Linear regression makes a lot of sense since it can be done streaming in a single pass. We have it as an aggregator in Hail: https://hail.is/docs/devel/aggregators.html#hail.expr.aggreg...

I wonder how they're doing logistic regression. It doesn't have a closed form and normally solved with an iterative, multi-pass optimization. edit: Read more closely. They're using (non-stochastic) batch gradient descent.