Advanced Statistical Methods Meet Machine Learning

24 Sep 2018, 09:00 → 25 Sep 2018, 17:00 US/Central

D172 (Argonne - Bldg 241)

Salman Habib (Argonne National Laboratory)

The workshop on "Advanced Statistical Methods Meet Machine Learning" will cover a number of topics in Bayesian statistics, non-parametric regression, model selection and object classification in the presence of noisy and low-dimensional data, unsupervised learning contrasting classical methods and deep learning, parameter sampling in high-dimensional spaces, applications of optimization and iterative methods, and the use of high-performance computing in modern data-intensive science, with a focus on cosmological applications and statistical inverse problems.

Monday, 24 September

1 SciDAC-4 project overview

2 Emulation applications/discussion 1

"Classical" methods

10:30 Coffee Break / Informal Discussion

3 Emulation applications/discussion 2

AI Methods

12:00 Working Lunch

4 Statistical inverse problems 1

"Classical" Methods

5 Statistical inverse problems 2

AI Methods

15:00 Coffee Break / Informal Discussions

6 Discussion on model based vs. model-free methods

7 Summary of Day 1 -- Action Items

Tuesday, 25 September

8 Open problem priorities

9 Optimization applications in emulation and inverse problems

10:30 Coffee Break / Informal Discussion

10 HPC and ML applications/discussion 1

12:00 Working Lunch

11 HPC and ML applications/discussion 2

12 Discussion on new science cases 1

15:00 Coffee Break / Informal Discussion

13 Discussion on new science cases 2

16:30 Summary of Meeting