Dictionary Identification - Sparse Matrix-Factorisation via $\ell_1$-Minimisation

This article treats the problem of learning a dictionary providing sparse representations for a given signal class, via $\ell_1$-minimisation. The problem can also be seen as factorising a $\ddim \times \nsig$ matrix $Y=(y_1 \ldots y_\nsig), \, y_n\in \R^\ddim$ of training signals into a $\ddim \times \natoms$ dictionary matrix $\dico$ and a $\natoms \times \nsig$ coefficient matrix $\X=(x_1\ldots x_\nsig),\, x_n \in \R^\natoms$, which is sparse. The exact question studied here is when a dictionary coefficient pair $(\dico,\X)$ can be recovered as local minimum of a (nonconvex) $\ell_1$-criterion with input $Y=\dico \X$. First, for general dictionaries and coefficient matrices, algebraic conditions ensuring local identifiability are derived, which are then specialised to the case when the dictionary is a basis. Finally, assuming a random Bernoulli-Gaussian sparse model on the coefficient matrix, it is shown that sufficiently incoherent bases are locally identifiable with high probability. The perhaps surprising result is that the typically sufficient number of training samples $\nsig$ grows up to a logarithmic factor only linearly with the signal dimension, i.e. $\nsig \approx C \natoms \log \natoms$, in contrast to previous approaches requiring combinatorially many samples.


Published in:
IEEE Transactions on Information Theory, 56, 7, 3523 - 3539
Year:
2010
Publisher:
Institute of Electrical and Electronics Engineers
ISSN:
0018-9448
Keywords:
Laboratories:




 Record created 2009-05-01, last modified 2018-03-17

External link:
Download fulltext
URL
Rate this document:

Rate this document:
1
2
3
 
(Not yet reviewed)