search

ck37 / varimpact

Variable importance through targeted causal inference, with Alan Hubbard
57 stars 13 forks source link

Extremely slow and crushes my laptop #12

Closed Shafi2016 closed 4 years ago

Shafi2016 commented 4 years ago

I am using Legion Y740 (15") Gaming Laptop. This is considered as one of the powerful laptops. I have 461 features. I let codes run for three days; it still not complete the process then I stopped R forcefully. I tried other options too "parallel via multicore" and "parallel via SNOW." I let them run overnight then in the morning as I wanted to use different things; it just crushed, and the laptop went to an automatic restart option. Whenever I use parallel via SNOW , it captures the computer like a ghost. When I open anything, it crushes, and laptop restarts.

ck37 commented 4 years ago

461 features is a large number - you're running 461 observational studies effectively.

I recommend doing a subset first, such as 5 or 10, by setting the A_names argument to a smaller vector of variable names. That will then allow you to estimate the expected runtime.

Runtime depends on the SuperLearner library, the number of observations, number of variables analyzed, the size of the adjustment set (see adjust_cutoff setting), and other settings.