FRIDAY APRIL 21, 2017
MIT SDSCon, Statistics and Data Science Conference
MIT EG&G Education Center, 50 Vassar Street, Building 34 Room 401 (4th Floor)
Cambridge, MA 02139
Please note, lunch will be located in building 32, 6th floor of Dreyfoos Tower, room D677
8:00am: Registration and Breakfast
9:00am: Introductory remarks: Professor Devavrat Shah, Director of SDSC
9:15am: Panel Session 1:
- David Gamarnik, Professor, MIT, Sloan School of Management
- Tamara Broderick, Assistant Professor, MIT, EECS
- Stefanie Jegelka, Assistant Professor, MIT, EECS
- Philippe Rigollet, Associate Professor, MIT, Mathematics
- Guy Bresler, Assistant Professor, MIT, EECS
- Ankur Moitra, Assistant Professor, MIT, Mathematics
10:45am: Coffee Break
11:00am: Plenary Talk: Michael Steele, Professor, University of Pennsylvania, Statistics
Title: Expectations: What Makes Them Great?
Abstract: Everyone knows the Saint Petersburg paradox, but knowledge of the paradox is not always taken to heart when algorithms are designed for making real-time sequential decisions. There is vast — and beautiful —- literature that pays unflinching devotion to decision rules that are designed with the single objective of maximizing an expected reward. The Saint Petersburg paradox suggests that these rule might be disastrous, but in practice things tend to work out quite reasonably.
The question is: Why? In this talk, I will review some new theoretical work that helps to explain the puzzle. Typically, the key issue is the identification of those dynamic programs where the resulting decision rules are mathematically well-behaved from a probabilistic point of view. In one general class of problems, the Saint Petersburg terror is abated by finding good, easily computed, bounds on the variance of the realized reward. In other contexts, which include several classical combinatorial problems and some well-studied inventory management models, one can even show that the realized rewards are asymptotically normal. The menace of the Saint Petersburg paradox is not entirely overcome, but at least one gains some understanding of why so many dynamic programming solutions have served us so well for so long.
12:00pm: Lunch and poster session featuring SDSC students (located in Building 32, Stata Center, Dreyfoos Tower, 6th Floor Lounge)
1:15pm: Industry Session
- Alex Cosmas, Chief Scientist, Booz Allen Hamilton
- Brian Ulciny, Director, Data Science, Thomson Reuters Labs
1:45PM: Plenary Talk: Jennifer Listgarten, Senior Researcher, Microsoft Research
Title: Genetics to CRISPR Gene Editing with Machine Learning
Abstract: Molecular biology, healthcare and medicine have been slowly morphing into large-scale, data driven sciences dependent on machine learning and applied statistics. In this talk I will start by explaining some of the modelling challenges in finding the genetic underpinnings of disease, which is important for screening, treatment, drug development, and basic biological insight. Genome and epigenome-wide associations, wherein individual or sets of (epi)genetic markers are systematically scanned for association with disease are one window into disease processes. Naively, these associations can be found by use of a simple statistical test. However, a wide variety of structure and confounders lie hidden in the data, leading to both spurious and missed associations if not properly addressed. Much of this talk will focus on how to model these types of data. Once we uncover genetic causes, genome editing—which is about deleting or changing parts of the genetic code—will one day let us fix the genome in a bespoke manner. Editing will also help us to understand mechanisms of disease, enable precision medicine and drug development, to name just a few more important applications. I will close by discussing how we developed machine learning approaches to enable more effective CRISPR gene editing.
2:45PM Coffee Break
3:15PM: Panel Session 2:
- Emery Brown, Professor, MIT, Brain and Cognitive Sciences
- Rahul Mazumder, Assistant Professor, MIT, Sloan School of Management
- In Song Kim, Assistant Professor, MIT, Department of Political Science
- Alberto Abadie, Professor, MIT, Economics
4:30pm: Plenary Talk: James Stock, Professor, Harvard University, Economics
Title: Statistical Analysis of Climate Data
Abstract: The science of climate change has largely been the domain of increasingly sophisticated geophysics-based models of the earth climate system. A complementary approach is to abstract from the detailed mechanisms in climate models and instead use statistical tools to analyze historical time series data on temperature, emissions, and climate record. Although statistical approaches to climate change have limitations because of available data, they have the ability to transparent insights into climate relations, including measures of uncertainty. Because of their transparency, statistical approaches to climate change have the potential to contribute to the public discussion around climate change in a way that complex models cannot. This talk examines recent work on the statistical approach to climate change, and focuses on three topics: the link between anthropogenic emissions and global temperatures, estimates of the transient climate response, and some more specific areas in which statisticians can contribute to climate science. Addressing these questions brings statisticians into the familiar territory of identifying (dynamic) causal effects, challenges posed by persistent time series, and spatial-temporal modeling.
5:30pm: Closing Remarks