Data Challenge

This page is a complement to our challenge’s page at

Relating structure to long-time dynamics in supercooled liquids is a long-standing problem. Several ideas have been put forward and tested in the physics literature. The idea of the challenge is to address this problem from a different angle, using a data-driven approach.
The use of the ML approach for glasses was introduced by the team of Andrea J. Liu, in particular in collaboration with Ekin Dogus Cubuk and Samuel S. Schoenholz. See refs. below.

Here is our python code to handle Periodic Boundary conditions (with comments), which is also included in the input data .zip file.

The general review on glasses we mentioned in the slides is cited below. If you have trouble getting it (or others), please ask us (also, there is sci-hub).

« Prizes » for a high accuracy algorithm:
– Very likely publication in a high-profile physics journal
– Participation to the 2019 annual meeting of the Simons collaboration « Cracking the Glass Problem » in New York city
– Discussions and interactions with the Parisian members of the Simons collaboration « Cracking the Glass Problem »

– Nature Physics 2015 –  Schoenholz, Cubuk, Sussman,, A structural approach to relaxation in glassy liquids. Nat. Phys., 12(5), 469–471.
– J chem Phys B 2016 – Cubuk, Schoenholz, et. al., Structural Properties of Defects in Glassy Liquids. J chem Phys B, 120(26), 6139–6146.
– PNAS 2016 – Schoenholz, Cubuk, et. al., The Relationship Between Local Structure and Relaxation in Out-of-Equilibrium Glassy Systems. PNAS, 114(2), 263–267
– PRL 2015 – Cubuk, Schoenholz, et. al.,  Identifying structural flow defects in disordered solids using machine-learning methods. PRL, 114(10), 108001.
– RevModPhys 2011 – Berthier, Biroli,  Theoretical perspective on the glass transition and amorphous materials, Review of Modern Physics, 83, 587,

Les commentaires sont fermés.