Scalable Convolutional Sparse Coding

PhD Thesis Proposal Defence


Title: "Scalable Convolutional Sparse Coding"

by

Miss Yaqing WANG


Abstract:

Convolutional sparse coding (CSC) improves sparse coding by learning a 
shift-invariant dictionary from the data. It has been successfully extracting 
local patterns from various data types, such as trajectories, images, audios, 
videos, multi-spectral and light field images and biomedical data. However, 
most existing CSC algorithms operate in the batch mode and are computationally 
expensive. This lack of scalability restricts the use of CSC on large-scale 
data.

First, we propose a scalable online CSC algorithm for data sets of large 
quantity. The algorithm, which will be called Online Convolutional Sparse 
Coding (OCSC). The key is a reformulation of the CSC objective so that 
convolution can be handled easily in the frequency domain, and much smaller 
space is needed. To solve the resultant optimization problem, we use the 
alternating direction method of multipliers (ADMM), and its subproblems have 
efficient closed-form solutions. Theoretical analysis shows that the learned 
dictionary converges to a stationary point of the optimization problem. Results 
on large image data sets such as ImageNet, Flower and CIFAR-10 show that the 
proposed algorithm outperforms state-of-the-art batch and online CSC methods. 
It is more scalable, has faster convergence and better reconstruction 
performance.

Then, we enable CSC to learn with a large set of filters. Instead of convolving 
with a dictionary shared by all samples, we propose the use of a 
sample-dependent dictionary in which each filter is a linear combination of a 
small set of base filters learned from data. This added flexibility allows a 
large number of sample-dependent patterns to be captured, which is especially 
useful in the handling of large or high-dimensional data sets. Computationally, 
the resultant model can be efficiently learned by online learning. 
Specifically, the base filter can still be updated by OCSC, while the codes and 
combination weights can be learned by accelerated nonconvex proximal 
algorithms. Extensive experimental results on multiple kinds of data sets such 
as images, multispectral images, videos, and light field images show that the 
proposed methods outperform existing CSC algorithms with significantly reduced 
time and space complexities.

Finally, we propose some further improvement. For example, extending CSC to 
deal with unknown noise, designing multiscale filters, introducing stochastic 
version, and extending the proposed sample-dependent filters to convolutional 
neural networks and graph convolutional networks.


Date:			Monday, 12 November 2018

Time:                  	4:30pm - 6:30pm

Venue:                  Room 5501
                         (lifts 25/26)

Committee Members:	Prof. Lionel Ni (Supervisor)
 			Prof. James Kwok (Supervisor)
 			Dr. Qiong Luo (Chairperson)
 			Prof. Nevin Zhang


**** ALL are Welcome ****