Closed espinoj closed 6 years ago
@espinoj Mimbuild is done.
@espinoj I'm working on LiNGAM.
I have re-added LiNGAM as an algorithm in the interface (by adding the appropriate annotation to the LiNGAM wrapper. I have cleaned up the implementation of LiNGAM (in the search package). I implement precisely algorithm A from http://www.jmlr.org/papers/volume7/shimizu06a/shimizu06a.pdf
p. 2007, using our translation of FastICA from like a decade ago. For the 'pruning' step I use FGES with the causal order found by Algorithm A.
This could be improved. The above paper has suggestions for further refining the algorithm. We could try another implementation of ICA. But the above is correct.
@cg09
@cg09 Clark, once this is is merged you can try it.
@kvb2univpitt Can you merge it?
this is merged in development for @cg09 to try
Mimbuild This is absolutely essential Lingam Good for very large samples