Rahul Nair

Rahul Nair

Research Staff Member

IBM Research Europe

Biography

Rahul Nair is a Research Staff Member at IBM Research Europe in Dublin. His research interests are in technology and development (societal applications), trusted computing. His expertise is in optimization, machine learning applied across sectors particularly transportation, healthcare, and business computing. He holds a Ph.D. from University of Maryland College Park.

Recent Posts

Optimal Counterfactuals in Tree Ensembles

This is a brief review of a paper by Parmentier and Vidal on “Optimal Counterfactual Explanations in Tree Ensembles”. A counterfactual explanation, in the context of machine learning (ML) models, answers the question what is the minimal change in input (features) that yields a different (desirable) outcome?

Some notes on Models of Environments

Suppose you are given a dataset on some process with the following attributes. The process that generates the data is complex and not fully observable. The data only captures part of dynamics of the process.

Prospective Explanations

At NeurIPS 2021 this year, Pierpaolo Tomassi and I have a demonstrator that brings together a few ideas on surrogates, bayesian networks, and model understanding. The system (while its still up!

Partitioning a set

Integer programming approaches to set partitioning

Zero-Suppressed Decision Diagrams and Independent Sets

I stumbled across Binary Decision Diagrams (BDDs) by chance. They are an efficient data structure to represent sets of graphs. While a graph $G$ is a set of vertices $V$ along with a set of edges $E$ that connect the vertices, a graph set is a collection of subgraphs over the universe $V$.

Recent Publications

Quickly discover relevant content by filtering publications.

Artificial intelligence indocyanine green (ICG) perfusion for colorectal cancer intra-operative tissue classification

Classification of tissue based on near infrared imaging of flow of dyes

Robust transit line planning based on demand estimates obtained from mobile phones

Line planning with demand range estimates

User Driven Model Adjustment via Boolean Rule Explanations

ML systems that allow users to adjust decision boundaries

What Changed? Interpretable Model Comparison

Addresses the problem of distinguishing two machine learning (ML) models built for the same task in a human-interpretable way

Perfusion Quantification from Endoscopic Videos: Learning to Read Tumor Signatures

Describes methods for interoperative decision support during endoscopies.