SlideShare a Scribd company logo
1 of 90
Download to read offline
Not Enough Measurements,
Too Many Measurements
Michael McCann
Department of Computational
Mathematics, Science and Engineering
Michigan State University
UM CSP Seminar, Oct. 22, 2020
2
My collaborators
●
Michael Unser and
BIG (EPFL)
– Kyong Jin
– Laurène Donati
– Harshit Gupta
●
Sai Ravishankar
and SLIM (MSU)
– Avrajit Ghosh
3
Goal: reconstruct meaningful
images from measurements
what we have:
measurements
what we want: image
The Physics
4
A linear example: X-ray CT
5
A linear example: X-ray CT
6
A linear example: X-ray CT
7
Not Enough Measurements
●
measurements are expensive
– time, radiation exposure, …
●
want to take M << N measurements
●
it’s hopeless!
8
Circa 2008-2016, the next slide
would have been...
9
Sparsity to the rescue!
10
Too Many Measurements!
●
e.g., The Cancer Imaging Archive
– 2223 subjects x hundreds of images
– includes Patient CT Projection Data Library
(AKA Mayo Clinic Data)
11
Goal: reconstruct meaningful
images from measurements PLUS a
training set
what we have:
training
what we have:
measurements
what we want: image
The Physics
12
Supervised image reconstruction
13
Supervised image reconstruction
●
What do we pick for D?
●
Where does the training data come from?
●
How do we solve the fitting problem?
●
What is the structure of F?
14
Themes in designing F
15
Themes in designing F
●
augment a direct method
– e.g., mix together different FBPs (Pelt et al. 2013)
denoise an FBP (Jin et al. 2017)
16
Themes in designing F
●
augment a direct method
– e.g., mix together different FBPs (Pelt et al. 2013)
denoise an FBP (Jin et al. 2017)
●
take inspiration from variational methods
– unrolling, plug-and-play, ...
– e.g., learn the regularization (Aggarwal et al. 2019;
Gupta et al. 2018)
learn the gradient (Adler et al. 2017),
learn filters and nonlinearities (Hammernik et al.
2017)
17
Themes in designing F
●
augment a direct method
– e.g., mix together different FBPs (Pelt et al. 2013)
denoise an FBP (Jin et al. 2017)
●
take inspiration from variational methods
– unrolling, plug-and-play, ...
– e.g., learn the regularization (Aggarwal et al. 2019; Gupta et
al. 2018)
learn the gradient (Adler et al. 2017),
learn filters and nonlinearities (Hammernik et al. 2017)
●
throw away variational methods
– e.g., learn the entire measurement to image mapping (Zhu et
al. 2018)
18
Themes in designing F
●
augment a direct method
– e.g., mix together different FBPs (Pelt et al. 2013)
denoise an FBP (Jin et al. 2017)
●
take inspiration from variational methods
– unrolling, plug-and-play, ...
– e.g., learn the regularization (Aggarwal et al. 2019; Gupta et al. 2018)
learn the gradient (Adler et al. 2017),
learn filters and nonlinearities (Hammernik et al. 2017)
●
throw away variational methods
– e.g., learn the entire measurement to image mapping (Zhu et al. 2018)
●
work in the data domain
– e.g., learn to inpaint missing measurements (Ghani et al. 2019)
19
Themes in designing F
●
augment a direct method
– e.g., mix together different FBPs (Pelt et al. 2013)
denoise an FBP (Jin et al. 2017)
●
take inspiration from variational methods
– unrolling, plug-and-play, ...
– e.g., learn the regularization (Aggarwal et al. 2019; Gupta et al. 2018)
learn the gradient (Adler et al. 2017),
learn filters and nonlinearities (Hammernik et al. 2017)
●
throw away variational methods
– e.g., learn the entire measurement to image mapping (Zhu et al. 2018)
●
work in the data domain
– e.g., learn to inpaint missing measurements (Ghani et al. 2019)
●
many more...
20
Results: biomedical images
(Jin, McCann, Froustey, Unser 2017)
21
Results: biomedical images
(Jin, McCann, Froustey, Unser 2017)
22
Results: biomedical images
(Jin, McCann, Froustey, Unser 2017)
23
Perspectives
24
Perspectives
●
It’s not about H, it’s about 𝓗
25
Perspectives
●
It’s not about H, it’s about 𝓗
●
If we care about MSE, let’s do challenges
– fastMRI, Low Dose CT Grand Challenge, others?
26
Perspectives
●
It’s not about H, it’s about 𝓗
●
If we care about MSE, let’s do challenges
– fastMRI, Low Dose CT Grand Challenge, others?
●
Let’s not forget data fidelity and robustness
ground truth FBPConvNet
27
Summary so far
X-ray CT
low-dose X-ray CT
(“impossible”)
28
Summary so far
X-ray CT
supervised
low-dose X-ray CT
29
Next topic
X-ray CT
supervised
low-dose X-ray CT
30
Next topic
X-ray CT
supervised
low-dose X-ray CT
31
Next topic
X-ray CT
supervised
low-dose X-ray CT
32
Single-particle cryo-EM
●
~10^6 projections of a single particle
●
random orientations, (electron) optical effects,
SNR ≈ 0 dB
●
subnanometer resolution
33
Single-particle cryo-EM
●
~10^6 projections of a single particle
●
random orientations, (electron) optical effects,
SNR ≈ 0 dB
●
subnanometer resolution
34
Single-particle cryo-EM
●
~10^6 projections of a single particle
●
random orientations, (electron) optical effects,
SNR ≈ 0 dB
●
subnanometer resolution
35
Single-particle cryo-EM
●
~10^6 projections of a single particle
●
random orientations, (electron) optical effects,
SNR ≈ 0 dB
●
subnanometer resolution
right: Wu et al. 2020
36
Single-particle cryo-EM
what we have:
measurements
what we want:
image
The Physics
unknown
nuisance
37
Cryo-EM reconstruction
●
projection matching (Penczek et al. 1994)
– # of unknowns grows with data :(
●
marginalized maximum likelihood (Sigworth 1998)
– marginalization is computationally heavy
– what most software currently does
●
method of moments (Kam 1980; Sharon et al. 2020)
– avoids above problems
– makes efficient use of data
38
Generative adversarial networks
●
given samples of a distribution, generate new
samples from the same distribution
image: Goodfellow et al. 2014
39
Generative adversarial networks
●
given samples of a distribution, generate new
samples from the same distribution
●
parameterization of the generator is flexible
image: Goodfellow et al. 2014
40
Typical GAN
(Gupta, McCann, Donati, Unser. 2020)
41
Typical GAN
(Gupta, McCann, Donati, Unser. 2020)
42
Typical GAN
(Gupta, McCann, Donati, Unser. 2020)
43
(Gupta et al. 2020)
44
thisprojectiondoesnotexist.com
Not obviously useful!
(Gupta et al. 2020)
45
CryoGAN
(Gupta et al. 2020)
46
(Gupta et al. 2020)
47
(Gupta et al. 2020)
48
for more on CryoGAN, SPACE Webinar: https://
www.youtube.com/watch?v=J6UZBeU3Bm0
(Gupta et al. 2020)
49
Intermission
50
Back to low-dose X-ray CT
(Jin et al. 2017)
51
Back to low-dose X-ray CT
(Jin et al. 2017)
52
Back to low-dose X-ray CT
(Jin et al. 2017)
53
A “that’s funny…”moment
figures and table: Adler et al. 2017
54
Can we learn a regularizer from
data?
●
recall supervised reconstruction
●
our proposed “architecture”
55
Why learn a regularizer?
56
Why learn a regularizer?
●
can’t be worse than sparsity-based
reconstruction
57
Why learn a regularizer?
●
can’t be worse than sparsity-based
reconstruction
●
results in a a convex problem
– robust, includes data fidelity, decades of theory
58
Why learn a regularizer?
●
can’t be worse than sparsity-based
reconstruction
●
results in a a convex problem
– robust, includes data fidelity, decades of theory
●
joins proven architecture with data adaptivity
59
Why learn a regularizer?
●
can’t be worse than sparsity-based reconstruction
●
results in a a convex problem
– robust, includes data fidelity, decades of theory
●
joins proven architecture with data adaptivity
●
gives a hope of interpreting the learned part
60
Related work
●
Main approach: relax the l1 term
– Peyré et al. 2011; Mairal et al. 2012; Sprechmann et al. 2013; Chen et al. 2014
61
Related work
●
Main approach: relax the l1 term
– Peyré et al. 2011; Mairal et al. 2012; Sprechmann et al. 2013; Chen et al. 2014
62
Related work
●
Main approach: relax the l1 term
– Peyré et al. 2011; Mairal et al. 2012; Sprechmann et al. 2013; Chen et al. 2014
63
Related work
●
Main approach: relax the l1 term
– Peyré et al. 2011; Mairal et al. 2012; Sprechmann et al. 2013; Chen et al. 2014
64
Our approach (outline)
65
Our approach (outline)
1. solve the lower problem at the current W
66
Our approach (outline)
1. solve the lower problem at the current W
2. find a (local) closed-form solution, x*(W)
67
Our approach (outline)
1. solve the lower problem at the current W
2. find a (local) closed-form solution, x*(W)
3. substitute x*(W) into the upper level problem,
compute a gradient w.r.t. W, and descend
68
1. Solve the lower level problem
●
standard, convex problem
– ADMM, Chambolle-Pock
●
need fast, accurate solutions
– hyperparameter selection is tough
69
2. Find a closed-form solution
70
2. Find a closed-form solution
●
uniqueness? no, but
71
2. Find a closed-form solution
●
uniqueness? no, but
●
intuition: in a region (of W-space) where the
sign pattern of Wx*(W) does not change,
||Wx||1 is linear
72
2. Find a closed-form solution
●
uniqueness? no, but
●
intuition: in a region (of W-space) where the
sign pattern of Wx*(W) does not change,
||Wx||1 is linear
– see McCann and Ravishankar 2020 for A = I
73
2. Find a closed-form solution
●
from McCann and Ravishankar 2020
74
2. Find a closed-form solution
●
from McCann and Ravishankar 2020
●
even better, Ali and Tibshirani 2019
75
2. Find a closed-form solution
●
from McCann and Ravishankar 2020
●
even better, Ali and Tibshirani 2019
– unique minimum norm solution when b=0
76
3. Find the gradient
77
3. Find the gradient
●
with 1d signals, pytorch can autograd these
expressions
– Agrawal et al. 2019 makes it even easier
78
3. Find the gradient
●
with 1d signals, pytorch can autograd these
expressions
– Agrawal et al. 2019 makes it even easier
●
with images, things get tough
– W is potentially million X million
– no more SVDs or explicit inverses
– our solution: by hand + pytorch
79
Early experiments
●
image denoising
●
W is a set of 8, 3x3 convolutions
●
compare to
– BM3D, TV, DCT
– unsupervised learned regularizer
●
training
– SGD with increasing batch size
– takes ~hours
80
Early results
81
Comparison with TV
82
Comparison with DCT
83
Absolute summed filter responses
and filters
●
Learned filters
– Are not orthonormal (neither ortho- nor -normal)
– Penalize edges less than DCT
84
Taking a step back
85
Taking a step back
●
CNN performance on image reconstruction is
due to
– the CNN (architecture)
– training
86
Taking a step back
●
CNN performance on image reconstruction is
due to
– the CNN (architecture)
– training
●
remove the training: deep image prior
– Reinhard Heckel 1W-MINDS seminar:
https://www.youtube.com/watch?v=AvJgmbeupGY
87
Taking a step back
●
CNN performance on image reconstruction is
due to
– the CNN (architecture)
– training
●
remove the training: deep image prior
– Reinhard Heckel 1W-MINDS seminar:
https://www.youtube.com/watch?v=AvJgmbeupGY
●
remove the CNN: learned regularizer
88
Thanks for your attention!
michael.thompson.mccann@gmail.com
slides:
89
References
●
M. Wu, G. C. Lander, and M. A. Herzik, “Sub-2 Angstrom resolution structure determination using single-
particle cryo-EM at 200 keV,” Journal of Structural Biology: X, vol. 4, p. 100020, 2020, doi:
10.1016/j.yjsbx.2020.100020.
●
Pawel A. Penczek, Robert A. Grassucci, and Joachim Frank. “The ribosome at improved resolution:
New techniques for merging and orientation refinement in 3D cryo-electron microscopy of biological
particles”. In:Ultramicroscopy 53.3 (1994), pp. 251–270.
●
Fred J Sigworth. “A maximum-likelihood approach to single-particle image refinement”. In:Journal of
structural biology 122.3 (1998), pp. 328–339.
●
Z. Kam, “The reconstruction of structure from electron micrographs of randomly oriented particles,” Journal
of Theoretical Biology, vol. 82, no. 1, Art. no. 1, Jan. 1980, doi: 10.1016/0022-5193(80)90088-0.
●
N. Sharon, J. Kileel, Y. Khoo, B. Landa, and A. Singer, “Method of moments for 3D single particle ab initio
modeling with non-uniform distribution of viewing angles,” Inverse Problems, vol. 36, no. 4, Art. no. 4, Feb.
2020, doi: 10.1088/1361-6420/ab6139.
●
I. Goodfellow et al., “Generative Adversarial Nets,” in Advances in Neural Information Processing Systems
27, Ed.Z. Ghahramani, M. Welling, C. Cortes, N. D. Lawrence, and K. Q. Weinberger Curran Associates,
Inc., 2014, pp. 2672–2680.
●
J. Adler and O. Öktem, “Solving ill-posed inverse problems using iterative deep neural networks,” Inverse
Problems, vol. 33, no. 12, p. 124007, Nov. 2017.
90
References
●
H. Gupta, M. T. McCann, L. Donati, and M. Unser, “CryoGAN: A New Reconstruction Paradigm for
Single-Particle Cryo-EM via Deep Adversarial Learning,” bioRxiv, Mar. 2020, doi:
10.1101/2020.03.20.001016.
●
G. Peyré and J. M. Fadili, “Learning analysis sparsity priors,” in Sampling Theory and Applications,
Singapore, Singapore, May 2011, p. 4.
●
J. Mairal, F. Bach, and J. Ponce, “Task-driven dictionary learning,”IEEE Transactions on Pattern Analysis
and Machine Intelligence, vol. 34, no. 4, pp. 791–804, Apr. 2012.
●
P. Sprechmann, R. Litman, T. Ben Yakar, A. M. Bronstein, and G. Sapiro, “Supervised sparse analysis
and synthesis operators,” in Advances in Neural Information Processing Systems 26, 2013, pp. 908–
916.
●
M. T. McCann and S. Ravishankar, “Supervised Learning of Sparsity-Promoting Regularizers for
Denoising,” arXiv:2006.05521 [eess.IV], 2020-06-09.
●
A. Ali and R. J. Tibshirani, “The Generalized Lasso Problem and Uniqueness,” Electronic Journal of
Statistics, vol. 13, no. 2, Art. no. 2, 2019, doi: 10.1214/19-ejs1569.
●
A. Agrawal, B. Amos, S. Barratt, S. Boyd, S. Diamond, and J. Z. Kolter, “Differentiable Convex
Optimization Layers,” in Advances in Neural Information Processing Systems 32, Ed.H. Wallach, H.
Larochelle, A. Beygelzimer, F. dAlché-Buc, E. Fox, and R. Garnett Curran Associates, Inc., 2019, pp.
9562–9574.

More Related Content

What's hot

Dependent processes in Bayesian Nonparametrics
Dependent processes in Bayesian NonparametricsDependent processes in Bayesian Nonparametrics
Dependent processes in Bayesian NonparametricsJulyan Arbel
 
4 tracking objects of deformable shapes
4 tracking objects of deformable shapes4 tracking objects of deformable shapes
4 tracking objects of deformable shapesprjpublications
 
Tensorflow KR PR12(Season 3) : 251th Paper Review
Tensorflow KR PR12(Season 3) : 251th Paper ReviewTensorflow KR PR12(Season 3) : 251th Paper Review
Tensorflow KR PR12(Season 3) : 251th Paper ReviewChanghoon Jeong
 
Fusion Based Gaussian noise Removal in the Images using Curvelets and Wavelet...
Fusion Based Gaussian noise Removal in the Images using Curvelets and Wavelet...Fusion Based Gaussian noise Removal in the Images using Curvelets and Wavelet...
Fusion Based Gaussian noise Removal in the Images using Curvelets and Wavelet...CSCJournals
 
A Method for Automatic Detection of the Square Piece In The Borderless Square...
A Method for Automatic Detection of the Square Piece In The Borderless Square...A Method for Automatic Detection of the Square Piece In The Borderless Square...
A Method for Automatic Detection of the Square Piece In The Borderless Square...CSCJournals
 
Chapter2: Likelihood-based approach
Chapter2: Likelihood-based approach Chapter2: Likelihood-based approach
Chapter2: Likelihood-based approach Jae-kwang Kim
 
Comparison of the optimal design
Comparison of the optimal designComparison of the optimal design
Comparison of the optimal designAlexander Decker
 
Exploring Global Reflection Symmetry in Visual Arts
Exploring Global Reflection Symmetry in Visual ArtsExploring Global Reflection Symmetry in Visual Arts
Exploring Global Reflection Symmetry in Visual ArtsMohamed Elawady
 
Object Shape Representation by Kernel Density Feature Points Estimator
Object Shape Representation by Kernel Density Feature Points Estimator Object Shape Representation by Kernel Density Feature Points Estimator
Object Shape Representation by Kernel Density Feature Points Estimator cscpconf
 
Introduction of Inverse Problem and Its Applications
Introduction of Inverse Problem and Its ApplicationsIntroduction of Inverse Problem and Its Applications
Introduction of Inverse Problem and Its ApplicationsKomal Goyal
 
Bayesian Graphical Models for Structural Vector Autoregressive Processes - D....
Bayesian Graphical Models for Structural Vector Autoregressive Processes - D....Bayesian Graphical Models for Structural Vector Autoregressive Processes - D....
Bayesian Graphical Models for Structural Vector Autoregressive Processes - D....SYRTO Project
 
A Hybrid Technique for the Automated Segmentation of Corpus Callosum in Midsa...
A Hybrid Technique for the Automated Segmentation of Corpus Callosum in Midsa...A Hybrid Technique for the Automated Segmentation of Corpus Callosum in Midsa...
A Hybrid Technique for the Automated Segmentation of Corpus Callosum in Midsa...IJERA Editor
 

What's hot (14)

Dependent processes in Bayesian Nonparametrics
Dependent processes in Bayesian NonparametricsDependent processes in Bayesian Nonparametrics
Dependent processes in Bayesian Nonparametrics
 
4 tracking objects of deformable shapes
4 tracking objects of deformable shapes4 tracking objects of deformable shapes
4 tracking objects of deformable shapes
 
Tensorflow KR PR12(Season 3) : 251th Paper Review
Tensorflow KR PR12(Season 3) : 251th Paper ReviewTensorflow KR PR12(Season 3) : 251th Paper Review
Tensorflow KR PR12(Season 3) : 251th Paper Review
 
Fusion Based Gaussian noise Removal in the Images using Curvelets and Wavelet...
Fusion Based Gaussian noise Removal in the Images using Curvelets and Wavelet...Fusion Based Gaussian noise Removal in the Images using Curvelets and Wavelet...
Fusion Based Gaussian noise Removal in the Images using Curvelets and Wavelet...
 
A Method for Automatic Detection of the Square Piece In The Borderless Square...
A Method for Automatic Detection of the Square Piece In The Borderless Square...A Method for Automatic Detection of the Square Piece In The Borderless Square...
A Method for Automatic Detection of the Square Piece In The Borderless Square...
 
Chapter2: Likelihood-based approach
Chapter2: Likelihood-based approach Chapter2: Likelihood-based approach
Chapter2: Likelihood-based approach
 
Comparison of the optimal design
Comparison of the optimal designComparison of the optimal design
Comparison of the optimal design
 
Exploring Global Reflection Symmetry in Visual Arts
Exploring Global Reflection Symmetry in Visual ArtsExploring Global Reflection Symmetry in Visual Arts
Exploring Global Reflection Symmetry in Visual Arts
 
MUMS: Bayesian, Fiducial, and Frequentist Conference - Spatially Informed Var...
MUMS: Bayesian, Fiducial, and Frequentist Conference - Spatially Informed Var...MUMS: Bayesian, Fiducial, and Frequentist Conference - Spatially Informed Var...
MUMS: Bayesian, Fiducial, and Frequentist Conference - Spatially Informed Var...
 
Object Shape Representation by Kernel Density Feature Points Estimator
Object Shape Representation by Kernel Density Feature Points Estimator Object Shape Representation by Kernel Density Feature Points Estimator
Object Shape Representation by Kernel Density Feature Points Estimator
 
Introduction of Inverse Problem and Its Applications
Introduction of Inverse Problem and Its ApplicationsIntroduction of Inverse Problem and Its Applications
Introduction of Inverse Problem and Its Applications
 
Propensity albert
Propensity albertPropensity albert
Propensity albert
 
Bayesian Graphical Models for Structural Vector Autoregressive Processes - D....
Bayesian Graphical Models for Structural Vector Autoregressive Processes - D....Bayesian Graphical Models for Structural Vector Autoregressive Processes - D....
Bayesian Graphical Models for Structural Vector Autoregressive Processes - D....
 
A Hybrid Technique for the Automated Segmentation of Corpus Callosum in Midsa...
A Hybrid Technique for the Automated Segmentation of Corpus Callosum in Midsa...A Hybrid Technique for the Automated Segmentation of Corpus Callosum in Midsa...
A Hybrid Technique for the Automated Segmentation of Corpus Callosum in Midsa...
 

Similar to Not Enough Measurements, Too Many Measurements

Supervised Learning of Sparsity-Promoting Regularizers for Denoising
Supervised Learning of Sparsity-Promoting Regularizers for DenoisingSupervised Learning of Sparsity-Promoting Regularizers for Denoising
Supervised Learning of Sparsity-Promoting Regularizers for DenoisingMike McCann
 
TPDM Presentation Slide (ICCV23)
TPDM Presentation Slide (ICCV23)TPDM Presentation Slide (ICCV23)
TPDM Presentation Slide (ICCV23)Suhyeon Lee
 
Style gan2 review
Style gan2 reviewStyle gan2 review
Style gan2 reviewtaeseon ryu
 
Bayesian Generalization Error and Real Log Canonical Threshold in Non-negativ...
Bayesian Generalization Error and Real Log Canonical Threshold in Non-negativ...Bayesian Generalization Error and Real Log Canonical Threshold in Non-negativ...
Bayesian Generalization Error and Real Log Canonical Threshold in Non-negativ...Naoki Hayashi
 
Lec7: Medical Image Segmentation (I) (Radiology Applications of Segmentation,...
Lec7: Medical Image Segmentation (I) (Radiology Applications of Segmentation,...Lec7: Medical Image Segmentation (I) (Radiology Applications of Segmentation,...
Lec7: Medical Image Segmentation (I) (Radiology Applications of Segmentation,...Ulaş Bağcı
 
Adaptive physics–inspired facial animation
Adaptive physics–inspired facial animationAdaptive physics–inspired facial animation
Adaptive physics–inspired facial animationRichard Southern
 
Introduction geostatistic for_mineral_resources
Introduction geostatistic for_mineral_resourcesIntroduction geostatistic for_mineral_resources
Introduction geostatistic for_mineral_resourcesAdi Handarbeni
 
Finite Element Analysis in Metal Forming processes
Finite Element Analysis in Metal Forming processesFinite Element Analysis in Metal Forming processes
Finite Element Analysis in Metal Forming processesabhishek_hukkerikar
 
Large Scale Recommendation: a view from the Trenches
Large Scale Recommendation: a view from the TrenchesLarge Scale Recommendation: a view from the Trenches
Large Scale Recommendation: a view from the TrenchesAnne-Marie Tousch
 
Computational Giants_nhom.pptx
Computational Giants_nhom.pptxComputational Giants_nhom.pptx
Computational Giants_nhom.pptxThAnhonc
 
Optimal Control Problem and Power-Efficient Medical Image Processing Using Puma
Optimal Control Problem and Power-Efficient Medical Image Processing Using PumaOptimal Control Problem and Power-Efficient Medical Image Processing Using Puma
Optimal Control Problem and Power-Efficient Medical Image Processing Using PumaIJMER
 
SPIE 10059-36(Reheman Baikejiang)
SPIE 10059-36(Reheman Baikejiang)SPIE 10059-36(Reheman Baikejiang)
SPIE 10059-36(Reheman Baikejiang)Reheman Baikejiang
 
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...Chris Rackauckas
 
Tutorial on Theory and Application of Generative Adversarial Networks
Tutorial on Theory and Application of Generative Adversarial NetworksTutorial on Theory and Application of Generative Adversarial Networks
Tutorial on Theory and Application of Generative Adversarial NetworksMLReview
 
Image segmentation methods for brain mri images
Image segmentation methods for brain mri imagesImage segmentation methods for brain mri images
Image segmentation methods for brain mri imageseSAT Journals
 
Numerical and experimental investigation on existing structures: two seminars
Numerical and experimental investigation on existing structures: two seminarsNumerical and experimental investigation on existing structures: two seminars
Numerical and experimental investigation on existing structures: two seminarsPhD ISG, Sapienza University of Rome
 

Similar to Not Enough Measurements, Too Many Measurements (20)

Supervised Learning of Sparsity-Promoting Regularizers for Denoising
Supervised Learning of Sparsity-Promoting Regularizers for DenoisingSupervised Learning of Sparsity-Promoting Regularizers for Denoising
Supervised Learning of Sparsity-Promoting Regularizers for Denoising
 
TPDM Presentation Slide (ICCV23)
TPDM Presentation Slide (ICCV23)TPDM Presentation Slide (ICCV23)
TPDM Presentation Slide (ICCV23)
 
Style gan2 review
Style gan2 reviewStyle gan2 review
Style gan2 review
 
riken-RBlur-slides.pptx
riken-RBlur-slides.pptxriken-RBlur-slides.pptx
riken-RBlur-slides.pptx
 
Bayesian Generalization Error and Real Log Canonical Threshold in Non-negativ...
Bayesian Generalization Error and Real Log Canonical Threshold in Non-negativ...Bayesian Generalization Error and Real Log Canonical Threshold in Non-negativ...
Bayesian Generalization Error and Real Log Canonical Threshold in Non-negativ...
 
Lec7: Medical Image Segmentation (I) (Radiology Applications of Segmentation,...
Lec7: Medical Image Segmentation (I) (Radiology Applications of Segmentation,...Lec7: Medical Image Segmentation (I) (Radiology Applications of Segmentation,...
Lec7: Medical Image Segmentation (I) (Radiology Applications of Segmentation,...
 
Adaptive physics–inspired facial animation
Adaptive physics–inspired facial animationAdaptive physics–inspired facial animation
Adaptive physics–inspired facial animation
 
Introduction geostatistic for_mineral_resources
Introduction geostatistic for_mineral_resourcesIntroduction geostatistic for_mineral_resources
Introduction geostatistic for_mineral_resources
 
Finite Element Analysis in Metal Forming processes
Finite Element Analysis in Metal Forming processesFinite Element Analysis in Metal Forming processes
Finite Element Analysis in Metal Forming processes
 
Large Scale Recommendation: a view from the Trenches
Large Scale Recommendation: a view from the TrenchesLarge Scale Recommendation: a view from the Trenches
Large Scale Recommendation: a view from the Trenches
 
Computational Giants_nhom.pptx
Computational Giants_nhom.pptxComputational Giants_nhom.pptx
Computational Giants_nhom.pptx
 
Optimal Control Problem and Power-Efficient Medical Image Processing Using Puma
Optimal Control Problem and Power-Efficient Medical Image Processing Using PumaOptimal Control Problem and Power-Efficient Medical Image Processing Using Puma
Optimal Control Problem and Power-Efficient Medical Image Processing Using Puma
 
SPIE 10059-36(Reheman Baikejiang)
SPIE 10059-36(Reheman Baikejiang)SPIE 10059-36(Reheman Baikejiang)
SPIE 10059-36(Reheman Baikejiang)
 
Ds33717725
Ds33717725Ds33717725
Ds33717725
 
Ds33717725
Ds33717725Ds33717725
Ds33717725
 
Thesis presentation
Thesis presentationThesis presentation
Thesis presentation
 
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
 
Tutorial on Theory and Application of Generative Adversarial Networks
Tutorial on Theory and Application of Generative Adversarial NetworksTutorial on Theory and Application of Generative Adversarial Networks
Tutorial on Theory and Application of Generative Adversarial Networks
 
Image segmentation methods for brain mri images
Image segmentation methods for brain mri imagesImage segmentation methods for brain mri images
Image segmentation methods for brain mri images
 
Numerical and experimental investigation on existing structures: two seminars
Numerical and experimental investigation on existing structures: two seminarsNumerical and experimental investigation on existing structures: two seminars
Numerical and experimental investigation on existing structures: two seminars
 

Recently uploaded

21scheme vtu syllabus of visveraya technological university
21scheme vtu syllabus of visveraya technological university21scheme vtu syllabus of visveraya technological university
21scheme vtu syllabus of visveraya technological universityMohd Saifudeen
 
SLIDESHARE PPT-DECISION MAKING METHODS.pptx
SLIDESHARE PPT-DECISION MAKING METHODS.pptxSLIDESHARE PPT-DECISION MAKING METHODS.pptx
SLIDESHARE PPT-DECISION MAKING METHODS.pptxCHAIRMAN M
 
NO1 Best Powerful Vashikaran Specialist Baba Vashikaran Specialist For Love V...
NO1 Best Powerful Vashikaran Specialist Baba Vashikaran Specialist For Love V...NO1 Best Powerful Vashikaran Specialist Baba Vashikaran Specialist For Love V...
NO1 Best Powerful Vashikaran Specialist Baba Vashikaran Specialist For Love V...Amil baba
 
The Entity-Relationship Model(ER Diagram).pptx
The Entity-Relationship Model(ER Diagram).pptxThe Entity-Relationship Model(ER Diagram).pptx
The Entity-Relationship Model(ER Diagram).pptxMANASINANDKISHORDEOR
 
engineering chemistry power point presentation
engineering chemistry  power point presentationengineering chemistry  power point presentation
engineering chemistry power point presentationsj9399037128
 
Maher Othman Interior Design Portfolio..
Maher Othman Interior Design Portfolio..Maher Othman Interior Design Portfolio..
Maher Othman Interior Design Portfolio..MaherOthman7
 
Involute of a circle,Square, pentagon,HexagonInvolute_Engineering Drawing.pdf
Involute of a circle,Square, pentagon,HexagonInvolute_Engineering Drawing.pdfInvolute of a circle,Square, pentagon,HexagonInvolute_Engineering Drawing.pdf
Involute of a circle,Square, pentagon,HexagonInvolute_Engineering Drawing.pdfJNTUA
 
Seizure stage detection of epileptic seizure using convolutional neural networks
Seizure stage detection of epileptic seizure using convolutional neural networksSeizure stage detection of epileptic seizure using convolutional neural networks
Seizure stage detection of epileptic seizure using convolutional neural networksIJECEIAES
 
Seismic Hazard Assessment Software in Python by Prof. Dr. Costas Sachpazis
Seismic Hazard Assessment Software in Python by Prof. Dr. Costas SachpazisSeismic Hazard Assessment Software in Python by Prof. Dr. Costas Sachpazis
Seismic Hazard Assessment Software in Python by Prof. Dr. Costas SachpazisDr.Costas Sachpazis
 
Fuzzy logic method-based stress detector with blood pressure and body tempera...
Fuzzy logic method-based stress detector with blood pressure and body tempera...Fuzzy logic method-based stress detector with blood pressure and body tempera...
Fuzzy logic method-based stress detector with blood pressure and body tempera...IJECEIAES
 
Artificial intelligence presentation2-171219131633.pdf
Artificial intelligence presentation2-171219131633.pdfArtificial intelligence presentation2-171219131633.pdf
Artificial intelligence presentation2-171219131633.pdfKira Dess
 
Theory of Time 2024 (Universal Theory for Everything)
Theory of Time 2024 (Universal Theory for Everything)Theory of Time 2024 (Universal Theory for Everything)
Theory of Time 2024 (Universal Theory for Everything)Ramkumar k
 
Software Engineering Practical File Front Pages.pdf
Software Engineering Practical File Front Pages.pdfSoftware Engineering Practical File Front Pages.pdf
Software Engineering Practical File Front Pages.pdfssuser5c9d4b1
 
What is Coordinate Measuring Machine? CMM Types, Features, Functions
What is Coordinate Measuring Machine? CMM Types, Features, FunctionsWhat is Coordinate Measuring Machine? CMM Types, Features, Functions
What is Coordinate Measuring Machine? CMM Types, Features, FunctionsVIEW
 
Autodesk Construction Cloud (Autodesk Build).pptx
Autodesk Construction Cloud (Autodesk Build).pptxAutodesk Construction Cloud (Autodesk Build).pptx
Autodesk Construction Cloud (Autodesk Build).pptxMustafa Ahmed
 
History of Indian Railways - the story of Growth & Modernization
History of Indian Railways - the story of Growth & ModernizationHistory of Indian Railways - the story of Growth & Modernization
History of Indian Railways - the story of Growth & ModernizationEmaan Sharma
 
electrical installation and maintenance.
electrical installation and maintenance.electrical installation and maintenance.
electrical installation and maintenance.benjamincojr
 
Augmented Reality (AR) with Augin Software.pptx
Augmented Reality (AR) with Augin Software.pptxAugmented Reality (AR) with Augin Software.pptx
Augmented Reality (AR) with Augin Software.pptxMustafa Ahmed
 
CLOUD COMPUTING SERVICES - Cloud Reference Modal
CLOUD COMPUTING SERVICES - Cloud Reference ModalCLOUD COMPUTING SERVICES - Cloud Reference Modal
CLOUD COMPUTING SERVICES - Cloud Reference ModalSwarnaSLcse
 
Final DBMS Manual (2).pdf final lab manual
Final DBMS Manual (2).pdf final lab manualFinal DBMS Manual (2).pdf final lab manual
Final DBMS Manual (2).pdf final lab manualBalamuruganV28
 

Recently uploaded (20)

21scheme vtu syllabus of visveraya technological university
21scheme vtu syllabus of visveraya technological university21scheme vtu syllabus of visveraya technological university
21scheme vtu syllabus of visveraya technological university
 
SLIDESHARE PPT-DECISION MAKING METHODS.pptx
SLIDESHARE PPT-DECISION MAKING METHODS.pptxSLIDESHARE PPT-DECISION MAKING METHODS.pptx
SLIDESHARE PPT-DECISION MAKING METHODS.pptx
 
NO1 Best Powerful Vashikaran Specialist Baba Vashikaran Specialist For Love V...
NO1 Best Powerful Vashikaran Specialist Baba Vashikaran Specialist For Love V...NO1 Best Powerful Vashikaran Specialist Baba Vashikaran Specialist For Love V...
NO1 Best Powerful Vashikaran Specialist Baba Vashikaran Specialist For Love V...
 
The Entity-Relationship Model(ER Diagram).pptx
The Entity-Relationship Model(ER Diagram).pptxThe Entity-Relationship Model(ER Diagram).pptx
The Entity-Relationship Model(ER Diagram).pptx
 
engineering chemistry power point presentation
engineering chemistry  power point presentationengineering chemistry  power point presentation
engineering chemistry power point presentation
 
Maher Othman Interior Design Portfolio..
Maher Othman Interior Design Portfolio..Maher Othman Interior Design Portfolio..
Maher Othman Interior Design Portfolio..
 
Involute of a circle,Square, pentagon,HexagonInvolute_Engineering Drawing.pdf
Involute of a circle,Square, pentagon,HexagonInvolute_Engineering Drawing.pdfInvolute of a circle,Square, pentagon,HexagonInvolute_Engineering Drawing.pdf
Involute of a circle,Square, pentagon,HexagonInvolute_Engineering Drawing.pdf
 
Seizure stage detection of epileptic seizure using convolutional neural networks
Seizure stage detection of epileptic seizure using convolutional neural networksSeizure stage detection of epileptic seizure using convolutional neural networks
Seizure stage detection of epileptic seizure using convolutional neural networks
 
Seismic Hazard Assessment Software in Python by Prof. Dr. Costas Sachpazis
Seismic Hazard Assessment Software in Python by Prof. Dr. Costas SachpazisSeismic Hazard Assessment Software in Python by Prof. Dr. Costas Sachpazis
Seismic Hazard Assessment Software in Python by Prof. Dr. Costas Sachpazis
 
Fuzzy logic method-based stress detector with blood pressure and body tempera...
Fuzzy logic method-based stress detector with blood pressure and body tempera...Fuzzy logic method-based stress detector with blood pressure and body tempera...
Fuzzy logic method-based stress detector with blood pressure and body tempera...
 
Artificial intelligence presentation2-171219131633.pdf
Artificial intelligence presentation2-171219131633.pdfArtificial intelligence presentation2-171219131633.pdf
Artificial intelligence presentation2-171219131633.pdf
 
Theory of Time 2024 (Universal Theory for Everything)
Theory of Time 2024 (Universal Theory for Everything)Theory of Time 2024 (Universal Theory for Everything)
Theory of Time 2024 (Universal Theory for Everything)
 
Software Engineering Practical File Front Pages.pdf
Software Engineering Practical File Front Pages.pdfSoftware Engineering Practical File Front Pages.pdf
Software Engineering Practical File Front Pages.pdf
 
What is Coordinate Measuring Machine? CMM Types, Features, Functions
What is Coordinate Measuring Machine? CMM Types, Features, FunctionsWhat is Coordinate Measuring Machine? CMM Types, Features, Functions
What is Coordinate Measuring Machine? CMM Types, Features, Functions
 
Autodesk Construction Cloud (Autodesk Build).pptx
Autodesk Construction Cloud (Autodesk Build).pptxAutodesk Construction Cloud (Autodesk Build).pptx
Autodesk Construction Cloud (Autodesk Build).pptx
 
History of Indian Railways - the story of Growth & Modernization
History of Indian Railways - the story of Growth & ModernizationHistory of Indian Railways - the story of Growth & Modernization
History of Indian Railways - the story of Growth & Modernization
 
electrical installation and maintenance.
electrical installation and maintenance.electrical installation and maintenance.
electrical installation and maintenance.
 
Augmented Reality (AR) with Augin Software.pptx
Augmented Reality (AR) with Augin Software.pptxAugmented Reality (AR) with Augin Software.pptx
Augmented Reality (AR) with Augin Software.pptx
 
CLOUD COMPUTING SERVICES - Cloud Reference Modal
CLOUD COMPUTING SERVICES - Cloud Reference ModalCLOUD COMPUTING SERVICES - Cloud Reference Modal
CLOUD COMPUTING SERVICES - Cloud Reference Modal
 
Final DBMS Manual (2).pdf final lab manual
Final DBMS Manual (2).pdf final lab manualFinal DBMS Manual (2).pdf final lab manual
Final DBMS Manual (2).pdf final lab manual
 

Not Enough Measurements, Too Many Measurements

  • 1. Not Enough Measurements, Too Many Measurements Michael McCann Department of Computational Mathematics, Science and Engineering Michigan State University UM CSP Seminar, Oct. 22, 2020
  • 2. 2 My collaborators ● Michael Unser and BIG (EPFL) – Kyong Jin – Laurène Donati – Harshit Gupta ● Sai Ravishankar and SLIM (MSU) – Avrajit Ghosh
  • 3. 3 Goal: reconstruct meaningful images from measurements what we have: measurements what we want: image The Physics
  • 7. 7 Not Enough Measurements ● measurements are expensive – time, radiation exposure, … ● want to take M << N measurements ● it’s hopeless!
  • 8. 8 Circa 2008-2016, the next slide would have been...
  • 10. 10 Too Many Measurements! ● e.g., The Cancer Imaging Archive – 2223 subjects x hundreds of images – includes Patient CT Projection Data Library (AKA Mayo Clinic Data)
  • 11. 11 Goal: reconstruct meaningful images from measurements PLUS a training set what we have: training what we have: measurements what we want: image The Physics
  • 13. 13 Supervised image reconstruction ● What do we pick for D? ● Where does the training data come from? ● How do we solve the fitting problem? ● What is the structure of F?
  • 15. 15 Themes in designing F ● augment a direct method – e.g., mix together different FBPs (Pelt et al. 2013) denoise an FBP (Jin et al. 2017)
  • 16. 16 Themes in designing F ● augment a direct method – e.g., mix together different FBPs (Pelt et al. 2013) denoise an FBP (Jin et al. 2017) ● take inspiration from variational methods – unrolling, plug-and-play, ... – e.g., learn the regularization (Aggarwal et al. 2019; Gupta et al. 2018) learn the gradient (Adler et al. 2017), learn filters and nonlinearities (Hammernik et al. 2017)
  • 17. 17 Themes in designing F ● augment a direct method – e.g., mix together different FBPs (Pelt et al. 2013) denoise an FBP (Jin et al. 2017) ● take inspiration from variational methods – unrolling, plug-and-play, ... – e.g., learn the regularization (Aggarwal et al. 2019; Gupta et al. 2018) learn the gradient (Adler et al. 2017), learn filters and nonlinearities (Hammernik et al. 2017) ● throw away variational methods – e.g., learn the entire measurement to image mapping (Zhu et al. 2018)
  • 18. 18 Themes in designing F ● augment a direct method – e.g., mix together different FBPs (Pelt et al. 2013) denoise an FBP (Jin et al. 2017) ● take inspiration from variational methods – unrolling, plug-and-play, ... – e.g., learn the regularization (Aggarwal et al. 2019; Gupta et al. 2018) learn the gradient (Adler et al. 2017), learn filters and nonlinearities (Hammernik et al. 2017) ● throw away variational methods – e.g., learn the entire measurement to image mapping (Zhu et al. 2018) ● work in the data domain – e.g., learn to inpaint missing measurements (Ghani et al. 2019)
  • 19. 19 Themes in designing F ● augment a direct method – e.g., mix together different FBPs (Pelt et al. 2013) denoise an FBP (Jin et al. 2017) ● take inspiration from variational methods – unrolling, plug-and-play, ... – e.g., learn the regularization (Aggarwal et al. 2019; Gupta et al. 2018) learn the gradient (Adler et al. 2017), learn filters and nonlinearities (Hammernik et al. 2017) ● throw away variational methods – e.g., learn the entire measurement to image mapping (Zhu et al. 2018) ● work in the data domain – e.g., learn to inpaint missing measurements (Ghani et al. 2019) ● many more...
  • 20. 20 Results: biomedical images (Jin, McCann, Froustey, Unser 2017)
  • 21. 21 Results: biomedical images (Jin, McCann, Froustey, Unser 2017)
  • 22. 22 Results: biomedical images (Jin, McCann, Froustey, Unser 2017)
  • 24. 24 Perspectives ● It’s not about H, it’s about 𝓗
  • 25. 25 Perspectives ● It’s not about H, it’s about 𝓗 ● If we care about MSE, let’s do challenges – fastMRI, Low Dose CT Grand Challenge, others?
  • 26. 26 Perspectives ● It’s not about H, it’s about 𝓗 ● If we care about MSE, let’s do challenges – fastMRI, Low Dose CT Grand Challenge, others? ● Let’s not forget data fidelity and robustness ground truth FBPConvNet
  • 27. 27 Summary so far X-ray CT low-dose X-ray CT (“impossible”)
  • 28. 28 Summary so far X-ray CT supervised low-dose X-ray CT
  • 32. 32 Single-particle cryo-EM ● ~10^6 projections of a single particle ● random orientations, (electron) optical effects, SNR ≈ 0 dB ● subnanometer resolution
  • 33. 33 Single-particle cryo-EM ● ~10^6 projections of a single particle ● random orientations, (electron) optical effects, SNR ≈ 0 dB ● subnanometer resolution
  • 34. 34 Single-particle cryo-EM ● ~10^6 projections of a single particle ● random orientations, (electron) optical effects, SNR ≈ 0 dB ● subnanometer resolution
  • 35. 35 Single-particle cryo-EM ● ~10^6 projections of a single particle ● random orientations, (electron) optical effects, SNR ≈ 0 dB ● subnanometer resolution right: Wu et al. 2020
  • 36. 36 Single-particle cryo-EM what we have: measurements what we want: image The Physics unknown nuisance
  • 37. 37 Cryo-EM reconstruction ● projection matching (Penczek et al. 1994) – # of unknowns grows with data :( ● marginalized maximum likelihood (Sigworth 1998) – marginalization is computationally heavy – what most software currently does ● method of moments (Kam 1980; Sharon et al. 2020) – avoids above problems – makes efficient use of data
  • 38. 38 Generative adversarial networks ● given samples of a distribution, generate new samples from the same distribution image: Goodfellow et al. 2014
  • 39. 39 Generative adversarial networks ● given samples of a distribution, generate new samples from the same distribution ● parameterization of the generator is flexible image: Goodfellow et al. 2014
  • 40. 40 Typical GAN (Gupta, McCann, Donati, Unser. 2020)
  • 41. 41 Typical GAN (Gupta, McCann, Donati, Unser. 2020)
  • 42. 42 Typical GAN (Gupta, McCann, Donati, Unser. 2020)
  • 48. 48 for more on CryoGAN, SPACE Webinar: https:// www.youtube.com/watch?v=J6UZBeU3Bm0 (Gupta et al. 2020)
  • 50. 50 Back to low-dose X-ray CT (Jin et al. 2017)
  • 51. 51 Back to low-dose X-ray CT (Jin et al. 2017)
  • 52. 52 Back to low-dose X-ray CT (Jin et al. 2017)
  • 53. 53 A “that’s funny…”moment figures and table: Adler et al. 2017
  • 54. 54 Can we learn a regularizer from data? ● recall supervised reconstruction ● our proposed “architecture”
  • 55. 55 Why learn a regularizer?
  • 56. 56 Why learn a regularizer? ● can’t be worse than sparsity-based reconstruction
  • 57. 57 Why learn a regularizer? ● can’t be worse than sparsity-based reconstruction ● results in a a convex problem – robust, includes data fidelity, decades of theory
  • 58. 58 Why learn a regularizer? ● can’t be worse than sparsity-based reconstruction ● results in a a convex problem – robust, includes data fidelity, decades of theory ● joins proven architecture with data adaptivity
  • 59. 59 Why learn a regularizer? ● can’t be worse than sparsity-based reconstruction ● results in a a convex problem – robust, includes data fidelity, decades of theory ● joins proven architecture with data adaptivity ● gives a hope of interpreting the learned part
  • 60. 60 Related work ● Main approach: relax the l1 term – Peyré et al. 2011; Mairal et al. 2012; Sprechmann et al. 2013; Chen et al. 2014
  • 61. 61 Related work ● Main approach: relax the l1 term – Peyré et al. 2011; Mairal et al. 2012; Sprechmann et al. 2013; Chen et al. 2014
  • 62. 62 Related work ● Main approach: relax the l1 term – Peyré et al. 2011; Mairal et al. 2012; Sprechmann et al. 2013; Chen et al. 2014
  • 63. 63 Related work ● Main approach: relax the l1 term – Peyré et al. 2011; Mairal et al. 2012; Sprechmann et al. 2013; Chen et al. 2014
  • 65. 65 Our approach (outline) 1. solve the lower problem at the current W
  • 66. 66 Our approach (outline) 1. solve the lower problem at the current W 2. find a (local) closed-form solution, x*(W)
  • 67. 67 Our approach (outline) 1. solve the lower problem at the current W 2. find a (local) closed-form solution, x*(W) 3. substitute x*(W) into the upper level problem, compute a gradient w.r.t. W, and descend
  • 68. 68 1. Solve the lower level problem ● standard, convex problem – ADMM, Chambolle-Pock ● need fast, accurate solutions – hyperparameter selection is tough
  • 69. 69 2. Find a closed-form solution
  • 70. 70 2. Find a closed-form solution ● uniqueness? no, but
  • 71. 71 2. Find a closed-form solution ● uniqueness? no, but ● intuition: in a region (of W-space) where the sign pattern of Wx*(W) does not change, ||Wx||1 is linear
  • 72. 72 2. Find a closed-form solution ● uniqueness? no, but ● intuition: in a region (of W-space) where the sign pattern of Wx*(W) does not change, ||Wx||1 is linear – see McCann and Ravishankar 2020 for A = I
  • 73. 73 2. Find a closed-form solution ● from McCann and Ravishankar 2020
  • 74. 74 2. Find a closed-form solution ● from McCann and Ravishankar 2020 ● even better, Ali and Tibshirani 2019
  • 75. 75 2. Find a closed-form solution ● from McCann and Ravishankar 2020 ● even better, Ali and Tibshirani 2019 – unique minimum norm solution when b=0
  • 76. 76 3. Find the gradient
  • 77. 77 3. Find the gradient ● with 1d signals, pytorch can autograd these expressions – Agrawal et al. 2019 makes it even easier
  • 78. 78 3. Find the gradient ● with 1d signals, pytorch can autograd these expressions – Agrawal et al. 2019 makes it even easier ● with images, things get tough – W is potentially million X million – no more SVDs or explicit inverses – our solution: by hand + pytorch
  • 79. 79 Early experiments ● image denoising ● W is a set of 8, 3x3 convolutions ● compare to – BM3D, TV, DCT – unsupervised learned regularizer ● training – SGD with increasing batch size – takes ~hours
  • 83. 83 Absolute summed filter responses and filters ● Learned filters – Are not orthonormal (neither ortho- nor -normal) – Penalize edges less than DCT
  • 85. 85 Taking a step back ● CNN performance on image reconstruction is due to – the CNN (architecture) – training
  • 86. 86 Taking a step back ● CNN performance on image reconstruction is due to – the CNN (architecture) – training ● remove the training: deep image prior – Reinhard Heckel 1W-MINDS seminar: https://www.youtube.com/watch?v=AvJgmbeupGY
  • 87. 87 Taking a step back ● CNN performance on image reconstruction is due to – the CNN (architecture) – training ● remove the training: deep image prior – Reinhard Heckel 1W-MINDS seminar: https://www.youtube.com/watch?v=AvJgmbeupGY ● remove the CNN: learned regularizer
  • 88. 88 Thanks for your attention! michael.thompson.mccann@gmail.com slides:
  • 89. 89 References ● M. Wu, G. C. Lander, and M. A. Herzik, “Sub-2 Angstrom resolution structure determination using single- particle cryo-EM at 200 keV,” Journal of Structural Biology: X, vol. 4, p. 100020, 2020, doi: 10.1016/j.yjsbx.2020.100020. ● Pawel A. Penczek, Robert A. Grassucci, and Joachim Frank. “The ribosome at improved resolution: New techniques for merging and orientation refinement in 3D cryo-electron microscopy of biological particles”. In:Ultramicroscopy 53.3 (1994), pp. 251–270. ● Fred J Sigworth. “A maximum-likelihood approach to single-particle image refinement”. In:Journal of structural biology 122.3 (1998), pp. 328–339. ● Z. Kam, “The reconstruction of structure from electron micrographs of randomly oriented particles,” Journal of Theoretical Biology, vol. 82, no. 1, Art. no. 1, Jan. 1980, doi: 10.1016/0022-5193(80)90088-0. ● N. Sharon, J. Kileel, Y. Khoo, B. Landa, and A. Singer, “Method of moments for 3D single particle ab initio modeling with non-uniform distribution of viewing angles,” Inverse Problems, vol. 36, no. 4, Art. no. 4, Feb. 2020, doi: 10.1088/1361-6420/ab6139. ● I. Goodfellow et al., “Generative Adversarial Nets,” in Advances in Neural Information Processing Systems 27, Ed.Z. Ghahramani, M. Welling, C. Cortes, N. D. Lawrence, and K. Q. Weinberger Curran Associates, Inc., 2014, pp. 2672–2680. ● J. Adler and O. Öktem, “Solving ill-posed inverse problems using iterative deep neural networks,” Inverse Problems, vol. 33, no. 12, p. 124007, Nov. 2017.
  • 90. 90 References ● H. Gupta, M. T. McCann, L. Donati, and M. Unser, “CryoGAN: A New Reconstruction Paradigm for Single-Particle Cryo-EM via Deep Adversarial Learning,” bioRxiv, Mar. 2020, doi: 10.1101/2020.03.20.001016. ● G. Peyré and J. M. Fadili, “Learning analysis sparsity priors,” in Sampling Theory and Applications, Singapore, Singapore, May 2011, p. 4. ● J. Mairal, F. Bach, and J. Ponce, “Task-driven dictionary learning,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 34, no. 4, pp. 791–804, Apr. 2012. ● P. Sprechmann, R. Litman, T. Ben Yakar, A. M. Bronstein, and G. Sapiro, “Supervised sparse analysis and synthesis operators,” in Advances in Neural Information Processing Systems 26, 2013, pp. 908– 916. ● M. T. McCann and S. Ravishankar, “Supervised Learning of Sparsity-Promoting Regularizers for Denoising,” arXiv:2006.05521 [eess.IV], 2020-06-09. ● A. Ali and R. J. Tibshirani, “The Generalized Lasso Problem and Uniqueness,” Electronic Journal of Statistics, vol. 13, no. 2, Art. no. 2, 2019, doi: 10.1214/19-ejs1569. ● A. Agrawal, B. Amos, S. Barratt, S. Boyd, S. Diamond, and J. Z. Kolter, “Differentiable Convex Optimization Layers,” in Advances in Neural Information Processing Systems 32, Ed.H. Wallach, H. Larochelle, A. Beygelzimer, F. dAlché-Buc, E. Fox, and R. Garnett Curran Associates, Inc., 2019, pp. 9562–9574.