remove unused featurization code

ecg/featurize.py

@@ -1,9 +1,7 @@
 from sklearn import preprocessing
-import pywt
 import numpy as np
 from tqdm import tqdm
 import warnings
-import scipy.signal as scs
 
 class Normalizer(object):
     def __init__(self, strategy):
@@ -38,66 +36,3 @@ def transform(self, x):
         new_shape = (x.shape[0]*x.shape[1], x.shape[2])
         return self.scaler.transform(
             x.reshape(new_shape)).reshape(original_shape)
-
-
-class BandPassFilter(object):
-    def filt(self, data, lowcut=0.5, highcut=90, fs=200, order=5):
-        """
-        http://scipy.github.io/old-wiki/pages/Cookbook/ButterworthBandpass
-        """
-        nyq = 0.5 * fs
-        low = lowcut / nyq
-        high = highcut / nyq
-        b, a = scs.butter(order, [low, high], btype='bandpass')
-        y = scs.lfilter(b, a, data)
-        return y
-
-    def transform(self, x):
-        print('Applying Butterworth Filter...')
-        return np.array([self.filt(x_indiv) for x_indiv in x])
-
-
-class DiscreteWaveletTransformer(object):
-    def __init__(self, wavelet_fns, level):
-        self.transforms = wavelet_fns
-        self.level = level
-
-    def transform(self, x):
-        print('Applying Wavelet Transformations...')
-        x_new = []
-        for x_indiv in tqdm(x):
-            x_indiv_trans = []
-            for wavefn in self.transforms:
-                transform = np.array(pywt.wavedec(
-                    x_indiv, wavefn, level=self.level)[:2])
-                if(len(x_indiv_trans) > 0 and
-                        transform.shape[1] != len(x_indiv_trans[0])):
-                    warnings.warn(
-                        "Reshaping to proper length after wavelet transform")
-                    transform = transform[:, :len(x_indiv_trans[0])]
-                x_indiv_trans.extend(transform)
-            x_new.append(np.array(x_indiv_trans).T)
-        x_new = np.array(x_new)
-        return x_new
-
-
-class ContinuousWaveletTransformer(object):
-    def __init__(self, wavelet_fns):
-        self.transforms = wavelet_fns
-        self.widths = np.linspace(1, 100, 10)  # TODO: parameterize
-
-    def transform(self, x):
-        print('Applying Wavelet Transformations...')
-        x_new = []
-        for x_indiv in tqdm(x):
-            x_indiv_trans = []
-            for wavefn in self.transforms:
-                transform, _ = pywt.cwt(x_indiv, self.widths, 'mexh')
-                if(transform.shape[1] != len(x_indiv)):
-                    warnings.warn(
-                        "Reshaping to proper length after wavelet transform")
-                    transform = transform[:, :len(x_indiv)]
-                x_indiv_trans.extend(transform)
-            x_new.append(np.array(x_indiv_trans).T)
-        x_new = np.array(x_new)
-        return x_new

ecg/load.py

@@ -2,8 +2,6 @@
 from __future__ import division
 from __future__ import absolute_import
 
-from builtins import zip
-from builtins import range
 import json
 import numpy as np
 import fnmatch
@@ -17,7 +15,6 @@
 
 # FIXME: step and samp_rate and duration should be part of process, not load
 
-
 class Loader(object):
     def __init__(
             self,

ecg/process.py

@@ -1,56 +1,29 @@
 from __future__ import division
 from __future__ import print_function
-from builtins import dict
+
 import numpy as np
 import featurize
 import collections
 
-
 class Processor(object):
     def __init__(
         self,
         use_one_hot_labels=True,
         normalizer=False,
-        wavelet_fns=[],
-        wavelet_type='discrete',
         relabel_classes={},
         ignore_classes=[],
-        wavelet_level=1,
-        use_bandpass_filter=False,
         **kwargs
     ):
-        self.wavelet_fns = wavelet_fns
         self.normalizer = normalizer
-        self.wavelet_type = wavelet_type
         self.ignore_classes = ignore_classes
         self.relabel_classes = relabel_classes
-        self.wavelet_level = wavelet_level
         self.use_one_hot_labels = use_one_hot_labels
-        self.use_bandpass_filter = use_bandpass_filter
         self.n = None
         self.classes = None
         self.int_to_class = None
         self.class_to_int = None
 
     def process_x(self, fit):
-        if self.use_bandpass_filter is True:
-            bp_filter = featurize.BandPassFilter()
-            self.x_train = bp_filter.transform(self.x_train)
-            self.x_test = bp_filter.transform(self.x_test)
-
-        if len(self.wavelet_fns) != 0:
-            if (self.wavelet_type == 'discrete'):
-                wavelet_transformer = \
-                    featurize.DiscreteWaveletTransformer(
-                        self.wavelet_fns, self.wavelet_level)
-            elif (self.wavelet_type == 'continuous'):
-                wavelet_transformer = \
-                    featurize.ContinuousWaveletTransformer(self.wavelet_fns)
-            else:
-                raise ValueError("Wavelet type not defined.")
-            self.x_train = wavelet_transformer.transform(self.x_train)
-            self.x_test = wavelet_transformer.transform(self.x_test)
-
         if self.normalizer is not False:
             if fit is True and len(self.x_train) > 0:
                 self.n = featurize.Normalizer(self.normalizer)

ecg/train.py

@@ -2,9 +2,6 @@
 from __future__ import unicode_literals
 from __future__ import division
 from __future__ import absolute_import
-from builtins import open
-from builtins import int
-from builtins import str
 
 import argparse
 import json