MIA: Josh Batson, Noise2Self: Blind denoising by self-supervision
May 6, 2018
MIA Meeting: https://www.youtube.com/watch?v=jwp1MsSXOZ4&list=PLlMMtlgw6qNjROoMNTBQjAcdx53kV50cS&index=96&t=0s
Josh Batson
Chan Zuckerberg Biohub
Blind denoising by self-supervision
Abstract: Modern biomedical science is defined by noisy high-dimensional data, whether from microscopes (electron, light-sheet, confocal), sequencing (RNA-seq, ATAC-seq, Hi-C), or sensors (physiology, EEG). We present a general framework for denoising high-dimensional measurements which can be applied to any of these domains, and which requires no prior on the signal, no estimate of the noise, and no clean training data. The only assumption is that the noise exhibits statistical independence across different dimensions of the measurement, while the true signal exhibits some correlation. For a broad class of functions ("J-invariant"), it is then possible to estimate the performance of a denoiser from noisy data alone. This allows us to calibrate J-invariant versions of any parameterised denoising algorithm, from the single hyperparameter of a median filter to the millions of weights of a deep neural network. We demonstrate this on natural image and microscopy data, where we exploit noise independence between pixels, and on single-cell gene expression data, where we exploit independence between detections of individual molecules. Finally, we prove a theoretical lower bound on the performance of an optimal denoiser. This framework generalizes recent work on training neural nets from noisy images and on cross-validation for matrix factorization. Preprint here:
https://arxiv.org/abs/1901.11365
MIA Meeting: https://youtu.be/jwp1MsSXOZ4
For more information on the Broad Institute and MIA visit: http://www.broadinstitute.org/MIA
Copyright Broad Institute, 2019. All rights reserved.
Видео MIA: Josh Batson, Noise2Self: Blind denoising by self-supervision канала Broad Institute
MIA Meeting: https://www.youtube.com/watch?v=jwp1MsSXOZ4&list=PLlMMtlgw6qNjROoMNTBQjAcdx53kV50cS&index=96&t=0s
Josh Batson
Chan Zuckerberg Biohub
Blind denoising by self-supervision
Abstract: Modern biomedical science is defined by noisy high-dimensional data, whether from microscopes (electron, light-sheet, confocal), sequencing (RNA-seq, ATAC-seq, Hi-C), or sensors (physiology, EEG). We present a general framework for denoising high-dimensional measurements which can be applied to any of these domains, and which requires no prior on the signal, no estimate of the noise, and no clean training data. The only assumption is that the noise exhibits statistical independence across different dimensions of the measurement, while the true signal exhibits some correlation. For a broad class of functions ("J-invariant"), it is then possible to estimate the performance of a denoiser from noisy data alone. This allows us to calibrate J-invariant versions of any parameterised denoising algorithm, from the single hyperparameter of a median filter to the millions of weights of a deep neural network. We demonstrate this on natural image and microscopy data, where we exploit noise independence between pixels, and on single-cell gene expression data, where we exploit independence between detections of individual molecules. Finally, we prove a theoretical lower bound on the performance of an optimal denoiser. This framework generalizes recent work on training neural nets from noisy images and on cross-validation for matrix factorization. Preprint here:
https://arxiv.org/abs/1901.11365
MIA Meeting: https://youtu.be/jwp1MsSXOZ4
For more information on the Broad Institute and MIA visit: http://www.broadinstitute.org/MIA
Copyright Broad Institute, 2019. All rights reserved.
Видео MIA: Josh Batson, Noise2Self: Blind denoising by self-supervision канала Broad Institute
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
![Screening Applications - Julia Joung](https://i.ytimg.com/vi/LYhIvmW5aoc/default.jpg)
![MPG Primer: Challenges and opportunities for developing more generalizable polygenic risk scores](https://i.ytimg.com/vi/kKcAnPznDE4/default.jpg)
![MPG Primer: Harmonization of deeply sequenced genomes (2023)](https://i.ytimg.com/vi/GKlm1VYqkhs/default.jpg)
![Broad Fellow: Ahmed Badran](https://i.ytimg.com/vi/sIIqagDL3OE/default.jpg)
![MIA: Travis Gibson, The dysbiotic microbiome as a dynamical system; Younhun Kim, Dirichlet processes](https://i.ytimg.com/vi/9YEWTQqEkro/default.jpg)
![Paola Arlotta: Studying the development of brain organoids](https://i.ytimg.com/vi/9eSOxjWrMds/default.jpg)
![The Drug Repurposing Hub at the Broad Institute](https://i.ytimg.com/vi/Ku4bV1OLlRc/default.jpg)
![MIA: Bo Xia, In silico screening for genome structure; Ruochi Zhang, graph representation learning](https://i.ytimg.com/vi/dNwL0VdMlF8/default.jpg)
![MPG Primer: Chronos: a dynamic model for CRISPR data (2022)](https://i.ytimg.com/vi/1j14DgmoBAM/default.jpg)
![Broad Institute Artist-In-Residence - Lucy Kim](https://i.ytimg.com/vi/AWuQ56rNLSk/default.jpg)
![3. How to link genetic constraint score from gene of interest to protein structure using G2P Portal](https://i.ytimg.com/vi/Ygt_PtWBvcI/default.jpg)
![Elinor Karlsson on the genetics of unusual traits](https://i.ytimg.com/vi/YNfGEj-CgQ0/default.jpg)
![Broad Discovery Series: Biomedical science and machine learning: A two-way street (2023)](https://i.ytimg.com/vi/XH7jXajzi7c/default.jpg)
![Broad Institute: The Rare Cancer Dependency Map](https://i.ytimg.com/vi/ubIqNjsg_1g/default.jpg)
![MPG Primer: Sparse dictionary learning recovers pleiotropy from human cell fitness screens (2022)](https://i.ytimg.com/vi/k0bJ-SybfiM/default.jpg)
![EWSC: Machine Learning about Sudden Cardiac Death from ECG Waveforms, Ziad Obermeyer](https://i.ytimg.com/vi/51VxB9KUXPY/default.jpg)
![MPG Primer: The Lipid Droplet Knowledge Portal (2022)](https://i.ytimg.com/vi/lL26j7NOlpE/default.jpg)
![MPG Primer: Detecting & tracking contamination in SARS-CoV-2 sequencing w/spike-in controls (2022)](https://i.ytimg.com/vi/ERQlDHSTiIs/default.jpg)
![Broad Institute Artist-In-Residence - Naoe Suzuki](https://i.ytimg.com/vi/PvvP8ThTsQI/default.jpg)
![MPG Primer: Introduction to Linear Models (2021)](https://i.ytimg.com/vi/CGz00PB58Jg/default.jpg)
![Genome editing and the germline: A conversation](https://i.ytimg.com/vi/POIeIILDo7k/default.jpg)