via ZOOM Meeting
Please join us for the MINDS& CIS Seminar Series
Tuesday, June 2, 2020 at 12:00 pmEastern Time (US and Canada)
?Identification Theory in Segregated Graph Causal Models? by EliSherman(CS,JHU)
Seminar will be remote via Zoom
Join Zoom Meeting
MeetingID: 915 7891 8871
Abstract- In recent years there has been an explosion of interest incausal inference methodologies in the machine learning and broader data sciencecommunities. A key issue at the foundation of all causal analysis is theconcept of identification: the work of Pearl and others has sought to formallycharacterize when causal queries are estimable from available (i.e. observed)data given the assumed causal model. In this talk I’ll discuss extensions tothe classical latent-variable DAG identification framework that are particularlyrelevant when there is dependence among data samples, such as social networkingand spatial settings. Toward this end, I will introduce the segregated graphmodel, a super model for latent-variable DAGs, and argue for its use in thesedependent settings. I will then provide sound and complete identificationresults for ‘node’ (i.e. classical) interventions. Finally, I will describesound and complete results for the identification of ‘policy’ interventions,corresponding to a sequential decision-making setting, in segregated graphs anddemonstrate how these results generalize and nest several existingidentification results.
Bio – Eli Sherman is a PhD student in the Computer Science Department atJohns Hopkins University. He develops methods for obtaining causal inferencesin social networking and dependent data contexts as well as approaches forintervention tailoring. He is interested in applications of these methods tohealthcare, economics, and public policy. Eli is supervisedby Ilya Shpitserand is affiliated with the Malone Center for Engineering in Healthcare and theMathematical Institute for Data Science, from which hereceives support throughthe MINDS PhD Fellowship.
This event has a video call.