Advanced Statistics meets Machine Learning III Workshop

13 Nov 2019, 08:45 → 15 Nov 2019, 15:00 US/Central

Room 102 (Argonne - Bldg 201)

The "Advanced Statistics meets Machine Learning III" workshop will cover a number of topics in modern machine learning and advanced statistics, with a focus on cosmological applications. Topics to be discussed includes, but is not limited to, Bayesian inference, inverse problems, uncertainty quantification, likelihood free methods, noisy and low-dimensional data, unsupervised deep learning, dimensionality reduction, adversarial techniques, generative models and scaling of statistical methods to HPC. This workshop aims to bring together statistics and machine learning communities to exchange ideas, and to help solving problems related to cosmology. Apart from regular talks and discussions, we plan to have hands-on sessions where problem statements and data will be provided beforehand. We strongly encourage participants to provide their own interesting problems and relevant data to be worked on during the hands-on sessions. Short presentations related to the problem statements, data or solutions would be ideal.

Site Map Argonne_Map_(with_closure).pdf

Wednesday, 13 November

08:45 → 09:00 Registration

09:00 → 09:30 Introduction and Welcome

Salman Habib

09:30 → 10:00 FlowPM: Distributed Particle-Mesh N-body Simulation in TensorFlow - Presentation

Francois Lanusse

10:00 → 10:30 Open Discussion & Coffee Break

10:30 → 11:00 Statistical Computing on Large Parallel Architectures - Presentation

George Ostrouchov

11:00 → 11:30 Robust learning with implicit residual networks

Viktor Reshniak

11:30 → 12:00 Intro to toy problems

12:00 → 13:00 Open Discussions & Lunch

13:00 → 14:30 Hands-On Sessions & Group Discussions

14:30 → 15:00 Open Discussion & Coffee Break

15:00 → 16:30 Hands-On Sessions & Group Discussions

Thursday, 14 November

09:00 → 09:30 Learning to Solve Inverse Problems in Imaging and Sample Complexity - Presentation

Rebecca Willett

09:30 → 10:00 Tackling the limitations of conventional reduced-order models for nonlinear dynamical systems using convolutional autoencoders - Presentation

Romit Maulik

10:00 → 10:30 Open Discussion & Coffee Break

10:30 → 11:00 Uncertainty Quantification and Bayesian Deep Learning - Presentation

Elise Jennings

11:00 → 11:30 Efficient computation of extreme excursion probabilities for dynamical systems - Presentation

Vishwas Rao

11:30 → 12:00 Simulation-Based Disaggregation of LIBS Spectra

Earl Lawrence

12:00 → 13:00 Open Discussion & Lunch

13:00 → 14:30 Hands-On Sessions & Group Discussions

14:30 → 15:00 Open Discussion & Coffee Break

15:00 → 16:30 Hands-On Sessions & Group Discussions

17:30 → 19:30 Workshop Dinner

Friday, 15 November

09:00 → 10:00 Hands-on Session

10:00 → 12:00 Final Presentations

12:00 → 13:00 Open Discussion & Lunch

13:00 → 15:00 Wrap-Up Discussions & Conclusions