Md Musfiqur Rahman

I am a Ph.D. student working with Professor Murat Kocaoglu in the School of Electrical and Computer Engineering, at Purdue University.

I apply causal inference to solve trustworthy ML problems using deep generative models and large pre-trained models. Currently, I am exploring areas such as adaptive causal effect estimation, generalization in causal generative models, causal representation learning, and more.

I am a regular reviewer of papers in UAI, ICML, ICLR, AAAI, NeurIPS, AISTAT etc.

You can see my resume here.
Follow me: #Email # Twitter #Github #LinkedIn #Google Scholar

News!

[New 🎉🎉] Joined Amazon as a summer intern to contribute to the Pricing and Promotion team. Alhamdulillah!
[New 🎉🎉] Got enlisted as notable reviewer in ICLR.
[New 🎉🎉] Our paper titled "FeDCM: Federated Learning of Deep Causal Generative Models" is accepted in UAI 2025! .
[New 🎉🎉] Passed my Phd Preliminary examination!
[Fall 2024] Our paper titled "Conditional Generative Models are Sufficient to Sample from Any Causal Effect Estimand" is accepted in NeurIPS 2024! .
- Presentation: https://recorder-v3.slideslive.com/?share=96768&s=09f41687-e6b3-4daa-ab8e-e751bf277f67
- Paper: https://nips.cc/virtual/2024/poster/93180
[Summer 2024] Our paper titled “Modular Learning of Deep Causal Generative Models for High-dimensional Causal Inference” is accepted at ICML’24.
- Presentation: https://youtu.be/TvWHdvfHk2o
- Paper: https://icml.cc/virtual/2024/poster/33643
[Summer 2024] Joined Genentech as a summer intern to contribute in Causal ML projects.
[Spring 2023] Our paper 'Towards Modular Learning of Deep Causal Generative Models' has been accepted for poster presentation at SCIS 2023 and SPIGM 2023 workshops.
[Spring 2023] I have been identified as one of the Top Reviewers at UAI 2023.
[Spring 2023] Our paper "Finding Invariant Predictors Efficiently via Causal Structure" has been accepted at UAI 2023.

Research History

Research Assistant in the School of Electrical and Computer Engineering, Purdue University. (Aug 2021- Now)
- Proposed a diffusion-based approach to sample from any interventional distribution given an arbitrary causal graph containing latent confounders.
- Developed a deep causal generative model to learn high-dimensional data distribution in a modular fashion, enabling the use of pre-trained models for causal and counterfactual effect estimation.
- Achieved expertise in training deep generative models: classifier-free diffusion models, conditional GANs, auto-encoders, etc.
- Established a strong understanding of state-of-the-art algorithms of causal effects estimation, counterfactual identification, transportability, invariant prediction with domain shift, and more.
Research assistant in the Cognitive Agents and Interaction Lab (CAIL) in the Department of Computer Science and Engineering, University of Dhaka.

Accelerated recursive partition-based causal structure learning and reduced the number of Conditional independence tests for detecting false edges up to 86%.
Accomplished higher parallelism and reduced message size to solve Distributed Constraint Optimization Problem in 4% - 96% faster runtime.

RESEARCH AND PUBLICATIONS

Conditional Generative Models are Sufficient to Sample from Any Causal Effect Estimand

Authors: Md Musfiqur Rahman, Matt Jordan, Murat Kocaoglu
Accepted in: Thirty-eighth Annual Conference on Neural Information Processing Systems (NeurIPS 2024)
See manuscript.

Modular Learning of Deep Causal Generative Models for High-dimensional Causal Inference

Authors: Md Musfiqur Rahman, and Murat Kocaoglu
Accepted in: 41st International Conference on Machine Learning (ICML-2024).
See manuscript.

Towards Modular Learning of Deep Causal Generative Models.

Authors: Md Musfiqur Rahman, and Murat Kocaoglu
Accepted in: ICML 2023 Workshops: i) on Structured Probabilistic Inference & Generative Modeling and ii) on Spurious Correlations, Invariance, and Stability (See workshop paper)

Finding Invariant Predictors Efficiently via Causal Structure

Authors: Lee, Kenneth, Md Musfiqur Rahman, and Murat Kocaoglu
Accepted in: Uncertainty in Artificial Intelligence. PMLR, 2023. (See Full Paper)

Accelerating Recursive Partition-Based Causal Structure Learning Using An Improved Structure Refinement Approach

Authors: Md Musfiqur Rahman, Ayman Rasheed, Md. Mosaddek Khan, Mohammad Ali Javidian, Pooyan Jamshidi & Md. Mamun-Or-Rashid
Accepted in: 20th International Conference on Autonomous Agents and Multiagent Systems (AAMAS-2021). (See Full Paper)

Speeding Up Distributed Pseudo-tree Optimization Procedure with Cross Edge Consistency to Solve DCOPs

Authors: Mashrur Rashik, Md Musfiqur Rahman, Md. Mamun-or-Rashid, Md. Mosaddek Khan, Long Tran-Thanh & Nicholas R. Jennings
Accepted in : Springer Journal - Applied Intelligence (Impact Factor: 2.882) (2020) (See full journal)

Google Sites

Report abuse

Page details

Page updated

Google Sites

Report abuse

Md Musfiqur Rahman

News!

Research History

RESEARCH AND PUBLICATIONS

Conditional Generative Models are Sufficient to Sample from Any Causal Effect Estimand

Modular Learning of Deep Causal Generative Models for High-dimensional Causal Inference

Towards Modular Learning of Deep Causal Generative Models.

Finding Invariant Predictors Efficiently via Causal Structure

Accelerating Recursive Partition-Based Causal Structure Learning Using An Improved Structure Refinement Approach

Speeding Up Distributed Pseudo-tree Optimization Procedure with Cross Edge Consistency to Solve DCOPs