An approach introduced to run Stochastic gradient descent in parallel without locks
Currently, there are no issues on this topic. Create one.