[FRIAM] Abduction

[uǝlƃ ☣]

Thanks for that paper.  It forced me to remember (and look up) the discussion in Pearl's book ("Causality" 2000) about the Markov assumption and latent structure reduction.  Part of my reaction to John's statement about trying to find a time series that cannot be generated by a sequential machine was a result of Pearl's discussion.  The question I'm now worried about is the facility/frequency with which cyclic graphs can be "simulated" by DAGs (which is why I implied that everywhere we think there might be a convergence to something "real" would require a monotonic parameter).


On 12/31/18 12:35 PM, Frank Wimberly wrote:
> Try this link.  Now I remember that Thomas Richardson first described the
> algorithm and Danks and I implemented it.
> 
> http://scholar.google.com/scholar_url?url=https://arxiv.org/pdf/1302.3599&hl=en&sa=X&scisig=AAGBfm23iOsgxDPx5eHVIU1aXYbP1yc_ZA&nossl=1&oi=scholarr


-- 
☣ uǝlƃ