|
Guido Francisco Montúfar
Max Planck Institute for Mathematics in the Sciences
Inselstrasse 22 04103 Leipzig Germany
|
|
mis.mpg.de|
CV:
PhD scholarship, MPI MIS Leipzig
(01/2009 - present) International Max Planck Research School Mathematics in the Sciences
Research assistant at the Institute for Theoretical Physics, TU-Berlin
(03/2008 - 12/2008) Graduiertenkolleg SFB 787 Halbleiter-Nanophotonik, TP B1 Theorie photonischer Bauelemente am quantenoptischen Limit
Diplom Physiker, TU-Berlin
(12/2008)
Diplom Mathematiker, TU-Berlin
(08/2007) |
Research interests:
Information Theory
Mixture Models
Neural Networks
Information Geometry
|
|
Publications
|
Refinements of Universal Approximation Results for Restricted Boltzmann Machines and Deep Belief Networks,
G.M., Nihat Ay, [Neural Computation 23:5, 2011], [Preprint arXiv 1005.1593] 
Universally Typical Sets for Ergodic Sources of Multidimensional Data,
with T. Krüger, R. Seiler, and R. Siegmund-Schultze, [MPI MIS Preprint 20/2011], [Preprint arXiv 1105.0393]
Expressive Power and Approximation Errors of Restricted Boltzmann Machines,
G.M., J. Rauh, N. Ay, to appear in: NIPS 2011, [MPI MIS Preprint 27/2011]
Selection Criteria for Neuromanifolds of Stochastic Dynamics,
Nihat Ay, G.M., Johannes Rauh, to appear in: Advances in Cognitive Neurodynamics (III), [MPI MIS Preprint 15/2011] |
|
Conference Contributions
|
Mixture Models and Representational Power of RBM's, DBN's and DBM's,
[pdf]
G.M., NIPS 2010 - Deep Learning and Unsupervised Feature Learning Workshop.
Faces of the probability simplex contained in the closure of an exponential family and minimal mixture representations,
G.M., Information Geometry and its Applications III, Leipzig Germany 2010.
Theory of transport and photon-statistics in a biased nanostructure,
G.M., M. Richter, T. Brandes, and A. Knorr, Nano-Optoelectronics Workshop, Tokyo and Shonan Village Japan 2008. i-NOW 2008. |
|
Theses
|
Theory of transport and photon-statistics in a biased nanostructure,
G.M., 12.2008,
German Diplom in Physics, Institute for Theoretical Physics TU-Berlin, Advisor: A. Knorr and T. Brandes.
Q-Sanov Theorem for d ≥ 2 ,
G.M., 08.2007,
German Diplom in Mathematics, Institute for Mathematics TU-Berlin, Advisor: R. Seiler and J.-D. Deuschel. |
|
Talks
|
Geometry of Restricted Boltzmann Machines Towards Geometry of Deep Belief Networks,
31.08.2011, RIKEN Workshop on Information Geometry, RIKEN BSI Japan.
Geometry and Approximation Errors of Restricted Boltzmann Machines,
02.09.2011, The 5th Statistical Machine Learning Seminar at the Institute of Statistical Mathematics, Tachikawa, Tokyo.
Selection Criteria for Neuromanifolds of Stochastic Dynamics,
12.06.2011,
The 3rd International Conference on Cognitive Neurodynamics, Niseko Village, Hokkaido, Japan
On Exponential Families and the Expressive Power of Related Generative Models,
14.03.2011, Laboratoire d'Informatique des Systèmes Adaptatifs - Université de Montréal , [Abstract]
Mixtures from Exponential Families,
02.03.2011 15:00, Leon-Lichtenstein-Seminarraum MPI MIS Leipzig.
Universal approximation results for Restricted Boltzmann Machines and Deep Belief Networks,
16.02.2011 15:00, Leon-Lichtenstein-Seminarraum MPI MIS Leipzig.
Necessary conditions for RBM universal approximators,
18.01.2011, Mariánská Meeting of the Department of Decision-Making Theory of the Institute of Information Theory and Automation UTIA, Mariánská, Czech Republic
Information Geometry of Mean-Field Methods,
Fall School on Statistical Mechanics and 5th anual PhD Student Conference in Probability, MPI MIS 09.2009.
Quantum-Sanov-Theorem for correlated States in multidimensional Grids,
Dies Mathematicus, TU-Berlin 02.2008.
Quanten-Sanov-Theorem im mehrdimensionallen Fall,
Workshop on Complexity and Information Theory, MPI MIS 10.2007. |