About Me

I am a 4th year PhD student at Duke University, advised by Katherine A. Heller. I have been a visiting student researcher at UC Berkeley with Michael I. Jordan since August 2018. I have also been a student researcher at Google Brain since January 2019.

I am interested in the design and theoretical analysis of probabilistic inference algorithms. More generally, I strive to better understand uncertainty quantification which I believe is crucial for robust decision making and efficient lifelong learning in a dynamic environment mirroring our physical world.

I have been fortunate to receive the following fellowships: Facebook Emerging Scholar (2019-2021, declined), Pratt-Gardner Graduate Fellowship (2017-2019).

During my Ph.D., I have spent wonderful summers interning at Google Brain (2019) and Google Research NY (2018).

Before moving to Duke, I graduated with a Bachelor's and a Master's degree in Computer Science from Princeton University, where I was advised by Barbara E. Engelhardt.

Preprints

• Variational Refinement for Importance Sampling Using the Forward Kullback-Leibler Divergence
  G Jerfel*, S Wang*, C Fannjiang, KA Heller, YA Ma, MI Jordan; accepted at AABI 2021; submitted to AISTATS 2021.
• Deep Uncertainty and the Search for Proteins
  Z Mariet, G Jerfel, Z Wang, C Angermueller, D Belanger, S Vora, M Bileschi, L Colwell, D Sculley, D Tran, J Snoek; accepted at ML 4 Molecules @ NeurIPS 2020; submitted to ICML 2021.

Publications

• Combining Ensembles and Data Augmentation Can Harm Your Calibration
  Y Wen*, G Jerfel*, R Muller, MW Dusenberry, J Snoek, B Lakshminarayanan, D Tran; ICLR; 2021.
• Efficient and Scalable Bayesian Neural Nets with Rank-1 Factors
  MW Dusenberry*, G Jerfel*, Y Wen, YA Ma, J Snoek, K Heller, B Lakshminarayanan, D Tran; ICML ; [Code]; 2020.
• Improving Calibration of BatchEnsemble with Data Augmentation
  Y Wen*, G Jerfel*, R Muller, MW Dusenberry, J Snoek, B Lakshminarayanan, D Tran; Workshop on Uncertainty and Robustness at ICML ; [Code]; 2020.
• Analyzing the Role of Model Uncertainty for Electronic Health Records
  MW Dusenberry, D Tran, E Choi, J Kemp, J Nixon, G Jerfel, K Heller, AM Dai; ACM Conference on Health, Inference, and Learning (CHIL); 2020.
• Reconciling meta-learning and continual learning with online mixtures of tasks
  G Jerfel*, E Grant*, T Griffiths, KA Heller; NeurIPS Spotlight; 2019.
• AdaNet: A Scalable and Flexible Framework for Automatically Learning Ensembles
  C Weill, J Gonzalvo, V Kuznetsov ... G Jerfel, V Macko, B Adlam, M Mohri, C Cortes; Automated Machine Learning at ICML; [Code]; 2019.
• Measuring Calibration in Deep Learning
  J Nixon, M Dusenberry, L Zhang, G Jerfel, D Tran; Uncertainty and Robustness in Deep Visual Learning at CVPR; 2019.
• Dynamic Collaborative Filtering with Compound Poisson Factorization
  G Jerfel, ME Basbug, BE Engelhardt; AISTATS; 2017.

Technical Reports

• Non-asymptotic Analysis of Langevin Monte Carlo Algorithms: A Review of Three Influential Papers; Qualifying Exam Report; 2019.
• Boosting Stochastic Backpropagation for Variational Inference; Master's Thesis; 2017.
• Dynamic Compound-Poisson Factorization; Bachelor's Thesis; 2016.
• Adaptive Non-Uniform Feature Subsampling for Stochastic Variational Inference; Junior Independent Work (Contact for manuscript); 2015.

Links: