Sparsifying Dictionaries and Sensing Matrices¶
Functions for constructing sparsying dictionaries and sensing matrices¶
|
A dictionary/sensing matrix where entries are drawn independently from normal distribution. |
|
A dictionary/sensing matrix where entries are drawn independently from Rademacher distribution. |
|
A sensing matrix where exactly d entries are 1 in each column |
|
Generates a random orthonormal basis for \(\mathbb{R}^N\) |
|
Generates a random sensing matrix with orthonormal rows |
|
Hadamard matrices of size \(n imes n\) |
A Hadamard basis |
|
A dictionary consisting of identity basis and hadamard bases |
|
|
DCT Basis |
A dictionary consisting of identity and DCT bases |
|
A dictionary consisting of identity, Hadamard and DCT bases |
|
Fourier basis |
|
|
Builds a wavelet basis for a given decomposition level |
Dictionary properties¶
|
Computes the Gram matrix \(G = A^T A\) |
|
Computes the frame matrix \(G = A A^T\) |
Returns the coherence of a dictionary A along with indices of most correlated atoms |
|
|
Computes the coherence of a dictionary |
|
Computes the frame bounds (largest and smallest singular valuee) |
Computes the upper frame bound for a dictionary |
|
Computes the lower frame bound for a dictionary |
|
|
Computes the babel function for a dictionary (generalized coherence) |
Dictionary comparison¶
These functions are useful for comparing dictionaries during the dictionary learning process.
Mutual coherence between two dictionaries A and B along with indices of most correlated atoms |
|
|
“Mutual coherence between two dictionaries A and B |
|
Identifies how many atoms are very close between dictionaries A and B |
Grassmannian frames¶
|
Minimum achievable coherence for a Grassmannian frame |
|
Builds a Grassmannian frame starting from a random matrix |