BootstrapSampler takes in a ratio as sample size parameter

Engine/src/simulator/sampler/abstract_sampler.py

@@ -20,19 +20,19 @@ def __init__(self):
         Constructor
         '''
         self.dataset = None
-        self.nrObs = None
+        self.nr_obs = None
         self.data_hist = None
         self.sample_hists = None
-        self.nrSamples = None
+        self.nr_samples = None
 
     def bind_data(self, dataset):
         '''
         Bind dataset to Sampler in order to record statistics during sampling
         '''
         self.dataset = dataset
-        self.nrObs = len(dataset)
-        self.nrSamples = 0
-        self.data_hist = np.zeros(self.nrObs)   # how often each element has been sampled
+        self.nr_obs = len(dataset)
+        self.nr_samples = 0
+        self.data_hist = np.zeros(self.nr_obs)   # how often each element has been sampled
         self.sample_hists = []              # array of histograms for each sample
 
     @abstractmethod
@@ -46,42 +46,42 @@ def add_sample_histogram(self, sample_hist_arr):
         '''
         Add histogram for sample to the pool of sample histograms
         '''
-        assert len(sample_hist_arr) == self.nrObs, "Sample histogram doesn't contain all obs from the dataset"
+        assert len(sample_hist_arr) == self.nr_obs, "Sample histogram doesn't contain all obs from the dataset"
         self.sample_hists.append(sample_hist_arr)
         self.data_hist += sample_hist_arr
-        self.nrSamples += 1
+        self.nr_samples += 1
 
     def plot_data_histogram(self):
         '''
         Plot histogram showing the total sampled data
         '''
-        plt.plot(range(0,self.nrObs), self.data_hist)
+        plt.plot(range(0,self.nr_obs), self.data_hist)
         plt.show()
 
     def plot_sample_histogram(self, sample_number):
         '''
         Samples generated are indexed starting from 0
         '''
-        plt.plot(range(0,self.nrObs), self.sample_hists[sample_number])
+        plt.plot(range(0,self.nr_obs), self.sample_hists[sample_number])
         plt.show()
 
     def plot_sample_histograms(self):
         '''
         Plot all sample histograms on different subplots
-        Cracks for nrSamples < 3 (since subplot is then one dimensional), so return None
+        Cracks for nr_samples < 3 (since subplot is then one dimensional), so return None
         '''
-        if self.nrSamples < 3: return
-        hEdgeSubplot= int(math.ceil(math.sqrt(self.nrSamples)))
-        vEdgeSubplot= int(round(math.sqrt(self.nrSamples)))
+        if self.nr_samples < 3: return
+        hEdgeSubplot= int(math.ceil(math.sqrt(self.nr_samples)))
+        vEdgeSubplot= int(round(math.sqrt(self.nr_samples)))
         fig, axarr = plt.subplots(vEdgeSubplot, hEdgeSubplot, sharex = True, sharey = True)
-        for i in range(self.nrSamples):
+        for i in range(self.nr_samples):
             ix = i/hEdgeSubplot
             iy = i%hEdgeSubplot
-            axarr[ix, iy].plot(range(self.nrObs), self.sample_hists[i])
+            axarr[ix, iy].plot(range(self.nr_obs), self.sample_hists[i])
             axarr[ix,iy].axes.get_xaxis().set_visible(False)
             axarr[ix,iy].axes.get_yaxis().set_visible(False)
         #delete empty subplots
-        for i in range(self.nrSamples, vEdgeSubplot*hEdgeSubplot):
+        for i in range(self.nr_samples, vEdgeSubplot*hEdgeSubplot):
             ix = i/hEdgeSubplot
             iy = i%hEdgeSubplot
             fig.delaxes(axarr[ix,iy])

Engine/src/simulator/sampler/bootstrap_sampler.py

@@ -9,17 +9,27 @@
 import random
 
 class BootstrapSampler(AbstractSampler):
+    '''
+    Samples with replacement from given data at a size_ratio specified in the constructor
+    '''
 
-    def __init__(self, sampleSize = None):
+    def __init__(self, sample_size_ratio = None):
         '''
         Constructor
+        @param sample_ratio: size of the bootstrap sample as a percentage of the dataset's size [0,1].
+        Default value is to sample the same nr as the dataset's size (1)
         '''
-        self.sampleSize = sampleSize
+        if sample_size_ratio is not None and (sample_size_ratio < 0 or sample_size_ratio > 1):
+            raise Exception("sample_size_ratio not between 0 and 1")
+        self.sample_size_ratio = sample_size_ratio
+        self.sample_size = None
 
     def bind_data(self, dataset):
         super(type(self), self).bind_data(dataset)
-        if self.sampleSize is None: 
-            self.sampleSize = self.nrObs
+        if self.sample_size_ratio is None: 
+            self.sample_size_ratio = 1
+
+        self.sample_size = int(round(self.sample_size_ratio * self.nr_obs))
 
     def sample(self):
         '''
@@ -29,11 +39,11 @@ def sample(self):
         if self.dataset is None:
             raise Exception("Data not bound")
 
-        sample_hist = self.nrObs * [0]
+        sample_hist = self.nr_obs * [0]
         sample_data = []   # emtpy list, vstack later
 
-        for _ in range(self.sampleSize):
-            index = random.randint(0, self.nrObs-1)
+        for _ in range(self.sample_size):
+            index = random.randint(0, self.nr_obs-1)
             sample_data.append(self.dataset[index])
             sample_hist[index] += 1
 

Engine/src/tests/simulator/sampler/abstract_sampler_test.py

@@ -20,18 +20,17 @@ def setUp(self):
         self.asmpl.bind_data(self.dataset)
 
     def test_bind_data(self):
-        assert self.asmpl.nrObs == 3
+        assert self.asmpl.nr_obs == 3
         assert self.asmpl.dataset is not None
         assert (self.asmpl.data_hist == [0,0,0]).all()
-        assert self.asmpl.nrSamples == 0
+        assert self.asmpl.nr_samples == 0
 
     def test_add_sample_histogram(self):
         self.asmpl.add_sample_histogram([1,1,2])
         self.asmpl.add_sample_histogram([1,2,5])
-        print self.asmpl.data_hist
         assert (self.asmpl.data_hist == [2,3,7]).all()
         assert len(self.asmpl.sample_hists) == 2
-        assert self.asmpl.nrSamples == 2
+        assert self.asmpl.nr_samples == 2
 
     '''
     #Visual Test

Engine/src/tests/simulator/sampler/bootstrap_sampler_test.py

@@ -19,31 +19,29 @@ def tearDown(self):
 
 
     def test_constructor(self):
-        bs = BootstrapSampler(100)
-        assert bs.sampleSize == 100
+        bs = BootstrapSampler(0.5)
+        assert bs.sample_size_ratio == 0.5
+        self.assertRaises(Exception, BootstrapSampler, -0.1)
+        self.assertRaises(Exception, BootstrapSampler, 10)
 
     def test_data_not_bound(self):
-        bs = BootstrapSampler(100)
-        try:
-            bs.sample()
-        except Exception:
-            return
-        assert False
+        bs = BootstrapSampler()
+        self.assertRaises(Exception, bs.sample)
 
-    def test_sampleSize(self):
+    def test_sample_size(self):
         bs = BootstrapSampler()
-        assert bs.sampleSize == None
+        assert bs.sample_size == None
         bs.bind_data(self.data)
-        assert bs.sampleSize == self.data.shape[0]
+        assert bs.sample_size == self.data.shape[0]
 
     def test_sample(self):
-        bs = BootstrapSampler(100)
+        bs = BootstrapSampler(0.5)
         bs.bind_data(self.data)
         sample = bs.sample()
 
-        assert bs.nrSamples == 1
+        assert bs.nr_samples == 1
         assert len(bs.sample_hists) == 1
-        assert sample.shape[0] == 100
+        assert sample.shape[0] == int(round(0.5 * self.data.shape[0]))
         assert sample.shape[1] == self.data.shape[1]
         assert not (bs.data_hist == np.zeros(self.data.shape[0])).all()
     '''