted the idea into Orbit v2.
We are planning to test Orbit v2 in the coming sprint.
Distributed Machine Learning Library
Preparation for Orbit v2 testing