Datamodels: Understanding Predictions with Data and Data with Predictions

Andrew Ilyas, Sung Min Park, Logan Engstrom, Guillaume Leclerc, Aleksander Madry
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:9525-9587, 2022.

Abstract

We present a conceptual framework, datamodeling, for analyzing the behavior of a model class in terms of the training data. For any fixed “target” example x, training set S, and learning algorithm, a datamodel is a parameterized function 2S→R that for any subset of S′⊂S—using only information about which examples of S are contained in S′—predicts the outcome of training a model on S′ and evaluating on x. Despite the complexity of the underlying process being approximated (e.g. end-to-end training and evaluation of deep neural networks), we show that even simple linear datamodels successfully predict model outputs. We then demonstrate that datamodels give rise to a variety of applications, such as: accurately predicting the effect of dataset counterfactuals; identifying brittle predictions; finding semantically similar examples; quantifying train-test leakage; and embedding data into a well-behaved and feature-rich representation space.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-ilyas22a, title = {Datamodels: Understanding Predictions with Data and Data with Predictions}, author = {Ilyas, Andrew and Park, Sung Min and Engstrom, Logan and Leclerc, Guillaume and Madry, Aleksander}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {9525--9587}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/ilyas22a/ilyas22a.pdf}, url = {https://proceedings.mlr.press/v162/ilyas22a.html}, abstract = {We present a conceptual framework, datamodeling, for analyzing the behavior of a model class in terms of the training data. For any fixed “target” example $x$, training set $S$, and learning algorithm, a datamodel is a parameterized function $2^S \to \mathbb{R}$ that for any subset of $S’ \subset S$—using only information about which examples of $S$ are contained in $S’$—predicts the outcome of training a model on $S’$ and evaluating on $x$. Despite the complexity of the underlying process being approximated (e.g. end-to-end training and evaluation of deep neural networks), we show that even simple linear datamodels successfully predict model outputs. We then demonstrate that datamodels give rise to a variety of applications, such as: accurately predicting the effect of dataset counterfactuals; identifying brittle predictions; finding semantically similar examples; quantifying train-test leakage; and embedding data into a well-behaved and feature-rich representation space.} }
Endnote
%0 Conference Paper %T Datamodels: Understanding Predictions with Data and Data with Predictions %A Andrew Ilyas %A Sung Min Park %A Logan Engstrom %A Guillaume Leclerc %A Aleksander Madry %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-ilyas22a %I PMLR %P 9525--9587 %U https://proceedings.mlr.press/v162/ilyas22a.html %V 162 %X We present a conceptual framework, datamodeling, for analyzing the behavior of a model class in terms of the training data. For any fixed “target” example $x$, training set $S$, and learning algorithm, a datamodel is a parameterized function $2^S \to \mathbb{R}$ that for any subset of $S’ \subset S$—using only information about which examples of $S$ are contained in $S’$—predicts the outcome of training a model on $S’$ and evaluating on $x$. Despite the complexity of the underlying process being approximated (e.g. end-to-end training and evaluation of deep neural networks), we show that even simple linear datamodels successfully predict model outputs. We then demonstrate that datamodels give rise to a variety of applications, such as: accurately predicting the effect of dataset counterfactuals; identifying brittle predictions; finding semantically similar examples; quantifying train-test leakage; and embedding data into a well-behaved and feature-rich representation space.
APA
Ilyas, A., Park, S.M., Engstrom, L., Leclerc, G. & Madry, A.. (2022). Datamodels: Understanding Predictions with Data and Data with Predictions. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:9525-9587 Available from https://proceedings.mlr.press/v162/ilyas22a.html.

Related Material