Convolutional Analysis Operator Learning: Dependence on Training Data (2019-02-21T00:00:00.000000Z)

TL;DR

A series of results that lend insight into the impact of dataset size on the filter update in CAOL are presented, including a general deterministic bound on errors in the estimated filters, and a bound on the expected errors as the number of training samples increases.

Abstract

Convolutional analysis operator learning (CAOL) enables the unsupervised training of (hierarchical) convolutional sparsifying operators or autoencoders from large datasets. One can use many training images for CAOL, but a precise understanding of the impact of doing so has remained an open question. This letter presents a series of results that lend insight into the impact of dataset size on the filter update in CAOL. The first result is a general deterministic bound on errors in the estimated filters, and is followed by a bound on the expected errors as the number of training samples increases. The second result provides a high probability analogue. The bounds depend on properties of the training data, and we investigate their empirical values with real data. Taken together, these results provide evidence for the potential benefit of using more training data in CAOL.

Authors

B. Adcock

2 papers

Il Yong Chun

2 papers

J. Fessler

2 papers

Convolutional Analysis Operator Learning: Dependence on Training Data

TL;DR

Abstract

Authors

References23 items

Foundations of Machine Learning

Sample Complexity Bounds for Dictionary Learning from Vector- and Tensor-Valued Data

Convolutional analysis operator learning: Application to sparse-view CT : (Invited Paper)

Fast and convergent iterative image recovery using trained convolutional neural networks

Minimax Reconstruction Risk of Convolutional Sparse Dictionary Learning

Deep BCD-Net Using Identical Encoding-Decoding CNN Structures for Iterative Image Recovery

Convolutional Analysis Operator Learning: Acceleration and Convergence

Convolutional Dictionary Learning: Acceleration and Convergence

Convergent convolutional dictionary learning using Adaptive Contrast Enhancement (CDL-ACE): Application of CDL to image denoising

Learning Co-Sparse Analysis Operators With Separable Structures

$\ell_{0}$ Sparsifying Transform Learning With Efficient Optimal Updates and Convergence Guarantees

Data-driven tight frame construction and image denoising

A Mathematical Introduction to Compressive Sensing

Constrained Overcomplete Analysis Operator Learning for Cosparse Signal Modelling

Analysis Operator Learning and its Application to Image Reconstruction

The Elements of Statistical Learning: Data Mining, Inference, and Prediction, Second Edition by Trevor Hastie, Robert Tibshirani, Jerome Friedman

$rm K$-SVD: An Algorithm for Designing Overcomplete Dictionaries for Sparse Representation

The elements of statistical learning: data mining, inference and prediction

Das asymptotische Verteilungsgesetz der Eigenwerte linearer partieller Differentialgleichungen (mit einer Anwendung auf die Theorie der Hohlraumstrahlung)

Cambridge, UK

and W

K-SVD : An Algorithm for Designing of Overcomplete Dictionaries for Sparse Representation

New Perturbation Bounds for the Unitary Polar Factor

Field of Study

Journal Information

Name

Page

Volume

Venue Information

Name

Type

URL

Alternate Names