[Lisa_seminaires] Fwd: Visitor and Talk (Monday 2:30) about learning finite automata

Aaron Courville aaron.courville at gmail.com
Lun 17 Sep 06:27:29 EDT 2012


Hi,

There is a talk today at McGill on "Spectral Methods for Learning
Finite State Machines", details below. It seems like it could be
really interesting.

Cheers,
Aaron



---------- Forwarded message ----------
From: Doina Precup <dprecup at cs.mcgill.ca>
Date: Sat, Sep 15, 2012 at 9:26 PM
Subject: Fwd: Visitor and Talk (Monday 2:30) about learning finite automata
To: Aaron Courville <courvila at iro.umontreal.ca>


Hi Aaron,

Please post this to the LISA mailing list.  BTW, if I should be
sending these announcements to someone else, let me know

Thank you,
Doina


Begin forwarded message:

> From: Doina Precup <dprecup at cs.mcgill.ca>
> Date: September 15, 2012 9:24:26 PM EDT
> To: labrl at cs.mcgill.ca
> Cc: Denis Therien <denis at cs.mcgill.ca>
> Subject: Visitor and Talk (Monday 2:30) about learning finite automata
>
> Hi everyone,
>
> Borja Balle Pigem will be visiting Monday-Thursday next week.  He has been doing very interesting work on spectral methods for learning automata, related to many of our interests in learning with partial observability.  He will give a talk this Monday at 2:30, in MC 103:
>
> Title: Spectral Methods for Learning Finite State Machines
>
> Abstract: A spectral method for learning Hidden Markov Models was
> presented by Hsu, Kakade, Zhang in COLT 2009. Following this seminal
> work their method has been extended to many other probabilistic
> models, including: Predictive State Representations, Kernelized Hidden
> Markov Models, Latent Tree Graphical Models, High-dimensional Mixture
> Models, Finite State Transducers, and Probabilistic Context-Free
> Grammars. The main feature of the method is that it provides a
> local-minima-free alternative to EM which is fast, easy to implement,
> and amenable to theoretical analysis.
>
> In this talk I will give a general and elegant description of this
> spectral method in a framework that encompasses some of the results
> mentioned above. In particular, those related to models that can be
> defined using Finite States Machines. I will also discuss open
> problems and opportunities for applications.
>
> (based on joint work with X. Carreras, M. Mohri, and A. Quattoni)
>
> For more information: http://www.lsi.upc.edu/~bballe/
>
> Let me know when you would be available to meet with him during his stay.
>
> Best,
> Doina



-- 
Aaron C. Courville
Département d’Informatique et
de recherche opérationnelle
Université de Montréal
email:Aaron.Courville at gmail.com


Plus d'informations sur la liste de diffusion Lisa_seminaires