Elan Rosenfeld

UPDATE: I am now a research scientist at Google Research, and I am no longer maintaining this page. If you would like to get in touch, please reach out to
[firstname].[lastname]@gmail.com

Hi! I'm a PhD student in the Machine Learning Department at CMU where I am fortunate to be advised by Andrej Risteski and Pradeep Ravikumar. I graduated from CMU with degrees in Computer Science and Statistics & Machine Learning, where my senior thesis on the computational complexity of human-usable password schemas was advised by Manuel Blum and Santosh Vempala. Before entering grad school I spent two years as a software engineer at Google NYC where I worked on Search Research and Machine Intelligence (SRMI).

[CV] [Google Scholar]

Recent Updates

— One-Shot Strategic Classification will appear at ICML 2024.

— Gave a talk on Opposing Signals at ML Collective.

— Opposing Signals and Intervention Extrapolation will appear at ICLR 2024.

I'm interested in the foundations of machine learning with a focus on robustness, reliability, and security. My research emphasizes principled approaches to improving robustness and generalization, particularly under distribution shift or adversarial manipulation. Much of my work develops new models of data—both theoretical and conceptual—to help understand and solve failure modes of existing methods. I also frequently study questions in representation learning: when representations are identifiable and efficiently learnable, how they arise during training, and how to ensure they will robustly generalize.

Outliers with Opposing Signals Have an Outsized Effect on Neural Network Optimization
ICLR 2024

Elan Rosenfeld, Andrej Risteski
[arXiv] [ML Collective talk] [X thread]

(Almost) Provable Error Bounds Under Distribution Shift via Disagreement Discrepancy
NeurIPS 2023

Elan Rosenfeld, Saurabh Garg
[arXiv] [code] [poster] [X thread]

Domain-Adjusted Regression or: ERM May Already Learn Features Sufficient for Out-of-Distribution Generalization
NeurIPS DistShift 2022

Elan Rosenfeld, Pradeep Ravikumar, Andrej Risteski
[arXiv] [poster]

The Risks of Invariant Risk Minimization
ICLR 2021

Elan Rosenfeld, Pradeep Ravikumar, Andrej Risteski
[arXiv] [poster] [CMU AI Seminar talk] [2 minute video presentation]

Certified Adversarial Robustness via Randomized Smoothing
ICML 2019

Jeremy Cohen, Elan Rosenfeld, Zico Kolter
[arXiv] [code] [short talk] [Zico's Simons talk]

One-Shot Strategic Classification Under Unknown Costs
ICML 2024

Elan Rosenfeld, Nir Rosenfeld
[arXiv]

Identifying Representations for Intervention Extrapolation
ICLR 2024

Sorawit Saengkyongam, Elan Rosenfeld, Pradeep Ravikumar, Niklas Pfister, Jonas Peters
[arXiv]

Learning Linear Causal Representations from Interventions under General Nonlinear Mixing
NeurIPS 2023

Simon Buchholz*, Goutham Rajendran*, Elan Rosenfeld, Bryon Aragam, Bernhard Schölkopf, Pradeep Ravikumar
[arXiv]

Iterative Feature Matching: Toward Provable Domain Generalization with Logarithmic Environments
NeurIPS 2022

Yining Chen, Elan Rosenfeld, Mark Sellke, Tengyu Ma, Andrej Risteski
[arXiv] [poster]

Analyzing and Improving the Optimization Landscape of Noise-Contrastive Estimation
ICLR 2022

Bingbin Liu, Elan Rosenfeld, Pradeep Ravikumar, Andrej Risteski
[arXiv] [blog post]

Deep Attentive Variational Inference
ICLR 2022

Ifigeneia Apostolopoulou, Ian Char, Elan Rosenfeld, Artur Dubrawski
[OpenReview] [blog post]

An Online Learning Approach to Interpolation and Extrapolation in Domain Generalization
AISTATS 2021

Elan Rosenfeld, Pradeep Ravikumar, Andrej Risteski
[arXiv]

Certified Robustness to Label-Flipping Attacks via Randomized Smoothing
ICML 2020

Elan Rosenfeld, Ezra Winston, Pradeep Ravikumar, Zico Kolter
[arXiv] [code (.zip)] [blog post] [virtual poster/presentation]

Loss in the Crowd: Hidden Breakthroughs in Language Model Training
ICML 2024 Workshop on Mechanistic Interpretability

Sara Kangaslahti, Elan Rosenfeld, Naomi Saphra
[OpenReview]

APE: Aligning Pretrained Encoders to Quickly Learn Aligned Multimodal Representations
NeurIPS 2022 Workshop: Has It Trained Yet?

Elan Rosenfeld, Preetum Nakkiran, Hadi Pouransari, Oncel Tuzel, Fartash Faghri
[arXiv] [poster]

Self-Reflective Variational Autoencoder
ICLR 2021 Workshop: Hardware Aware Efficient Training

Ifigeneia Apostolopoulou, Elan Rosenfeld, Artur Dubrawski
[arXiv] [poster] [short presentation]

Human-Usable Password Schemas: Beyond Information-Theoretic Security
CMU Senior Thesis
Awarded “Exemplary Senior Honors Thesis”

Elan Rosenfeld, Santosh Vempala, Manuel Blum
[arXiv] [poster]

You found my thesis proposal!