activity
20172023
most citedUnderspecification Presents Challenges for Credibility in Modern Machine Learning

430 citations · 514 across the 9 of their papers we have counts for

collaborators

14 papers

cs.LG20228 cited

On Optimal Early Stopping: Over-informative versus Under-informative Parametrization

Ruoqi Shen, Liyao Gao, Yi-An Ma

Early stopping is a simple and widely used method to prevent over-training neural networks. We develop theoretical results to reveal the relationship between the optimal early stop…

cs.LG20211 cited

When is the Convergence Time of Langevin Algorithms Dimension Independent? A Composite Optimization Viewpoint

Yoav Freund, Yi-An Ma, Tong Zhang

There has been a surge of works bridging MCMC sampling and optimization, with a specific focus on translating non-asymptotic convergence guarantees for optimization problems into t…

stat.ML20216 cited

Variational Refinement for Importance Sampling Using the Forward Kullback-Leibler Divergence

Ghassen Jerfel, Serena Wang, Clara Fannjiang +3

Variational Inference (VI) is a popular alternative to asymptotically exact sampling in Bayesian inference. Its main workhorse is optimization over a reverse Kullback-Leibler diver…

cs.AI20211 cited

Quantifying Uncertainty in Deep Spatiotemporal Forecasting

Dongxia Wu, Liyao Gao, Xinyue Xiong +4

Deep learning is gaining increasing popularity for spatiotemporal forecasting. However, prior works have mostly focused on point estimates without quantifying the uncertainty of th…

cs.LG202120 cited

DeepGLEAM: A hybrid mechanistic and deep learning model for COVID-19 forecasting

Dongxia Wu, Liyao Gao, Xinyue Xiong +4

We introduce DeepGLEAM, a hybrid model for COVID-19 forecasting. DeepGLEAM combines a mechanistic stochastic simulation model GLEAM with deep learning. It uses deep learning to lea…

cs.LG2020430 cited

Underspecification Presents Challenges for Credibility in Modern Machine Learning

Alexander D'Amour, Katherine Heller, Dan Moldovan +37

ML models often exhibit unexpectedly poor behavior when they are deployed in real-world domains. We identify underspecification as a key reason for these failures. An ML pipeline i…