The STAMPS webinars are open to everyone.
September 2 - Welcome meeting #
Welcome newcomers and returning participants!
September 9, 1:30PM EDT - Public Webinar : Lukas Heinrich (Technical University Munich)
September 16 - Donata Giglio (University of Colorado Boulder) #
Title: Advancing our understanding of the climate system: challenges and opportunities from the Argo array
Description: The Argo program has provided an unprecedented view of the global ocean temperature, salinity, pressure, and biogeochemical fields. We will discuss scientific advances that the (roughly) two and a half million Argo profiles available make possible, with a focus on challenges and opportunities for statisticians.
Note: Advanced Data Analysis (ADA) related talk
September 23 - Julia Walchessen #
Title: Learning Likelihood Surfaces for Spatial Processes with Computationally Intensive or Intractable Likelihood Functions
Abstract: In spatial statistics, fast and accurate parameter estimation coupled with a means of uncertainty quantification can be a challenging task because the likelihood functions might be slow to evaluate or intractable. In this work, we propose using convolutional neural networks to learn the likelihood function for a spatial process. Through a specifically designed classification task, our neural network implicitly learns the likelihood function, even in situations where the likelihood is not explicitly available. To demonstrate our approach, we compare the neural network maximum likelihood parameter estimates to parameter estimates using standard methods such as exact maximum likelihood and composite likelihood for two different spatial processes—a Gaussian Process, which has a computationally intensive likelihood function, and a Brown-Resnick Process, which has an intractable likelihood function. We also compare the learned likelihood surfaces to the exact and composite likelihood surfaces for the Gaussian Process and Brown-Resnick Process, respectively. We conclude that our method provides fast and accurate parameter estimation with a potential method of uncertainty quantification in situations where standard methods are either undesirably slow or inaccurate.
September 30 - Coty Jen (Chemical Engineering at CMU) #
Title: Chemical Fingerprinting of Wildfire Smoke
Description: Each year, the US is blanketed by wildfire smoke produced from western North American wildfires. Wildfire smoke is composed of thousands of different molecules that vary depending on what fuels are burned. This talk will cover current wildfire analytical science and the importance of predicting fuel composition in helping forest managers reduce wildfire risk.
Note: Advanced Data Analysis (ADA) related talk
October 7 - Hamish Gordon (Chemical Engineering at CMU) #
Title: Understanding how fires affect climate by constraining the parameter space of an IPCC climate model
Description: Fires have substantial, but uncertain, effects on Earth’s climate. By comparing climate model output to atmospheric observations, this project aims to develop statistical techniques to reduce uncertainties in a large ensemble of climate model simulations and apply the techniques to constrain the effects of smoke from fires on climate.
Note: Advanced Data Analysis (ADA) related talk
October 14, 1:30PM EDT - Public Webinar : Benjamin Nachman (Lawrence Berkeley National Laboratory)
October 21 - No meeting #
Enjoy fall break!
October 27, 3:30PM EDT - Public Webinar : Kaze Wong (Flatiron Institute)
October 28 - No meeting (tentative) #
TBA
November 4 - LANL Statistics Group #
Topic: LANL ADA projects overview
Note: Advanced Data Analysis (ADA) related talk
November 11 - Purvasha Chakravarti (University College London) #
Title: Robust Signal Detection using a Classifier Decorrelated through Optimal Transport (CDOT)
November 18 - No meeting #
November 25 - No meeting #
Enjoy Thanksgiving break!
December 2 - Woonyoung Chang (Statistics and Data Science at CMU) #
Title: Chemical Fingerprinting of Smoke
Note: This is an ADA project presentation for the Department of Statistics & Data Science at CMU. Advisors: Ann Lee (StatsDS) and Coty Jen (Chemical Engineering).
