nikcleju@0: # -*- coding: utf-8 -*-
nikcleju@0: """
nikcleju@17: Defines standard parameters for approximate reconstruction simulation
nikcleju@0: 
nikcleju@17: Author: Nicolae Cleju
nikcleju@0: """
nikcleju@17: __author__ = "Nicolae Cleju"
nikcleju@17: __license__ = "GPL"
nikcleju@17: __email__ = "nikcleju@gmail.com"
nikcleju@0: 
nikcleju@0: import numpy
nikcleju@17: 
nikcleju@17: # Solver algorithms to run
nikcleju@0: from algos import *
nikcleju@0: 
nikcleju@17: # Test parameters
nikcleju@15: paramstest = dict()
nikcleju@21: #paramstest['algosN'] = nesta,      # tuple of algorithms not depending on lambda
nikcleju@21: #paramstest['algosL'] = lambda_sl0,  # tuple of algorithms depending on lambda (ABS-lambda)
nikcleju@21: paramstest['algosN'] = gap,mixed_sl0,mixed_bp,nesta      # tuple of algorithms not depending on lambda
nikcleju@21: paramstest['algosL'] = lambda_sl0,lambda_bp,lambda_ompeps,lambda_tst 
nikcleju@15: paramstest['d'] = 50.0
nikcleju@15: paramstest['sigma'] = 2.0
nikcleju@15: paramstest['deltas'] = numpy.array([0.05, 0.45, 0.95])
nikcleju@15: paramstest['rhos'] = numpy.array([0.05, 0.45, 0.95])
nikcleju@15: #deltas = numpy.array([0.95])
nikcleju@15: #deltas = numpy.arange(0.05,1.,0.05)
nikcleju@15: #rhos = numpy.array([0.05])
nikcleju@17: paramstest['numvects'] = 10;  # Number of vectors to generate
nikcleju@21: paramstest['SNRdb'] = 40.;    # This is norm(signal)/norm(noise), so power, not energy
nikcleju@15: # Values for lambda
nikcleju@15: #lambdas = [0 10.^linspace(-5, 4, 10)];
nikcleju@15: paramstest['lambdas'] = numpy.array([0., 0.0001, 0.01, 1, 100, 10000])
nikcleju@15: paramstest['savedataname'] = 'approx_pt_stdtest.mat'
nikcleju@15: paramstest['saveplotbase'] = 'approx_pt_stdtest_'
nikcleju@15: paramstest['saveplotexts'] = ('png','pdf','eps')
nikcleju@0: 
nikcleju@0: 
nikcleju@21: # Prove 11 convergence
nikcleju@19: paramsl1prove = dict()
nikcleju@19: paramsl1prove['algosN'] = nesta,      # tuple of algorithms not depending on lambda
nikcleju@19: paramsl1prove['algosL'] = lambda_bp,  # tuple of algorithms depending on lambda (ABS-lambda)
nikcleju@19: paramsl1prove['d'] = 50.0
nikcleju@19: paramsl1prove['sigma'] = 2.0
nikcleju@19: paramsl1prove['deltas'] = numpy.array([0.8]) # m = 0.8*50 = 40
nikcleju@19: paramsl1prove['rhos'] = numpy.array([0.15]) # l = d - rho*m = 50 - 0.15*40 = 44
nikcleju@19: paramsl1prove['numvects'] = 100;  # Number of vectors to generate
nikcleju@19: paramsl1prove['SNRdb'] = 40.;    # This is norm(signal)/norm(noise), so power, not energy
nikcleju@19: paramsl1prove['lambdas'] = numpy.array([0.00001, 0.0001, 0.001, 0.01, 0.1, 1, 10, 100, 1000, 10000])
nikcleju@19: paramsl1prove['savedataname'] = 'prove_approx_ell1.mat'
nikcleju@19: paramsl1prove['saveplotbase'] = 'prove_approx_ell1'
nikcleju@19: paramsl1prove['saveplotexts'] = ('png','pdf','eps')
nikcleju@19: 
nikcleju@19: 
nikcleju@0: # Standard parameters 1
nikcleju@0: # All algorithms, 100 vectors
nikcleju@17: # d = 50, sigma = 2, delta and rho full resolution (0.05 step), lambdas = 0, 1e-4, 1e-2, 1, 100, 10000
nikcleju@17: # Noise 40db
nikcleju@17: params1 = dict()
nikcleju@17: params1['algosN'] = gap,mixed_sl0,mixed_bp,nesta      # tuple of algorithms not depending on lambda
nikcleju@17: params1['algosL'] = lambda_sl0,lambda_bp,lambda_ompeps,lambda_tst  # tuple of algorithms depending on lambda (ABS-lambda)
nikcleju@17: params1['d'] = 50.0
nikcleju@21: params1['sigma'] = 1.2
nikcleju@17: params1['deltas'] = numpy.arange(0.05,1.,0.05)
nikcleju@17: params1['rhos'] = numpy.arange(0.05,1.,0.05)
nikcleju@21: params1['numvects'] = 100;  # Number of vectors to generate
nikcleju@17: params1['SNRdb'] = 40.;    # This is norm(signal)/norm(noise), so power, not energy
nikcleju@17: params1['lambdas'] = numpy.array([0., 0.0001, 0.01, 1, 100, 10000])
nikcleju@21: params1['savedataname'] = 'approx_pt_params1.mat'
nikcleju@21: params1['saveplotbase'] = 'approx_pt_params1_'
nikcleju@17: params1['saveplotexts'] = ('png','pdf','eps')
nikcleju@0: 
nikcleju@0: # Standard parameters 2
nikcleju@0: # All algorithms, 100 vectors
nikcleju@17: # d = 20, sigma = 10, delta and rho full resolution (0.05 step), lambdas = 0, 1e-4, 1e-2, 1, 100, 10000
nikcleju@17: # Noise 40db
nikcleju@17: params2 = dict()
nikcleju@17: params2['algosN'] = gap,mixed_sl0,mixed_bp,nesta      # tuple of algorithms not depending on lambda
nikcleju@17: params2['algosL'] = lambda_sl0,lambda_bp,lambda_ompeps,lambda_tst  # tuple of algorithms depending on lambda (ABS-lambda)
nikcleju@21: params2['d'] = 50.0
nikcleju@21: params2['sigma'] = 2
nikcleju@17: params2['deltas'] = numpy.arange(0.05,1.,0.05)
nikcleju@17: params2['rhos'] = numpy.arange(0.05,1.,0.05)
nikcleju@21: params2['numvects'] = 100;  # Number of vectors to generate
nikcleju@17: params2['SNRdb'] = 40.;    # This is norm(signal)/norm(noise), so power, not energy
nikcleju@17: params2['lambdas'] = numpy.array([0., 0.0001, 0.01, 1, 100, 10000])
nikcleju@21: params2['savedataname'] = 'approx_pt_params2.mat'
nikcleju@21: params2['saveplotbase'] = 'approx_pt_params2_'
nikcleju@17: params2['saveplotexts'] = ('png','pdf','eps')
nikcleju@0:   
nikcleju@0:   
nikcleju@17: # Standard parameters 3
nikcleju@0: # All algorithms, 100 vectors
nikcleju@17: # d = 50, sigma = 2, delta and rho full resolution (0.05 step), lambdas = 0, 1e-4, 1e-2, 1, 100, 10000
nikcleju@17: # Identical with 1 but with 20dB SNR noise
nikcleju@17: params3 = dict()
nikcleju@17: params3['algosN'] = gap,mixed_sl0,mixed_bp,nesta      # tuple of algorithms not depending on lambda
nikcleju@17: params3['algosL'] = lambda_sl0,lambda_bp,lambda_ompeps,lambda_tst  # tuple of algorithms depending on lambda (ABS-lambda)
nikcleju@17: params3['d'] = 50.0
nikcleju@21: params3['sigma'] = 1.2
nikcleju@17: params3['deltas'] = numpy.arange(0.05,1.,0.05)
nikcleju@17: params3['rhos'] = numpy.arange(0.05,1.,0.05)
nikcleju@21: params3['numvects'] = 100;  # Number of vectors to generate
nikcleju@17: params3['SNRdb'] = 20.;    # This is norm(signal)/norm(noise), so power, not energy
nikcleju@17: params3['lambdas'] = numpy.array([0., 0.0001, 0.01, 1, 100, 10000])
nikcleju@21: #params3['lambdas'] = numpy.array([0., 0.01, 0.1, 1, 10, 100])
nikcleju@21: params3['savedataname'] = 'approx_pt_params3.mat'
nikcleju@21: params3['saveplotbase'] = 'approx_pt_params3_'
nikcleju@17: params3['saveplotexts'] = ('png','pdf','eps')
nikcleju@0:           
nikcleju@0: # Standard parameters 4
nikcleju@0: # All algorithms, 100 vectors
nikcleju@17: # d = 20, sigma = 10, delta and rho full resolution (0.05 step), lambdas = 0, 1e-4, 1e-2, 1, 100, 10000
nikcleju@17: # Identical to 2 but with 20dB SNR noise
nikcleju@17: params4 = dict()
nikcleju@17: params4['algosN'] = gap,mixed_sl0,mixed_bp,nesta      # tuple of algorithms not depending on lambda
nikcleju@17: params4['algosL'] = lambda_sl0,lambda_bp,lambda_ompeps,lambda_tst  # tuple of algorithms depending on lambda (ABS-lambda)
nikcleju@21: params4['d'] = 50.0
nikcleju@21: params4['sigma'] = 2.0
nikcleju@17: params4['deltas'] = numpy.arange(0.05,1.,0.05)
nikcleju@17: params4['rhos'] = numpy.arange(0.05,1.,0.05)
nikcleju@21: params4['numvects'] = 100;  # Number of vectors to generate
nikcleju@17: params4['SNRdb'] = 20.;    # This is norm(signal)/norm(noise), so power, not energy
nikcleju@17: params4['lambdas'] = numpy.array([0., 0.0001, 0.01, 1, 100, 10000])
nikcleju@21: #params4['lambdas'] = numpy.array([0., 0.01, 0.1, 1, 10, 100])
nikcleju@21: params4['savedataname'] = 'approx_pt_params4.mat'
nikcleju@21: params4['saveplotbase'] = 'approx_pt_params4_'
nikcleju@17: params4['saveplotexts'] = ('png','pdf','eps')
nikcleju@21: 
nikcleju@21: 
nikcleju@21: # Standard parameters 5
nikcleju@21: # All algorithms, 100 vectors
nikcleju@21: # d = 50, sigma = 2, delta and rho full resolution (0.05 step), lambdas = 0, 1e-4, 1e-2, 1, 100, 10000
nikcleju@21: # VIRTUALLY NO NOISE, Noise 200db
nikcleju@21: params5 = dict()
nikcleju@21: params5['algosN'] = gap,mixed_sl0,mixed_bp,nesta      # tuple of algorithms not depending on lambda
nikcleju@21: params5['algosL'] = lambda_sl0,lambda_bp,lambda_ompeps,lambda_tst  # tuple of algorithms depending on lambda (ABS-lambda)
nikcleju@21: params5['d'] = 50.0
nikcleju@21: params5['sigma'] = 1.2
nikcleju@21: params5['deltas'] = numpy.arange(0.05,1.,0.05)
nikcleju@21: params5['rhos'] = numpy.arange(0.05,1.,0.05)
nikcleju@21: params5['numvects'] = 10;  # Number of vectors to generate
nikcleju@21: params5['SNRdb'] = 200.;    # This is norm(signal)/norm(noise), so power, not energy
nikcleju@21: params5['lambdas'] = numpy.array([0., 0.0001, 0.01, 1, 100, 10000])
nikcleju@21: params5['savedataname'] = 'approx_pt_params5.mat'
nikcleju@21: params5['saveplotbase'] = 'approx_pt_params5_'
nikcleju@21: params5['saveplotexts'] = ('png','pdf','eps')
nikcleju@21: 
nikcleju@21: params3sl0 = dict()
nikcleju@21: params3sl0['algosN'] = mixed_sl0,      # tuple of algorithms not depending on lambda
nikcleju@21: params3sl0['algosL'] = lambda_sl0,  # tuple of algorithms depending on lambda (ABS-lambda)
nikcleju@21: params3sl0['d'] = 50.0
nikcleju@21: params3sl0['sigma'] = 1.2
nikcleju@21: params3sl0['deltas'] = numpy.arange(0.05,1.,0.25)
nikcleju@21: params3sl0['rhos'] = numpy.arange(0.05,1.,0.25)
nikcleju@21: #params3sl0['deltas'] = numpy.array([0.85])
nikcleju@21: #params3sl0['rhos'] = numpy.array([0.15])
nikcleju@21: params3sl0['numvects'] = 10;  # Number of vectors to generate
nikcleju@21: params3sl0['SNRdb'] = 20.;    # This is norm(signal)/norm(noise), so power, not energy
nikcleju@21: #params3sl0['lambdas'] = numpy.array([0., 0.01, 0.1, 1, 10, 100])
nikcleju@21: params3sl0['lambdas'] = numpy.array([0.0001, 10000])
nikcleju@21: params3sl0['savedataname'] = 'approx_pt_params3sl0.mat'
nikcleju@21: params3sl0['saveplotbase'] = 'approx_pt_params3sl0_'
nikcleju@21: params3sl0['saveplotexts'] = ('png','pdf','eps')