Scope

Machine learning is one of the key technologies managing high-dimensional data continuously accumulating every day in various disciplines of natural sciences. This technology actually has a close relationship to statistical physics, and several statistical-mechanical studies on some machine-learning issues, such as evaluation of the ideal performance of designed systems and modeling biological systems, have been conducted for decades, mainly based on techniques of statistical mechanics of random-spin systems. This trend is continuing and expanding, and now starts to incorporate methods of non-equilibrium statistical mechanics which recently has a significant advance.
On this opportunity, this workshop provides active researchers on the related issues with an occasion to exchange and discuss their ideas, to deepen this very interdisciplinary area of science and more.



Program

July 21st 22nd 23rd 24th
Physics approach to general inference problems Deep learning and the ingredients Recent advance in inference algorithms Inference and model construction of biological phenomena
09:45-10:00
Opening
09:45-11:00
Salakhutdinov
09:45-11:00
Decelle
09:45-11:00
Ferrari
10:00-11:15
Kappen
11:00-11:15
coffee break
11:00-11:15
coffee break
11:00-11:15
coffee break
11:15-11:30
coffee break
11:15-12:30
Sohl-Dickstein
11:15-12:30
Ricci-Tersenghi
11:15-12:30
Mora
11:30-12:00
Miyama
12:00-12:30
Tokuda
12:30-13:45
Lunch
12:30-13:45
Lunch
12:30-13:45
Lunch
12:30-13:45
Lunch
13:45-14:30
Poster Preview and Introduction
13:45-14:15
Karakida
13:45-18:00
Free discussion
13:45-14:15
Ogawa
14:15-15:15
Sohl-Dickstein
14:15-14:45
Marinari
14:30-18:00
Free discussion
14:45-18:00
Free discussion
15:15-18:00
Free discussion




July 21 (Tue.)
09:45-10:00 Opening
10:00-11:15 Bert (HJ) Kappen "Adaptive importance sampling for control and inference" download (pdf)
11:15-11:30 coffee break
11:30-12:00 Masamichi J Miyama "Sparse modeling approach to STM data analysis"
12:00-12:30 Satoru Tokuda "Phase transition phenomena in Bayesian statistics"
12:30-13:45 Lunch
13:45-14:30 Poster Preview and Introduction
The poster presenters who want to show abstract of their researches to participants in the first week may use this time to introduce their contents in a few minutes. If you want to use several slides, please send them to the head of the organizer in advance, mohzeki@i.kyoto-u.ac.jp.
14:30-18:00 Free discussion

July 22 (Wed.)
09:45-11:00 Ruslan Salakhutdinov "Recent Advances in Deep Learning" download (pdf)
11:00-11:15 coffee break
11:15-12:30 Jascha Sohl-Dickstein "Deep Unsupervised Learning using Nonequilibrium Thermodynamics" download (pdf)
12:30-13:45 Lunch
13:45-14:15 Ryo Karakida "Equilibrium Analysis of Representation Learning in Gaussian RBMs"
14:15-15:15 Jascha Sohl-Dickstein "Hamiltonian Monte Carlo"
15:15-18:00 Free discussion

July 23 (Thu.)
09:45-11:00 Aurelien Decelle "Improving inference in the Ising inverse problem: Mean-field aspect and Decimation" download (pdf)
11:00-11:15 coffee break
11:15-12:30 Federico Ricci-Tersenghi "Improved mean-field approximations for inferring marginals and model parameters" download (pdf)
12:30-13:45 Lunch
13:45-18:00 Free discussion

July 24 (Fri.)
09:45-11:00 Ulisse Ferrari "Approximate Newton Methond for the Inverse Ising Problem" download (pdf)
11:00-11:15 coffee break
11:15-12:30 Thierry Mora "Inferring the critical dynamics of a neural population" download (pdf)
12:30-13:45 Lunch
13:45-14:15 Noriaki Ogawa "Pattern formation in Fish Retina from Physical Model Approach"
14:15-14:45 Enzo Marinari "Cross correlations of the American baby names"
14:45-18:00 Free discussion


Poster Session

Poster List (Machine Learning)


Invited Speakers

Aurelien Decelle (Rome La Sapienza)

Ulisse Ferrari (Ecole normale supérieure)

Bert (HJ) Kappen (Radboud)

Thierry Mora (École normale supérieure)

Ruslan Salakhutdinov (Toronto)

Jascha Sohl-Dickstein (Stanford)

Federico Ricci-Tersenghi (Rome La Sapienza)




Organizers

Masayuki Ohzeki (Kyoto)

Tomoyuki Obuchi (Tokyo Tech)

Muneki Yasuda (Yamagata)


Designed by CSS.Design Sample