cs1 0.1.1

Compressed Sensing library for 1D Spectroscopic Profiling Data

cs1

Compressed Sensing library for 1D Spectroscopic Profiling Data

module	sub-module	description
cs1.basis	cs1.basis.common	commonly used non-adaptive CS transform bases
	cs1.basis.adaptive	adaptive CS transform bases, e.g., LDA (linear discriminant analysis)-based, EBP (eigenvector-based projection)
cs1.cs		basic functions for CS sensing, recovery, hyper-parameter grid-search, etc.
cs1.metrics		CS-related metrics, e.g., mutual coherence, sparsity, MSE, KLD
cs1.security	cs1.security.tvsm	time-variant sensing matrix mechanism
cs1.domain	cs1.domain.audio	contains functions for audio and other one-dimensional signal processing. e.g., wave file I/O, lossy compression, ECG simulator
	cs1.domain.image	contains functions for image processing. e.g., image CS, lossy compression

Installation

pip install cs1

A simple startup

import cs1

# Generate common non-adaptive bases and save to a local pickle file.
# The generation process can be very slow, so save it for future use.
cs1.basis.common.Generate_PSIs(n, savepath = 'PSIs_' + str(n) + '.pkl') # n is the data/signal dimensionality

# load back bases
file = open('PSIs_' + str(n) + '.pkl','rb')
PSIs = pickle.load(file)
file.close()

# sparsity analysis
cs1.metrics.analyze_sparsity(x, PSIs)

# compare different bases and sampling ratio on a single sample
rmses = cs1.cs.GridSearch_Sensing_n_Recovery(x, PSIs, solver = 'LASSO') # returns relative MSEs

low-level cs functions

from cs1.basis.common import *

dftmtx()
dctmtx()
hwtmtx()

from cs1.cs import *

sensing()
recovery()

from cs1.metrics import *

mutual_coherence()
...

singal processing functions for audio / image domains

from cs1.domain.audio import *

simulate_ECG()
dct_lossy_signal_compression()
dft_lossy_signal_compression()

from cs1.domain.image import *

img_dct()
img_dft()
dct_lossy_image_compression()
dft_lossy_image_compression()

adaptive cs bases

from cs1.basis.adaptive import *

PSI, _ = EBP(X) # X is a m-by-n training dataset. PSI is the EBP basis
PSI, _, _ = LDA(X, y, display = True) # X and y are training dataset. PSI is the LDA basis.