Flatiron Internal Conference 2022: Flatiron-wide Algorithms and Mathematics

Oct 27, 2022, 8:15 AM → Oct 28, 2022, 9:00 PM America/New_York

2nd Floor: Ingrid Daubechies Auditorium (162 5th Avenue)

FWAM is a two-day internal conference which has the goal of introducing and reviewing numerical tools of broad and significant usefulness to Flatiron researchers across centers. The conference this year will be organized into 4.5 broad topics, with a combination of "tutorial/overview" talks offering practical and accessible introductions, and "case-study" talks showcasing research applications.

 

The broad topics are

 

(1a) Efficient representation of low-dimensional functions

(1b) Efficient representation of high-dimensional functions (tensor network methods)

(2) Ordinary differential equations

(3) High performance computing

(4) Practical tools for machine learning and data science

 

FWAM is planned as an in-person event. Lunch will be provided, and there will be a closing reception. Please register using this form.

 

Update: recordings of certain FWAM22 talks are now available here. Slides for some of the talks can also be found by clicking on the talk in the Contribution List tab, clicking on a talk, and looking under "Presentation Materials".

Demo for Manas Rachh's talk

Thursday, October 27

9:30 AM Welcome

Efficient representation of low-dimensional functions: Interpolation and integration in low dimensions (tutorial/overview)

Convener: Manas Rachh (Flatiron)

Efficient representation of low-dimensional functions: Maximum entropy closure for kinetic theories of complex fluids (case study)

Convener: Scott Weady (CCB)

Coffee break

Efficient representation of high-dimensional functions (tensor network methods): Compressing functions with tensor networks: Applications to PDEs and DFTs (tutorial/overview)

Conveners: Matt Fishman (CCQ), Miles Stoudenmire (CCQ)

12:30 PM Lunch

Efficient representation of high-dimensional functions (tensor network methods): Tensor network compression for high dimensional integration and its application to Feynman diagrams (case study)

Convener: Olivier Parcollet (CCQ)

Ordinary differential equations: Numerical solution of ODEs: a practical guide (tutorial/overview)

Convener: Fruzsina Agocs (CCM)

3:20 PM Coffee break

Ordinary differential equations: Physical considerations for when time integration of gas dynamics fails (case study)

Convener: Chris White (CCA)

Ordinary differential equations: Solving ODEs in a Bayesian model (case study)

Convener: Charles Margossian (CCM)

Friday, October 28

High performance computing: What every programmer should know about HPC (tutorial/overview)

Convener: Dhairya Malhotra (CCM)

High performance computing: How not to write a high-performance N-body code (case study)

Convener: Lehman Garrison (SCC)

11:00 AM Coffee break

High performance computing: GPUs: the good, the bad, and the ugly (tutorial/overview)

Conveners: Geraud Krawezik (SCC), Robert Blackwell (SCC)

12:20 PM Lunch

Practical tools for machine learning and data science: Intro to machine learning by building Scikit-Learn estimators from scratch (tutorial/overview)

Convener: Michael Eickenberg (CCM)

Practical tools for machine learning and data science: Building a connectome of the mini-wasp brain using deep learning (case study)

Convener: Jingpeng Wu (CCN)

3:10 PM Coffee break

Practical tools for machine learning and data science: Stochastic gradient descent, tricks, learning rates, and momentum (tutorial/overview)

Convener: Robert Gower (CCM)

Practical tools for machine learning and data science: Improving exoplanet detection with discriminative linear regression (case study)

Convener: Lily Zhao (CCA)

5:00 PM Reception