hoaphumanoid

thanks for the comment, would this definition be clearer: "Recall at k is the proportion of relevant items found in the top-k recommendations"?

hey one of the authors here. We are planning to create some notebooks on recommendation about recommendation algos :-)

Sometimes the notebook doesn't render, here there is a different option: https://nbviewer.jupyter.org/github/miguelgfierro/sciblog_su...

Miguel Fierro from Microsoft here: Joseph Sirosh session will be broadcasted live at 2.15pm ET in this link: https://myignite.microsoft.com/sessions/56555?source=session...

Yes, it was initially forked from an internal boosted trees library from Microsoft. The guys who created it are Microsoft Research Asia

The LightGBM implementation on GPU is based on this paper: https://arxiv.org/abs/1706.08359 they use several smart techniques to make the computation faster. One is how they create histograms of features that are computed in parallel in the GPU

In deep learning, the algorithm that is used, back propagation, is basically a chained matrix multiplication, which is good for the GPU structure. However, for boosted trees the optimization problem is a little bit different, there is a sorting and then a computation of a gradient and hessian. In this case, it is not as efficient for GPU than deep learning.