Image-to-Image Regression with Distribution-Free Uncertainty Quantification and Applications in Imaging

Anastasios N Angelopoulos, Amit Pal Kohli, Stephen Bates, Michael Jordan, Jitendra Malik, Thayer Alshaabi, Srigokul Upadhyayula, Yaniv Romano
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:717-730, 2022.

Abstract

Image-to-image regression is an important learning task, used frequently in biological imaging. Current algorithms, however, do not generally offer statistical guarantees that protect against a model’s mistakes and hallucinations. To address this, we develop uncertainty quantification techniques with rigorous statistical guarantees for image-to-image regression problems. In particular, we show how to derive uncertainty intervals around each pixel that are guaranteed to contain the true value with a user-specified confidence probability. Our methods work in conjunction with any base machine learning model, such as a neural network, and endow it with formal mathematical guarantees{—}regardless of the true unknown data distribution or choice of model. Furthermore, they are simple to implement and computationally inexpensive. We evaluate our procedure on three image-to-image regression tasks: quantitative phase microscopy, accelerated magnetic resonance imaging, and super-resolution transmission electron microscopy of a Drosophila melanogaster brain.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-angelopoulos22a, title = {Image-to-Image Regression with Distribution-Free Uncertainty Quantification and Applications in Imaging}, author = {Angelopoulos, Anastasios N and Kohli, Amit Pal and Bates, Stephen and Jordan, Michael and Malik, Jitendra and Alshaabi, Thayer and Upadhyayula, Srigokul and Romano, Yaniv}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {717--730}, 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/angelopoulos22a/angelopoulos22a.pdf}, url = {https://proceedings.mlr.press/v162/angelopoulos22a.html}, abstract = {Image-to-image regression is an important learning task, used frequently in biological imaging. Current algorithms, however, do not generally offer statistical guarantees that protect against a model’s mistakes and hallucinations. To address this, we develop uncertainty quantification techniques with rigorous statistical guarantees for image-to-image regression problems. In particular, we show how to derive uncertainty intervals around each pixel that are guaranteed to contain the true value with a user-specified confidence probability. Our methods work in conjunction with any base machine learning model, such as a neural network, and endow it with formal mathematical guarantees{—}regardless of the true unknown data distribution or choice of model. Furthermore, they are simple to implement and computationally inexpensive. We evaluate our procedure on three image-to-image regression tasks: quantitative phase microscopy, accelerated magnetic resonance imaging, and super-resolution transmission electron microscopy of a Drosophila melanogaster brain.} }
Endnote
%0 Conference Paper %T Image-to-Image Regression with Distribution-Free Uncertainty Quantification and Applications in Imaging %A Anastasios N Angelopoulos %A Amit Pal Kohli %A Stephen Bates %A Michael Jordan %A Jitendra Malik %A Thayer Alshaabi %A Srigokul Upadhyayula %A Yaniv Romano %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-angelopoulos22a %I PMLR %P 717--730 %U https://proceedings.mlr.press/v162/angelopoulos22a.html %V 162 %X Image-to-image regression is an important learning task, used frequently in biological imaging. Current algorithms, however, do not generally offer statistical guarantees that protect against a model’s mistakes and hallucinations. To address this, we develop uncertainty quantification techniques with rigorous statistical guarantees for image-to-image regression problems. In particular, we show how to derive uncertainty intervals around each pixel that are guaranteed to contain the true value with a user-specified confidence probability. Our methods work in conjunction with any base machine learning model, such as a neural network, and endow it with formal mathematical guarantees{—}regardless of the true unknown data distribution or choice of model. Furthermore, they are simple to implement and computationally inexpensive. We evaluate our procedure on three image-to-image regression tasks: quantitative phase microscopy, accelerated magnetic resonance imaging, and super-resolution transmission electron microscopy of a Drosophila melanogaster brain.
APA
Angelopoulos, A.N., Kohli, A.P., Bates, S., Jordan, M., Malik, J., Alshaabi, T., Upadhyayula, S. & Romano, Y.. (2022). Image-to-Image Regression with Distribution-Free Uncertainty Quantification and Applications in Imaging. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:717-730 Available from https://proceedings.mlr.press/v162/angelopoulos22a.html.

Related Material