Seminar

Solving Complex Problems with Design Thinking and Data Science

Thursday, November 30, 2017 - 11:00am to 12:00pm
New York, NY 10027
United States

Columbia Data Science Institute Industry Innovation Seminars


Presented by Lucas Saloumi (Data Scientist), Rick Winslow (Head of Commercial Digital Innovation) and Zeina Zeitouni (Product Manager)

Capital One is a pioneer in applying both artificial intelligence and design thinking to make banking smarter, more intuitive and personal. Presenters from the Commercial Banking Digital Innovation team will show how designers and data scientists can work together on cross-functional teams to solve complex problems and build compelling new products. They will share examples of how machine learning classification and outlier detection has been applied in their business. And they will discuss the challenges faced when deploying intelligent systems at scale, all while keeping customers and business partners engaged in the process.

Some Optimization Problems in Smart Grids: Steven Low, Caltech

Friday, November 3, 2017 - 11:00am to 12:00pm
Columbia University
New York, NY 10027
United States

Data Science Institute | Sense, Collect & Move Data Center Seminar

Speaker: Steven Low, Caltech

Optimal power flow (OPF) is fundamental because it underlies numerous power system operation and planning problems. In this talk, I will give a sample of optimization problems in the management of a large network of distributed energy resources. The nonlinearity of power flow equations leads to the nonconvexity of OPF, one of the main computational challenges in power system applications. We describe a method to deal with nonconvexity through semidefinite relaxation. Semidefinite programs are hard to scale to large OPF problems. We describe a highly scalable distributed solution based on ADMM. These algorithms are offline in that they iterate until the computation has converged before applying the final solution to the grid, and are therefore not suitable for real-time optimization of distributed energy resources at scale. We describe realtime OPF that explicitly exploits the network as a power flow equation solver and characterize its performance in tracking changing network conditions. Finally, in practice, not all network nodes have sensors or can be controlled. We characterize controllability and observability of power flow dynamics in terms of the spectral properties of its Laplacian matrix. This characterization can be used to optimize the placement of sensors and actuators in the grid.

How the New Availability of Urban and Industrial Data are Impacting Our World from Public Safety to Jet Engines

Thursday, May 11, 2017 - 4:00pm to 5:30pm
Schapiro Hall (CEPSR) 750 (Costa Engineering Commons)
Columbia University
New York, NY 10027
United States

Columbia Data Science Institute Industry Innovation Seminars

(Talk 2 of 2)


Peter Marx, Vice President, Advanced Projects, GE Digital, Adjunct Professor, USC

Estimating Causal Effect of Ads in a Real-Time Bidding Platform

Wednesday, April 26, 2017 - 5:00pm to 6:00pm
Davis Auditorium | The Schapiro Center | Columbia University
530 West 120th Street
New York, NY 10027
United States

Columbia Data Science Innovation Seminars

Prasad Chalasani
SVP, Data Science
MediaMath

Title: Estimating Causal Effect of Ads in a Real-Time Bidding Platform

Abstract: A real-time bidding platform responds to incoming ad-opportunities ("bid requests") by deciding whether or not to submit a bid and how much to bid. If the submitted bid wins, the user is shown an ad. Advertisers hope that ad-exposure leads to an increased likelihood of a desired action, such as a click or conversion (purchase, etc). So an important quantity that advertisers want to measure is the causal effect of advertising, namely, what is the response probability of an exposed user, compared with the counterfactual (un-observable) response-rate of the user if they were not exposed to the ad. In an ideal randomized test, the user is randomly assigned to test or control AFTER the submitted bid is won, and test users are served the ad in the normal way, while control users are not. While this is ideal from a statistical perspective, in practice this approach has the drawback that money spent by advertisers is wasted when a user is assigned to control. At Media Math we have developed a methodology for causal effect measurement where users are assigned to test or control BEFORE bid submission. One challenge here is that not all test-group users are exposed to an ad; only a winning bid results in ad exposure, and the winning population can have a significant bias. This talk will describe our approach to handle this and other challenges to ad impact measurement in this setting, and how we use MCMC Gibbs sampling to arrive at confidence intervals for ad-impact.

Closing the Gap Between Digital Technology and Prevention of Disease Using Data & Analytics

Thursday, February 23, 2017 - 5:00pm to 6:00pm
United States

Columbia Data Science Innovation Seminars

Evan Garmaise, Data Scientist
Junghoon Woo, Data Scientist

In 2016, the Department of Health and Human Services announced the certification of the Diabetes Prevention Program (DPP). The DPP aims to reach 86 million pre-diabetic Medicare participants in the United States through education, training, and lifestyle coaching. According to the physician payment rule recently announced, Medicare will be reimbursing both digital and in-person versions of the DPP; however, it remained unclear how the parameters for the digital version will be set. Due to the relatively short history of digital DPP, little is known regarding the mechanism of weight loss when the services are rendered through a mobile app. This will be critical for CMS to set payment mechanism by outcomes as it announced. To better understand the mechanism of weight loss by digital DPP solutions, and to help CMS make the most informed decision on the payment rule, we have collaborated with one of the few certified digital DPP.

Pages

Subscribe to RSS - Seminar
Back to Top