def plot_pr_curve(data,startPoints,filename,legend_string):
mlab.addpath('/home/cchansen/snp_analysis/common')
xlabel = 'Recall'
ylabel = 'Precision'
xlim = n.array([0,1])
ylim = n.array([0,1])
mlab.create_pdf_plot(data,startPoints,'',filename,xlabel,14,ylabel,14,xlim,ylim,legend_string)
def __init__(self, sigma = 0.2, _lambda = np.exp(-15), Nadd = 150, output = 0, maxIter = np.inf,
epsilon = 0.001, delta_k = 1, tempInt = 0.95, epsilon_back = 0.001, flyComputeK = 0,
deselect = 0, CV = None, lambdas = np.r_[[np.exp(-12)], np.exp(np.arange(-10,-3))],
sigmas = np.sqrt(1. / (2. * 2.**np.arange(-5, 3)))):
self.sigma = sigma
self._lambda = _lambda
self.Nadd = Nadd
self.output = output
self.maxIter = maxIter
self.epsilon = epsilon
self.delta_k = delta_k
self.tempInt = tempInt
self.epsilon_back = epsilon_back
self.flyComputeK = flyComputeK
self.deselect = deselect
self.CV = CV
self.lambdas = lambdas
self.sigmas = sigmas
this_dir, this_filename = os.path.split(__file__)
src_dir = os.path.join(this_dir, "ivmSoftware4.3/src")
mlab.addpath(src_dir)
def load_matlab():
"""
Imports and starts matlab bridge if not started.
"""
global mlab
global MatlabError
from mlabwrap import mlab
from mlabraw import error as MatlabError
abspath = os.path.abspath(__file__)
features = os.path.join(os.path.dirname(os.path.dirname(abspath)),
"features")
toolbox = os.path.join(features, "matlab-chroma-toolbox")
mlab.addpath(toolbox)
def generateBSAinput(scale=10):
from mlabwrap import mlab
#mlab.addpath('../')
mlab.addpath('/home/mammoth/dejan/simtools/AuditoryToolbox')
mlab.addpath('/home/mammoth/dejan/simtools/RCToolbox')
mlab.addpath('/home/mammoth/dejan/simtools/RCToolbox/spike_coding')
mlab.addpath('/home/mammoth/dejan/simtools/RCToolbox/utility')
mlab.addpath('/home/mammoth/dejan/simtools/speech')
currDir = os.getcwd()
mlab.cd('/home/mammoth/dejan/simtools/speech')
mlab.startup()
#mlab.cd(currDir)
mlab.cd('/home/mammoth/dejan/simtools/RCToolbox')
InputDist = mlab.preprocessed_speech('scale', scale)
h5filename = 'spkdata_%d.h5' % scale
N = mlab.get(InputDist, 'size').flatten()
N_rev = mlab.get(InputDist, 'rev_size').flatten()
NN = N + N_rev
print "Saving inputs..."
for i in range(NN):
stimulus = PreprocessedSpeechStimulus()
S = mlab.generate_input(InputDist, i + 1)
nc = mlab.length(S.channel).flatten()
for c in range(nc):
channel = Channel(S.channel[c].data.flatten())
stimulus.channel.append(channel)
stimulus.Tsim = S.info.Tstim.flatten()[0]
stimulus.file = S.info.file
stimulus.speaker = S.info.speaker.flatten()[0]
stimulus.utterance = S.info.utterance.flatten()[0]
stimulus.digit = S.info.digit.flatten()[0]
stimulus.reversed = S.info.reversed.flatten()[0] == 1
if stimulus.reversed:
grpname = stimulus.file + "_rev"
else:
grpname = stimulus.file
print "%d: %s" % (i, grpname)
stimulus.save(filename=h5filename, grpname=grpname)
import scipy as sp
import scipy.io
import plca
from string import lower
import csv
logging.basicConfig(level=logging.INFO,
format='%(levelname)s %(name)s %(asctime)s '
'%(filename)s:%(lineno)d %(message)s')
logger = logging.getLogger('segmenter')
try:
from mlabwrap import mlab
mlab.addpath('coversongs')
except:
logger.warning('Unable to import mlab module. Feature extraction '
'and evaluation will not work.')
def extract_features(wavfilename, fctr=400, fsd=1.0, type=1):
"""Computes beat-synchronous chroma features from the given wave file
Calls Dan Ellis' chrombeatftrs Matlab function.
"""
if lower(wavfilename[-4:]) == '.csv':
logger.info('CSV filename reading preprocessed features from %s',
wavfilename)
csvr = csv.reader(open(wavfilename, 'rb'), delimiter=',')
feats = np.array([[float(x) for x in row] for row in csvr])
"""
import os
import sys
import tempfile
import numpy as np
# MATLAB wrapper, we remove the dperecation warnings
import warnings
warnings.filterwarnings('ignore',category=DeprecationWarning)
from mlabwrap import mlab
warnings.filterwarnings('default',category=DeprecationWarning)
# kalman toolbox path
kal_toolbox_path = '/home/thierry/Columbia/Imputation/PythonSrc/KalmanAll'
mlab.warning('off','all') # hack, remove warnings
# some functions are redefined but who cares!
mlab.addpath(kal_toolbox_path)
mlab.addpath(os.path.join(kal_toolbox_path,'Kalman'))
mlab.addpath(os.path.join(kal_toolbox_path,'KPMstats'))
mlab.addpath(os.path.join(kal_toolbox_path,'KPMtools'))
mlab.warning('on','all')
def learn_kalman(data,A,C,Q,R,initx,initV,niter,diagQ=1,diagR=1):
"""
Main function, take initial parameters and train a Kalman filter.
We assume a model:
x(t+1) = A*x(t) + w(t), w ~ N(0, Q), x(0) ~ N(init_x, init_V)
y(t) = C*x(t) + v(t), v ~ N(0, R)
INPUT
data - one sequence, one observation per col
A - DxD matrix, D dimension of hidden state
from mlabwrap import mlab
import pycuda.autoinit
import pycuda.driver as drv
import numpy
from pycuda.compiler import SourceModule
mlab.addpath('../FasihSarStuff/', nout=0)
# Params
clight = 299792458.0
block_width = 16
block_height = 16
# Matlab data loading
data = mlab.helper3DSAR()
#data = mlab.helperMTI()
data = mlab.rangeCompress(data)
mdouble = mlab.double
do = lambda x: float(mdouble(x)[0, 0])
nint = numpy.int32
nfloat = numpy.float32
rp = mdouble(data.upsampled_range_profiles)
im = numpy.zeros_like(data.im_final).astype(numpy.complex64)
[Nimg_height, Nimg_width] = im.shape
delta_pixel_x = numpy.diff(data.x_vec)[0, 0]
delta_pixel_y = numpy.diff(data.y_vec)[0, 0]
c__4_deltaF = clight / (4.0 * do(data.deltaF))
# --------------------------------------------------------------------------- #
# example.py
# Tarik Tosun, 2012-07-19
# Description:
# Demonstrates what you can do with python-retargeter.
# --------------------------------------------------------------------------- #
import numpy as np
import scipy as sp
from mlabwrap import mlab
# when mlabwrap is imported, it starts an instance of matlab which runs in the
# background. Nearly any matlab function may be called as: 'mlab.[function]'
mlab.addpath(mlab.genpath('../minimal'))
# ----------------------------------------------------------- #
# Creating Kinematic Chain Models:
# ----------------------------------------------------------- #
# pr2larm(length_total) returns a simulated PR2 left arm of specified total
# length. (pr2rarm returns a simulated PR2 right arm.)
# Degrees of freedom:
# [shoulder_yaw, shoulder_tilt, shoulder_roll, elbow_flex]
L = 300
pr2 = mlab.pr2rarm(L)
# human24(lengths) returns a 2-link, 4-dof chain similar to a human arm. the
# lengths vector specifies link lengths.
# Degrees of freedom:
# [shoulder_roll, shoulder_yaw, shoulder_pitch, elbow_pitch]
from mlabwrap import mlab as matlab
# give an interface to the mlab system through cobra.matlab
cobra.matlab = matlab
matlab.__doc__ = """
This is an mlabwrap connection to MATLAB which can be used to call
MATLAB functions. For example, if model is a python model, the following
can be used to optimize the model in MATLAB:
> matlab_model = cobra.mlab.cobra_model_object_to_cobra_matlab_struct(model)
> result = cobra.matlab.optimizeCbModel(matlab_model)
Any MATLAB function can be called this way"""
# add path with module's python scripts to the MATLAB path
mlab_path = os.path.join(cobra.__path__[0], 'mlab', 'matlab_scripts')
matlab.addpath(mlab_path)
_possible_cobra_locations = ["~/MATLAB/cobra", "~/cobra",
"~/Documents/MATLAB/cobra", "~/Documents/opencobra/matlab/cobra"]
def init_matlab_toolbox(matlab_cobra_path=None, discover_functions=True):
"""initialize the matlab cobra toolbox, and load its functions
into mlab's namespace (very useful for ipython tab completion)
matlab_cobra_path: the path to the directory containing the MATLAB
cobra installation. Using the default None will attempt to find the
toolbox in the MATLAB path
discover_functions: Whether mlabwrap should autodiscover all cobra toolbox
functions in matlab. This is convenient for tab completion, but may take
some time."""
from mlabwrap import mlab
import pycuda.autoinit
import pycuda.driver as drv
import numpy
from pycuda.compiler import SourceModule
mlab.addpath('../FasihSarStuff/', nout=0)
# Params
clight = 299792458.0
block_width = 16
block_height = 16
# Matlab data loading
data = mlab.helper3DSAR()
#data = mlab.helperMTI()
data = mlab.rangeCompress(data)
mdouble = mlab.double;
do = lambda x: float(mdouble(x)[0,0])
nint = numpy.int32
nfloat = numpy.float32
rp = mdouble(data.upsampled_range_profiles)
im = numpy.zeros_like(data.im_final).astype(numpy.complex64)
[Nimg_height, Nimg_width] = im.shape
delta_pixel_x = numpy.diff(data.x_vec)[0,0]
delta_pixel_y = numpy.diff(data.y_vec)[0,0]
c__4_deltaF = clight / (4.0 * do(data.deltaF))
except ImportError:
print 'Unable to import mlab module. Attempting to install...'
os.system('cd %s; python setup.py build' % MLABWRAPDIR)
basedir = '%s/build/' % MLABWRAPDIR
sys.path.extend([os.path.join(basedir, x) for x in os.listdir(basedir)
if x.startswith('lib')])
from mlabwrap import mlab
try:
mlab.sin(1)
except:
# Re-initialize the broken connection to the Matlab engine.
import mlabraw
mlab._session = mlabraw.open()
mlab.addpath(os.path.join(CURRDIR, 'coversongs'))
def extract_features(track, fctr=400, fsd=1.0, type=1):
"""Computes beat-synchronous chroma features.
Uses Dan Ellis' chrombeatftrs Matlab function (via the mlabwrap
module, which is included with this feature extractor).
See http://labrosa.ee.columbia.edu/projects/coversongs
for more details.
Parameters
----------
track : gordon Track instance
fctr : float
import optparse
import os
import sys
import numpy as np
import scipy as sp
import scipy.io
import plca
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger('segmenter')
try:
from mlabwrap import mlab
mlab.addpath('coversongs')
except:
logger.warning('Unable to import mlab module. Feature extraction '
'and evaluation will not work.')
def extract_features(wavfilename, fctr=400, fsd=1.0, type=1):
"""Computes beat-synchronous chroma features from the given wave file
Calls Dan Ellis' chrombeatftrs Matlab function.
"""
x, fs = mlab.wavread(wavfilename, nout=2)
feats, beats = mlab.chrombeatftrs(x, fs, fctr, fsd, type, nout=2)
return feats, beats.flatten()
import os
import sys
import numpy as np
import scipy as sp
import scipy.io
import plca
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("segmenter")
try:
from mlabwrap import mlab
mlab.addpath("coversongs")
except:
logger.warning("Unable to import mlab module. Feature extraction " "and evaluation will not work.")
def extract_features(wavfilename, fctr=400, fsd=1.0, type=1):
"""Computes beat-synchronous chroma features from the given wave file
Calls Dan Ellis' chrombeatftrs Matlab function.
"""
x, fs = mlab.wavread(wavfilename, nout=2)
feats, beats = mlab.chrombeatftrs(x, fs, fctr, fsd, type, nout=2)
return feats, beats.flatten()
def segment_song(
Also, can use codebook encoding. T. Bertin-Mahieux (2010) Columbia University [email protected] """ import os import sys import glob import numpy as np import scipy.io as sio rondir = 'ronwsiplca' from ronwsiplca import segmenter as SEGMENTER from mlabwrap import mlab mlab.addpath(os.path.abspath(rondir)) mlab.addpath(os.path.abspath('.')) # beatles directories on my machines (TBM) _enfeats_dir = os.path.expanduser('~/Columbia/InfiniteListener/beatles_enbeatfeats') _audio_dir = os.path.expanduser('~/Columbia/InfiniteListener/beatles_audio') _seglab_dir = os.path.expanduser('~/Columbia/InfiniteListener/beatles_seglab') def get_all_files(basedir,pattern='*.wav') : """ From a root directory, go through all subdirectories and find all files that fit the pattern. Return them in a list. """ allfiles = [] for root, dirs, files in os.walk(basedir):
import os
import sys
import numpy as np
from mlabwrap import mlab
import scikits.audiolab as AUDIOLAB
# makes sure we have the right matlab files
# save their absolute path
_code_dir = os.path.dirname(__file__)
_find_landmarks_path = os.path.join(os.path.abspath(_code_dir),'find_landmarks.m')
if not os.path.exists(_find_landmarks_path):
print "can't find find_lanmarks.m, not same place as get_landmarks?"
print "get_landmarks.py dir:",_code_dir
raise ImportError
mlab.addpath(_code_dir)
def wavread(path):
"""
Wrapper around scikits functions
Returns: wavdata, sample rate, encoding type
See pyaudiolab or scikits.audiolab for more information
"""
return AUDIOLAB.wavread(path)
def find_landmarks_from_wav(wavpath):
"""
utility function, open wav, calls find_landmarks
"""
wav = wavread(wavpath)
Also, can use codebook encoding. T. Bertin-Mahieux (2010) Columbia University [email protected] """ import os import sys import glob import numpy as np import scipy.io as sio rondir = 'ronwsiplca' from ronwsiplca import segmenter as SEGMENTER from mlabwrap import mlab mlab.addpath(os.path.abspath(rondir)) mlab.addpath(os.path.abspath('.')) # beatles directories on my machines (TBM) _enfeats_dir = os.path.expanduser( '~/Columbia/InfiniteListener/beatles_enbeatfeats') _audio_dir = os.path.expanduser('~/Columbia/InfiniteListener/beatles_audio') _seglab_dir = os.path.expanduser('~/Columbia/InfiniteListener/beatles_seglab') def get_all_files(basedir, pattern='*.wav'): """ From a root directory, go through all subdirectories and find all files that fit the pattern. Return them in a list. """ allfiles = []
import os
import sys
import numpy as np
from mlabwrap import mlab
import scikits.audiolab as AUDIOLAB
# makes sure we have the right matlab files
# save their absolute path
_code_dir = os.path.dirname(__file__)
_find_landmarks_path = os.path.join(os.path.abspath(_code_dir), "find_landmarks.m")
if not os.path.exists(_find_landmarks_path):
print "can't find find_lanmarks.m, not same place as get_landmarks?"
print "get_landmarks.py dir:", _code_dir
raise ImportError
mlab.addpath(_code_dir)
def wavread(path):
"""
Wrapper around scikits functions
Returns: wavdata, sample rate, encoding type
See pyaudiolab or scikits.audiolab for more information
"""
return AUDIOLAB.wavread(path)
def find_landmarks_from_wav(wavpath):
"""
utility function, open wav, calls find_landmarks
"""
"""
import os
import sys
import tempfile
import numpy as np
# MATLAB wrapper, we remove the dperecation warnings
import warnings
warnings.filterwarnings('ignore', category=DeprecationWarning)
from mlabwrap import mlab
warnings.filterwarnings('default', category=DeprecationWarning)
# kalman toolbox path
kal_toolbox_path = '/home/thierry/Columbia/Imputation/PythonSrc/KalmanAll'
mlab.warning('off', 'all') # hack, remove warnings
# some functions are redefined but who cares!
mlab.addpath(kal_toolbox_path)
mlab.addpath(os.path.join(kal_toolbox_path, 'Kalman'))
mlab.addpath(os.path.join(kal_toolbox_path, 'KPMstats'))
mlab.addpath(os.path.join(kal_toolbox_path, 'KPMtools'))
mlab.warning('on', 'all')
def learn_kalman(data, A, C, Q, R, initx, initV, niter, diagQ=1, diagR=1):
"""
Main function, take initial parameters and train a Kalman filter.
We assume a model:
x(t+1) = A*x(t) + w(t), w ~ N(0, Q), x(0) ~ N(init_x, init_V)
y(t) = C*x(t) + v(t), v ~ N(0, R)
INPUT
data - one sequence, one observation per col
A - DxD matrix, D dimension of hidden state