Me
Jennifer Brennan
Causal Inference and Experimental Design
Google Research

Google Scholar | Resume
| Google Research homepage

About

Jennifer Brennan works on experimental design in the Market Algorithms group at Google Research. Jennifer's research interests include experimental design and causal inference for marketplaces and platforms, especially settings in which interactions between the arms of an experiment cause standard methods to fail.

Jennifer received her PhD in Computer Science and Engineering from the University of Washington in 2022, advised by Kevin Jamieson, and her BS in Mathematics and Computer Science from Harvey Mudd College. Jennifer has developed experimental designs for identifying promising antibiotic combinations, bias-minimizing designs for martketplace experiments with interference, and techniques for training memory-hungry deep learning models. She is grateful to have been supported by an NSF Graduate Research Fellowship.


Publications

Cluster-Randomized Designs for One-Sided Bipartite Experiments, Jennifer Brennan, Vahab Mirrokni, Jean Pouget-Abadie. NeurIPS 2022. Paper

Sample-Efficient Identification of High-Dimensional Antibiotic Synergy with the Normalized Diagonal Sampling Design, Jennifer Brennan, Lalit Jain, Sofia Garman, Ann E. Donnelly, Erik S. Wright, Kevin Jamieson. PLOS Computational Biology, 2022. Paper

BAM: Bayes with Adaptive Memory, Josue Nassar, Jennifer Brennan, Ben Evans, Kendall Lowrey, ICLR 2022.

Dynamic Tensor Rematerialization, Marisa Kirisame, Steven Lyubomirsky, Altan Haan, Jennifer Brennan, Mike He, Jared Roesch, Tianqi Chen, Zachary Tatlock, ICLR 2021. Paper | Talk | Slides

Estimating the Number and Effect Sizes of Non-null Hypotheses, Jennifer Brennan, Ramya Korlakai Vinayak and Kevin Jamieson, ICML 2020. Paper | Talk | Slides

Reconciliation feasibility in the presence of gene duplication, loss, and coalescence with multiple individuals per species, Jennifer Rogers, Andrew Fishberg, Nora Youngs and Yi-Chieh Wu, BMC Bioinformatics, 2017. 18:292 Paper