Welcome to the 2016 Symposium!
Data science is changing financial markets and is now particularly applicable to developing better risk analytics. On October 7th at UC Berkeley’s Memorial Stadium, Berkeley’s Consortium for Data Analytics in Risk hosted our 2nd annual Symposium. The program featured plenary talks by Craig Boutilier, principal scientist at Google; Dan diBartolomeo, CEO of Northfield Information Services; and Mike Jordan, Professor of Statistics and Computer Science at UC Berkeley. Jessica Donohue (Chief Innovation Officer at State Street Global Exchange) and Lisa Goldberg (CDAR Co-Director) opened the day’s program with an overview of CDAR’s first year’s successes and goals for the future. Jeff Bohn of State Street Global Exchange will led a discussion on the impact of emerging ESG (Economic, Social, and Corporate Governance) standards and practices. Alex Papanicolaou (CDAR Postdoctoral Researcher) and Alex Shkolnik (CRMR Postdoctoral Researcher, CDAR Affiliate) discussed the ways that statistical methods and theories can, or sometimes cannot, be useful in finance research.
On October 6, 2016, in partnership with BIDS, CDAR hosted a Research Spotlight in the Doe Library from 6-8:30pm. This kickoff to our annual symposium featured graduate students’ and postdoctoral researchers’ work in financial data analytics, and applications of statistical processes in risk management. We heard about up-and-coming research projects in finance and data analytics, met peers from industry and academia, and enjoyed refreshments including wine chosen by our sommelier.
CDAR Symposium Agenda
UC Berkeley – California Memorial Stadium
Friday, October 07, 2016
|8:30 – 9:40 a.m.||Breakfast|
|9:35 – 9:40 a.m.||Welcome|
|9:40 – 10:15 a.m.||CDAR’s First Year, Looking Forward
Symposium Opening slides 2016
|10:15 – 11:15 a.m.||Incorporation of Text News Analytics in Risk Assessment
Dan’s Symposium slides 2016
Dan DiBartolomeo, President, Northfield
Analytical models in finance all share some basic concepts. Financial market participants observe some period of past events they deem relevant, build a statistical model of the observed data, and then make the heroic assumption that events in the future will be like those in the past. While almost every financial institution has extensive risk modeling systems in place (as often mandated by regulators) the Global Financial Crisis has shown that such systems are frequently grossly inadequate. What is missing from nearly all models is an explicit recognition of how the present is different from the past, and therefore how the short term future is also likely to be different from the past. By defining “news” explicitly as the information set that informs us of the differences between past and present, we can condition our estimates of the distribution of future outcomes more robustly.
|11:15 – 11:30 a.m.||Break|
|11:30 a.m. – 12:30 p.m.||Exploiting Myopic Prediction Models in Reinforcement Learning
Craig Boutilier, Principal Scientist, Google
Overview of several techniques for solving large-scale reinforcement learning problems of the type that might commonly arise in advertising and recommendation contexts. We place special emphasis on techniques that exploit the data and models that are used for traditional “myopic” prediction of user behavior (e.g., CTR) to readily construct policies that optimize long-term, cumulative versions of these metrics. We outline challenges and potential solutions that arise in model-free RL in such settings, and derive novel new model-based techniques for the solution of large factored Markov decision processes.
|12:30 – 1:45 p.m.||Lunch|
|1:45 – 2:45 p.m.||What Can Statistical Methods do (or not do) for Finance?
Modern statistical methods hold much promise for researchers and academics concerned with the measurement and management of financial risk. But as in classical statistics, understanding the confines in which a particular method is useful can be crucial. In this interactive session, we will explore some successes and limitations of statistical methods with examples that touch upon bias, inference, forecasting and related themes.
|2:45 – 3:45 p.m.||On Computational Thinking, Inferential Thinking and Data Science Mike Jordan’s Symposium Slides 2016
Mike Jordan, Professor, UC Berkeley
The rapid growth in the size and scope of datasets in science and technology has created a need for novel foundational perspectives on data analysis that blend the inferential and computational sciences. That classical perspectives from these fields are not adequate to address emerging problems in “Big Data” is apparent from their sharply divergent nature at an elementary level—in computer science, the growth of the number of data points is a source of “complexity” that must be tamed via algorithms or hardware, whereas in statistics, the growth of the number of data points is a source of “simplicity” in that inferences are generally stronger and asymptotic results can be invoked. On a formal level, the gap is made evident by the lack of a role for computational concepts such as “runtime” in core statistical theory and the lack of a role for statistical concepts such as “risk” in core computational theory. I present several research vignettes aimed at bridging computation and statistics, including the problem of inference under privacy and communication constraints, and methods for trading off the speed and accuracy of inference.
|3:45 – 4:00 p.m.||Move downstairs to Stadium Club & refreshments served|
|4:00 – 5:15 p.m.||ESG Discussion
Moderator: Jeff Bohn, State Street
This ESG panel will look at issues related to data, methodologies and performance as related to portfolio risk modeling and portfolio performance. In particular, the panel will drill into the environment, social and governance issues separately to consider which themes will have more or less impact on financial portfolios.
|5:15 – 5:30 p.m.||Closing Remarks
Bob Anderson, CDAR