1

From Search to Sampling: Generative Models for Robust Algorithmic Recourse

We propose a generative model for recourse that outputs a distribution over likely recourse instances.

Prateek Garg, Lokesh Nagalapatti, Sunita Sarawagi

Robust Root Cause Diagnosis using In-Distribution Interventions

Identifying root cause of anomalies using interventions rather estimated from a learned SCM

Lokesh Nagalapatti, Ashutosh Srivastava, Sunita Sarawagi, Amit Sharma

Tab-Shapley: Identifying Top-k Tabular Data Quality Insights

Identifying group anomalies (covering both columns and rows) in Tabular Data using Game theoretic models.

Manisha Padala, Lokesh Nagalapatti, Atharv Tyagi, Ramasuri Narayanam, Shiv Kumar Saini

Continuous Treatment Effect estimation using Gradient Interpolation and Kernel Smoothing

A two-pronged strategy to infer treatment effects for continuous treatments.

Lokesh Nagalapatti, Akshay Iyer, Abir De, Sunita Sarawagi

Gradient Coresets for Federated Learning

Finding coresets in federated learning.

Durga Sivasubramanian (=), Lokesh Nagalapatti, Rishabh Iyer, Ganesh Ramakrishnan

PairNet: Training with Observed Pairs to Estimate Individual Treatment Effect

A simple loss on pairs of observational instances outperforms (almost) SOTA approaches on Treatment Effect Estimation

Lokesh Nagalapatti, Pranava Singhal (=), Avishek Ghosh, Sunita Sarawagi

Learning Recourse on Instance Environment to Enhance Prediction Accuracy

Learning recourse on the environment that generates data instances.

Lokesh Nagalapatti, Guntakanti Sai Koushik, Abir De, Sunita Sarawagi

Is your data relevant?: Dynamic selection of data in Federated Learning.

Using Reinforcement Learning based techniques to enable clients in Federated Learning derive updates only from relevant data.

Lokesh Nagalapatti, Ruhi Sharma Mittal, Ramasuri Narayanam

Game of Gradients - Mitigating irrelevant clients in Federated Learning

Using Cooperative Game theory approaches to mitigate the impace of noisy clients in Federated Learning.

Lokesh Nagalapatti, Ramasuri Narayanam

Ranking Data Slices for ML Model Validation: A Shapley Value Approach

Shapley values based approach to detect slices of data that needs to be analyzed to debug the performance of a black-box ML model.

Eitan Farchi, Ramasuri Narayanam, Lokesh Nagalapatti