Check out our new paper on predictive learning on sign-valued hidden Markov trees (to be submitted shortly to JMLR)! This paper not only extends prior work on noiseless tree structure learning (see related papers by Bresler & Karzand (MIT EECS), cited in the paper, of course), but also develops new theoretical insights and proof techniques.
This is joint work with Anand Sarwate and Kostas Nikolakakis from ECE Rutgers.
Feel free to download the paper by clicking here.
There will also be an announcement on arXiv shortly.