Discussion Series: Controversial Topics in Precision Medicine and Machine Learning


This series of discussions will be held in person at SAMSI, as well as broadcasted via the WebEx platform. The goal of this webinar series is to create a platform for productive discussion on controversial topics. During our Wednesday sessions, leaders from industry and academia will give a 10-15 minute talk about controversial problems they have encountered in their research or applications. After the introductory talk, the room will be open to a group discussion of the presented topic, problems, and possible solutions. These discussions could inspire research ideas or even spark collaborations between participants.

** Notice of Consent **

SAMSI values the proprietary and intellectual property of our participants. The materials presented at our various workshops and programs are in high demand by event participants and the applied mathematics and statistics community that comprise our audience. Therefore, we encourage all of our invited speakers to share their materials, as appropriate, in order to pass along the valuable research that is being done in your field of study and is a focus of this event. In addition, unless SAMSI is give written approval from our speakers we ARE NOT authorized to share the materials presented at this event.

Please click HERE to complete a SAMSI Consent form for this event. SAMSI appreciates your time and willingness to share this valuable content with others and we hope you enjoy this event!
For any questions or concerns about our consent policy, please contact us at: communications@samsi.info


Wednesdays 2:30 pm-3:30pm at SAMSI, Research Triangle Park, NC.

This informative discussion series is being offered through a collaboration between SAMSI and Duke University Computer Science. For questions about this series, contact Alina Jade Barnett and Lesia Semenova – they are the organizers for this event.

February 13, 2019 2:30pm – 3:30pm

Lecture: Causal inference

Location: SAMSI Classroom
Speaker: Fan Li, Associate Professor, Department of Statistical Science, Duke University


Fan Li will discuss a few challenging problems in causal inference emerging from the era of precision medicine and machine learning. Some topics will include:
1) What do we mean by precision medicine? Prediction vs. Causal?
2) What do we mean by machine learning (ML)? Is ML a magic bullet?
3) What type of problems in precision medicine that ML is mostly effective (compared to traditional methods like regression)?
4) Different frameworks of causal inference: is this a conflict of ideology and/or personality rather than substance? Does it really matter in practice?

February 27, 2019 2:30pm – 3:30pm

Lecture: Unequal Performance Across Groups in Face Image Classification

Location: SAMSI Classroom
Speaker: Dr. Kush R Varshney, Principal Research Staff Member and Manager, Learning and Decision Making, IBM Research AI, IBM


Recent work shows unequal performance of commercial face classification services in the gender classification task across intersectional groups defined by skin type and gender. Accuracy on dark-skinned females is significantly worse than on any other group. In this discussion, we will give a broad context to this issue and analyze why it is happening.

April 10, 2019 2:30pm – 3:30pm

Lecture: The Need for Interpretable Machine Learning Models in Precision Medicine

Location: SAMSI Classroom
Speaker: Eric Laber, Ph.D., Associate Professor, Department of Statistics, North Carolina State University


To be determined

May 1, 2019 2:30pm – 3:30pm

Lecture: Including Machine Learning in Clinical Practice

Location: SAMSI Classroom
Speaker: David Carlson, Ph.D., Assistant Professor of Civil and Environmental Engineering, Duke University


It is becoming increasingly common to include machine learning techniques in clinical analysis with significant promises to improve predictions, with the implication that such techniques will improve clinical outcomes. However, it is often unclear and controversial on how to actually include the built models in clinical practice. I will first briefly discuss some of my recent research on methods to learn improved bio-markers from clinical trials, and then talk about some of the issues in deployment.