3 ms·
I programmed MPI about 15 years ago, during a summer at the MIT AI Lab. I implemented neural network feature creation for a backgammon agent ("Automated featur
by bravura 14y ago
I programmed MPI about 15 years ago, during a summer at the MIT AI Lab.
I implemented neural network feature creation for a backgammon agent ("Automated feature selection to maximize learning in artificial intelligence").
Nowadays, I mainly do parallel machine learning on machines with higher network latency. I haven't used MPI since.
- apawloski 14y agoWhat do you use now instead of MPI?
- pm90 14y agoHe talked about higher latency so I'm guessing just sockets
- bravura 14y agoIncorrect. With the sort of work I do nowadays (large-scale ML and NLP), I generally need very little synchronization, i.e. my tasks are usually embarrassingly parallel. I typically save final results in a centralized store (DB or NFS) and look at it there. I also use Hadoop where appropriate.