Example #1
0
def persistence():

    print "Persistence"
    numpy.random.seed(42)
    n_dims = 2
    n_class = 12
    size = 100
    labels = numpy.random.random_integers(0, 256, size)
    samples = numpy.zeros((size, n_dims), dtype=type)

    print "Generating data"
    for i in range(0, size):
        t = 6.28 * numpy.random.random_sample()
        samples[i][0] = 2 * labels[i] + 1.5 * numpy.cos(t)
        samples[i][1] = 2 * labels[i] + 1.5 * numpy.sin(t)

    print "Creating dense classifier"
    classifier = algo.svm_dense(0, n_dims=n_dims, seed=42)

    print "Adding sample vectors to dense classifier"
    for y, x_list in zip(labels, samples):
        x = numpy.array(x_list, dtype=type)
        classifier.add_sample(float(y), x)

    print "Pickling dense classifier"
    cPickle.dump(classifier, open('test', 'wb'))
    classifier = cPickle.load(open('test', 'rb'))

    print "Training dense classifier"
    classifier.train(gamma=1, C=10, eps=1e-1)

    print "Predicting with dense classifier"
    print classifier.predict(samples[0])

    print "Creating 0/1 classifier"
    classifier01 = algo.svm_01(n_dims=n_dims, seed=42)

    print "Adding sample vectors to 0/1 classifier"
    for y, x_list in zip(labels, samples):
        x = numpy.array(x_list, dtype=type)
        classifier01.add_sample(float(y), x)

    print "Training 0/1 classifier"
    classifier01.train(gamma=1. / 3., C=100, eps=1e-1)

    print "Pickling 0/1 classifier"
    cPickle.dump(classifier01, open('test', 'wb'))
    classifier01 = cPickle.load(open('test', 'rb'))

    print "Predicting with 0/1 classifier"
    print classifier01.predict(numpy.array(samples[0], dtype=type))
Example #2
0
def persistence():
    
    print "Persistence"
    numpy.random.seed(42)
    n_dims = 2
    n_class = 12
    size = 100
    labels = numpy.random.random_integers(0, 256, size)
    samples = numpy.zeros((size, n_dims), dtype=type)

    print "Generating data"
    for i in range(0, size):
        t = 6.28 * numpy.random.random_sample()
        samples[i][0] = 2 * labels[i] + 1.5 * numpy.cos(t)
        samples[i][1] = 2 * labels[i] + 1.5 * numpy.sin(t)

    print "Creating dense classifier"
    classifier = algo.svm_dense(0, n_dims = n_dims, seed=42)
    
    print "Adding sample vectors to dense classifier"
    for y, x_list in zip(labels, samples):
        x = numpy.array(x_list, dtype=type)
        classifier.add_sample(float(y), x)

    print "Pickling dense classifier"
    cPickle.dump(classifier, open('test', 'wb'))
    classifier = cPickle.load(open('test', 'rb'))

    print "Training dense classifier"
    classifier.train(gamma = 1, C = 10, eps=1e-1)

    print "Predicting with dense classifier"
    print classifier.predict(samples[0])

    print "Creating 0/1 classifier"
    classifier01 = algo.svm_01(n_dims = n_dims, seed=42)

    print "Adding sample vectors to 0/1 classifier"
    for y, x_list in zip(labels, samples):
        x = numpy.array(x_list, dtype=type)
        classifier01.add_sample(float(y), x)

    print "Training 0/1 classifier"
    classifier01.train(gamma = 1./3., C = 100, eps=1e-1)

    print "Pickling 0/1 classifier"
    cPickle.dump(classifier01, open('test', 'wb'))
    classifier01 = cPickle.load(open('test', 'rb'))

    print "Predicting with 0/1 classifier"
    print classifier01.predict(numpy.array(samples[0], dtype=type))
Example #3
0
    def testPersistentSize(self):
        for _ in range(2):
            # Multiple of 8 because of sse.
            nDims = 32

            nClass = _RGEN.randint(4, 8)
            size = _RGEN.randint(20, 50)
            labels = _RGEN.random_integers(0, nClass - 1, size)
            samples = _RGEN.normal(size=(size, nDims)).astype(_DTYPE)

            classifier = svm_dense(0, nDims, seed=_SEED, probability=True)

            for y, xList in zip(labels, samples):
                x = np.array(xList, dtype=_DTYPE)
                classifier.add_sample(float(y), x)

            classifier.train(gamma=1.0 / 3.0, C=100, eps=1e-1)
            classifier.cross_validate(2, gamma=0.5, C=10, eps=1e-3)

            s = classifier.__getstate__()
            print("nDims=",
                  nDims,
                  "nClass=",
                  nClass,
                  "n_vectors=",
                  size,
                  end=' ')
            print("dense:", len(s), classifier.persistent_size(), end=' ')
            self.assertEqual(len(s), classifier.persistent_size())

            classifier01 = svm_01(0, nDims, seed=_SEED, probability=True)

            for y, xList in zip(labels, samples):
                x = np.array(xList, dtype=_DTYPE)
                classifier01.add_sample(float(y), x)

            classifier01.train(gamma=1.0 / 3.0, C=100, eps=1e-1)
            classifier01.cross_validate(2, gamma=0.5, C=10, eps=1e-3)

            s = classifier01.__getstate__()
            print("0/1", len(s), classifier01.persistent_size())
            self.assertEqual(len(s), classifier01.persistent_size())
Example #4
0
  def testPersistentSize(self):
    for _ in range(2):
      # Multiple of 8 because of sse.
      nDims = 32

      nClass = _RGEN.randint(4, 8)
      size = _RGEN.randint(20, 50)
      labels = _RGEN.random_integers(0, nClass - 1, size)
      samples = _RGEN.normal(size=(size, nDims)).astype(_DTYPE)

      classifier = svm_dense(0, nDims, seed=_SEED, probability=True)

      for y, xList in zip(labels, samples):
        x = np.array(xList, dtype=_DTYPE)
        classifier.add_sample(float(y), x)

      classifier.train(gamma=1.0/3.0, C=100, eps=1e-1)
      classifier.cross_validate(2, gamma=0.5, C=10, eps=1e-3)

      s = classifier.__getstate__()
      print "nDims=", nDims, "nClass=", nClass, "n_vectors=", size,
      print "dense:", len(s), classifier.persistent_size(),
      self.assertEqual(len(s), classifier.persistent_size())

      classifier01 = svm_01(0, nDims, seed=_SEED, probability=True)

      for y, xList in zip(labels, samples):
        x = np.array(xList, dtype=_DTYPE)
        classifier01.add_sample(float(y), x)

      classifier01.train(gamma=1.0/3.0, C=100, eps=1e-1)
      classifier01.cross_validate(2, gamma=0.5, C=10, eps=1e-3)

      s = classifier01.__getstate__()
      print "0/1", len(s), classifier01.persistent_size()
      self.assertEqual(len(s), classifier01.persistent_size())
Example #5
0
def persistence():

    print("Persistence")
    numpy.random.seed(42)
    n_dims = 2
    n_class = 12
    size = 100
    labels = numpy.random.random_integers(0, 256, size)
    samples = numpy.zeros((size, n_dims), dtype=type)

    print("Generating data")
    for i in range(0, size):
        t = 6.28 * numpy.random.random_sample()
        samples[i][0] = 2 * labels[i] + 1.5 * numpy.cos(t)
        samples[i][1] = 2 * labels[i] + 1.5 * numpy.sin(t)

    print("Creating dense classifier")
    classifier = algo.svm_dense(0, n_dims=n_dims, seed=42)

    print("Adding sample vectors to dense classifier")
    for y, x_list in zip(labels, samples):
        x = numpy.array(x_list, dtype=type)
        classifier.add_sample(float(y), x)

    print("Serializing dense classifier")

    # CHH proto is not supported
    # schema = classifier.getSchema()
    # with open("test", "w+b") as f:
    #     # Save
    #     proto = schema.new_message()
    #     classifier.write(proto)
    #     proto.write(f)
    #
    #     # Load
    #     f.seek(0)
    #     proto2 = schema.read(f)
    #     classifier = algo.svm_dense.read(proto2)

    print("Training dense classifier")
    classifier.train(gamma=1, C=10, eps=1e-1)

    print("Predicting with dense classifier")
    print(classifier.predict(samples[0]))

    print("Creating 0/1 classifier")
    classifier01 = algo.svm_01(n_dims=n_dims, seed=42)

    print("Adding sample vectors to 0/1 classifier")
    for y, x_list in zip(labels, samples):
        x = numpy.array(x_list, dtype=type)
        classifier01.add_sample(float(y), x)

    print("Training 0/1 classifier")
    classifier01.train(gamma=1. / 3., C=100, eps=1e-1)

    print("Serializing 0/1 classifier")

    # CHH proto is not supported
    # schema = classifier01.getSchema()
    # with open("test", "w+b") as f:
    #     # Save
    #     proto = schema.new_message()
    #     classifier01.write(proto)
    #     proto.write(f)
    #
    #     # Load
    #     f.seek(0)
    #     proto2 = schema.read(f)
    #     classifier01 = algo.svm_01.read(proto2)

    print("Predicting with 0/1 classifier")
    print(classifier01.predict(numpy.array(samples[0], dtype=type)))
Example #6
0
def persistence():
    
    print "Persistence"
    numpy.random.seed(42)
    n_dims = 2
    n_class = 12
    size = 100
    labels = numpy.random.random_integers(0, 256, size)
    samples = numpy.zeros((size, n_dims), dtype=type)

    print "Generating data"
    for i in range(0, size):
        t = 6.28 * numpy.random.random_sample()
        samples[i][0] = 2 * labels[i] + 1.5 * numpy.cos(t)
        samples[i][1] = 2 * labels[i] + 1.5 * numpy.sin(t)

    print "Creating dense classifier"
    classifier = algo.svm_dense(0, n_dims = n_dims, seed=42)
    
    print "Adding sample vectors to dense classifier"
    for y, x_list in zip(labels, samples):
        x = numpy.array(x_list, dtype=type)
        classifier.add_sample(float(y), x)

    print "Serializing dense classifier"

    schema = classifier.getSchema()
    with open("test", "w+b") as f:
        # Save
        proto = schema.new_message()
        classifier.write(proto)
        proto.write(f)

        # Load
        f.seek(0)
        proto2 = schema.read(f)
        classifier = algo.svm_dense.read(proto2)

    print "Training dense classifier"
    classifier.train(gamma = 1, C = 10, eps=1e-1)

    print "Predicting with dense classifier"
    print classifier.predict(samples[0])

    print "Creating 0/1 classifier"
    classifier01 = algo.svm_01(n_dims = n_dims, seed=42)

    print "Adding sample vectors to 0/1 classifier"
    for y, x_list in zip(labels, samples):
        x = numpy.array(x_list, dtype=type)
        classifier01.add_sample(float(y), x)

    print "Training 0/1 classifier"
    classifier01.train(gamma = 1./3., C = 100, eps=1e-1)

    print "Serializing 0/1 classifier"
    schema = classifier01.getSchema()
    with open("test", "w+b") as f:
        # Save
        proto = schema.new_message()
        classifier01.write(proto)
        proto.write(f)

        # Load
        f.seek(0)
        proto2 = schema.read(f)
        classifier01 = algo.svm_01.read(proto2)

    print "Predicting with 0/1 classifier"
    print classifier01.predict(numpy.array(samples[0], dtype=type))