Dan Moldovan

TL;DR: This work shows the need to explicitly account for underspecification in modeling pipelines that are intended for real-world deployment in any domain, and shows that this problem appears in a wide variety of practical ML pipelines.

TL;DR: It is found that increasing both the training set and model sizes significantly improve the distributional shift robustness and it is shown that, perhaps surprisingly, simple changes in the preprocessing can significantly mitigate robustness issues in some cases.

TL;DR: In this article, the authors investigate the impact of pre-training data size, the model scale, and the data preprocessing pipeline on distributional shift robustness of CNNs.

TL;DR: Techniques from the field of automatic differentiation that can give researchers expressive power, performance and strong usability are explored, including source-code transformation (SCT), flexible gradient surgery, efficient in-place array operations, and higher-order derivatives.

TL;DR: This work describes how the use of staged programming in Python, via source code transformation, offers a midpoint between these two library design patterns, capturing the benefits of both machine learning and imperative programming.