Python keras.initializations.normal() Examples

The following are 26 code examples of keras.initializations.normal(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module keras.initializations , or try the search function .
Example #1
Source File: vaegan_cifar.py    From MachineLearning with Apache License 2.0 6 votes vote down vote up 
def generator(batch_size, gf_dim, ch, rows, cols):
    model = Sequential()

    model.add(
        Dense(gf_dim * 8 * rows[0] * cols[0], batch_input_shape=(batch_size, z_dim), name="g_h0_lin", init=normal))
    model.add(Reshape((rows[0], cols[0], gf_dim * 8)))
    model.add(BN(mode=2, axis=3, name="g_bn0", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconvolution2D(gf_dim * 4, 5, 5, output_shape=(batch_size, rows[1], cols[1], gf_dim * 4), subsample=(2, 2),
                              name="g_h1", border_mode="same", init=normal))
    model.add(BN(mode=2, axis=3, name="g_bn1", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconvolution2D(gf_dim * 2, 5, 5, output_shape=(batch_size, rows[2], cols[2], gf_dim * 2), subsample=(2, 2),
                              name="g_h2", border_mode="same", init=normal))
    model.add(BN(mode=2, axis=3, name="g_bn2", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconvolution2D(ch, 5, 5, output_shape=(batch_size, rows[3], cols[3], ch), subsample=(2, 2), name="g_h3",
                              border_mode="same", init=normal))
    model.add(Activation("tanh"))

    return model
Example #2
Source File: vaegan_cifar.py    From MachineLearning with Apache License 2.0 6 votes vote down vote up 
def encoder(batch_size, df_dim, ch, rows, cols):
    model = Sequential()
    X = Input(batch_shape=(batch_size, rows[-1], cols[-1], ch))
    model = Convolution2D(df_dim, 5, 5, subsample=(2, 2), border_mode="same",
                          name="e_h0_conv", dim_ordering="tf", init=normal)(X)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim * 2, 5, 5, subsample=(2, 2), border_mode="same",
                          name="e_h1_conv", dim_ordering="tf")(model)
    model = BN(mode=2, axis=3, name="e_bn1", gamma_init=mean_normal, epsilon=1e-5)(model)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim * 4, 5, 5, subsample=(2, 2), name="e_h2_conv", border_mode="same",
                          dim_ordering="tf", init=normal)(model)
    model = BN(mode=2, axis=3, name="e_bn2", gamma_init=mean_normal, epsilon=1e-5)(model)
    model = LeakyReLU(.2)(model)
    model = Flatten()(model)

    mean = Dense(z_dim, name="e_h3_lin", init=normal)(model)
    logsigma = Dense(z_dim, name="e_h4_lin", activation="tanh", init=normal)(model)
    meansigma = Model([X], [mean, logsigma])
    return meansigma
Example #3
Source File: vaegan_cifar.py    From MachineLearning with Apache License 2.0 6 votes vote down vote up 
def discriminator(batch_size, df_dim, ch, rows, cols):
    X = Input(batch_shape=(batch_size, rows[-1], cols[-1], ch))
    model = Convolution2D(df_dim, 5, 5, subsample=(2, 2), border_mode="same",
                          name="d_h0_conv", dim_ordering="tf", init=normal)(X)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim * 2, 5, 5, subsample=(2, 2), border_mode="same",
                          name="d_h1_conv", dim_ordering="tf", init=normal)(model)
    model = BN(mode=2, axis=3, name="d_bn1", gamma_init=mean_normal, epsilon=1e-5)(model)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim * 4, 5, 5, subsample=(2, 2), border_mode="same",
                          name="d_h2_conv", dim_ordering="tf", init=normal)(model)

    dec = BN(mode=2, axis=3, name="d_bn3", gamma_init=mean_normal, epsilon=1e-5)(model)
    dec = LeakyReLU(.2)(dec)
    dec = Flatten()(dec)
    dec = Dense(1, name="d_h3_lin", init=normal)(dec)

    output = Model([X], [dec, model])

    return output
Example #4
Source File: vaegan_svhn.py    From MachineLearning with Apache License 2.0 6 votes vote down vote up 
def generator(batch_size, gf_dim, ch, rows, cols):
    model = Sequential()

    model.add(
        Dense(gf_dim * 8 * rows[0] * cols[0], batch_input_shape=(batch_size, z_dim), name="g_h0_lin", init=normal))
    model.add(Reshape((rows[0], cols[0], gf_dim * 8)))
    model.add(BN(mode=2, axis=3, name="g_bn0", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconvolution2D(gf_dim * 4, 5, 5, output_shape=(batch_size, rows[1], cols[1], gf_dim * 4), subsample=(2, 2),
                              name="g_h1", border_mode="same", init=normal))
    model.add(BN(mode=2, axis=3, name="g_bn1", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconvolution2D(gf_dim * 2, 5, 5, output_shape=(batch_size, rows[2], cols[2], gf_dim * 2), subsample=(2, 2),
                              name="g_h2", border_mode="same", init=normal))
    model.add(BN(mode=2, axis=3, name="g_bn2", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconvolution2D(ch, 5, 5, output_shape=(batch_size, rows[3], cols[3], ch), subsample=(2, 2), name="g_h3",
                              border_mode="same", init=normal))
    model.add(Activation("tanh"))

    return model
Example #5
Source File: vaegan_svhn.py    From MachineLearning with Apache License 2.0 6 votes vote down vote up 
def encoder(batch_size, df_dim, ch, rows, cols):
    model = Sequential()
    X = Input(batch_shape=(batch_size, rows[-1], cols[-1], ch))
    model = Convolution2D(df_dim, 5, 5, subsample=(2, 2), border_mode="same",
                          name="e_h0_conv", dim_ordering="tf", init=normal)(X)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim * 2, 5, 5, subsample=(2, 2), border_mode="same",
                          name="e_h1_conv", dim_ordering="tf")(model)
    model = BN(mode=2, axis=3, name="e_bn1", gamma_init=mean_normal, epsilon=1e-5)(model)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim * 4, 5, 5, subsample=(2, 2), name="e_h2_conv", border_mode="same",
                          dim_ordering="tf", init=normal)(model)
    model = BN(mode=2, axis=3, name="e_bn2", gamma_init=mean_normal, epsilon=1e-5)(model)
    model = LeakyReLU(.2)(model)
    model = Flatten()(model)

    mean = Dense(z_dim, name="e_h3_lin", init=normal)(model)
    logsigma = Dense(z_dim, name="e_h4_lin", activation="tanh", init=normal)(model)
    meansigma = Model([X], [mean, logsigma])
    return meansigma
Example #6
Source File: vaegan_svhn.py    From MachineLearning with Apache License 2.0 6 votes vote down vote up 
def discriminator(batch_size, df_dim, ch, rows, cols):
    X = Input(batch_shape=(batch_size, rows[-1], cols[-1], ch))
    model = Convolution2D(df_dim, 5, 5, subsample=(2, 2), border_mode="same",
                          name="d_h0_conv", dim_ordering="tf", init=normal)(X)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim * 2, 5, 5, subsample=(2, 2), border_mode="same",
                          name="d_h1_conv", dim_ordering="tf", init=normal)(model)
    model = BN(mode=2, axis=3, name="d_bn1", gamma_init=mean_normal, epsilon=1e-5)(model)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim * 4, 5, 5, subsample=(2, 2), border_mode="same",
                          name="d_h2_conv", dim_ordering="tf", init=normal)(model)

    dec = BN(mode=2, axis=3, name="d_bn3", gamma_init=mean_normal, epsilon=1e-5)(model)
    dec = LeakyReLU(.2)(dec)
    dec = Flatten()(dec)
    dec = Dense(1, name="d_h3_lin", init=normal)(dec)

    output = Model([X], [dec, model])

    return output
Example #7
Source File: autoencoder.py    From research with BSD 3-Clause "New" or "Revised" License 6 votes vote down vote up 
def generator(batch_size, gf_dim, ch, rows, cols):

    model = Sequential()

    model.add(Dense(gf_dim*8*rows[0]*cols[0], batch_input_shape=(batch_size, z_dim), name="g_h0_lin", init=normal))
    model.add(Reshape((rows[0], cols[0], gf_dim*8)))
    model.add(BN(mode=2, axis=3, name="g_bn0", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconv2D(gf_dim*4, 5, 5, subsample=(2, 2), name="g_h1", init=normal))
    model.add(BN(mode=2, axis=3, name="g_bn1", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconv2D(gf_dim*2, 5, 5, subsample=(2, 2), name="g_h2", init=normal))
    model.add(BN(mode=2, axis=3, name="g_bn2", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconv2D(gf_dim, 5, 5, subsample=(2, 2), name="g_h3", init=normal))
    model.add(BN(mode=2, axis=3, name="g_bn3", gamma_init=mean_normal, epsilon=1e-5))
    model.add(Activation("relu"))

    model.add(Deconv2D(ch, 5, 5, subsample=(2, 2), name="g_h4", init=normal))
    model.add(Activation("tanh"))

    return model
Example #8
Source File: cppn.py    From cppn-keras with MIT License 5 votes vote down vote up 
def my_init(shape, name=None):
    return initializations.normal(shape, scale=1.2, name=name)
Example #9
Source File: vaegan_cifar.py    From MachineLearning with Apache License 2.0 5 votes vote down vote up 
def mean_normal(shape, mean=1., scale=0.02, name=None):
    return K.variable(np.random.normal(loc=mean, scale=scale, size=shape), name=name)
Example #10
Source File: audiounet.py    From audio-super-res with MIT License 5 votes vote down vote up 
def normal_init(shape, dim_ordering='tf', name=None):
    return normal(shape, scale=1e-3, name=name, dim_ordering=dim_ordering)
Example #11
Source File: audiotfilm.py    From audio-super-res with MIT License 5 votes vote down vote up 
def normal_init(shape, dim_ordering='tf', name=None):
    return normal(shape, scale=1e-3, name=name, dim_ordering=dim_ordering)
Example #12
Source File: dnn.py    From audio-super-res with MIT License 5 votes vote down vote up 
def normal_init(shape, dim_ordering='tf', name=None):
    return normal(shape, scale=0.0000001, name=name, dim_ordering=dim_ordering)
Example #13
Source File: LUNA_unet.py    From Luna2016-Lung-Nodule-Detection with MIT License 5 votes vote down vote up 
def gaussian_init(shape, name=None, dim_ordering=None):
   return initializations.normal(shape, scale=0.001, name=name, dim_ordering=dim_ordering)
Example #14
Source File: transition.py    From research with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def cleanup(data):
  X = data[0]
  sh = X.shape
  X = X.reshape((-1, 3, 160, 320))
  X = np.asarray([cv2.resize(x.transpose(1, 2, 0), (160, 80)) for x in X])
  X = X/127.5 - 1.
  X = X.reshape((sh[0], (time+out_leng)*4, 80, 160, 3))
  Z = np.random.normal(0, 1, (X.shape[0], z_dim))
  return Z, X[:, ::4]
Example #15
Source File: transition.py    From research with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def mean_normal(shape, mean=1., scale=0.02, name=None):
    return K.variable(np.random.normal(loc=mean, scale=scale, size=shape), name=name)
Example #16
Source File: autoencoder.py    From research with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def encoder(batch_size, df_dim, ch, rows, cols):

    model = Sequential()
    X = Input(batch_shape=(batch_size, rows[-1], cols[-1], ch))
    model = Convolution2D(df_dim, 5, 5, subsample=(2, 2), border_mode="same",
                          name="e_h0_conv", dim_ordering="tf", init=normal)(X)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim*2, 5, 5, subsample=(2, 2), border_mode="same",
                          name="e_h1_conv", dim_ordering="tf")(model)
    model = BN(mode=2, axis=3, name="e_bn1", gamma_init=mean_normal, epsilon=1e-5)(model)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim*4, 5, 5, subsample=(2, 2), name="e_h2_conv", border_mode="same",
                          dim_ordering="tf", init=normal)(model)
    model = BN(mode=2, axis=3, name="e_bn2", gamma_init=mean_normal, epsilon=1e-5)(model)
    model = LeakyReLU(.2)(model)

    model = Convolution2D(df_dim*8, 5, 5, subsample=(2, 2), border_mode="same",
                          name="e_h3_conv", dim_ordering="tf", init=normal)(model)
    model = BN(mode=2, axis=3, name="e_bn3", gamma_init=mean_normal, epsilon=1e-5)(model)
    model = LeakyReLU(.2)(model)
    model = Flatten()(model)

    mean = Dense(z_dim, name="e_h3_lin", init=normal)(model)
    logsigma = Dense(z_dim, name="e_h4_lin", activation="tanh", init=normal)(model)
    meansigma = Model([X], [mean, logsigma])
    return meansigma
Example #17
Source File: autoencoder.py    From research with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def cleanup(data):
  X = data[0][:64, -1]
  X = np.asarray([cv2.resize(x.transpose(1, 2, 0), (160, 80)) for x in X])
  X = X/127.5 - 1.
  Z = np.random.normal(0, 1, (X.shape[0], z_dim))
  return Z, X
Example #18
Source File: autoencoder.py    From research with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def mean_normal(shape, mean=1., scale=0.02, name=None):
    return K.variable(np.random.normal(loc=mean, scale=scale, size=shape), name=name)
Example #19
Source File: conditional.py    From research with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def mean_normal(shape, mean=1., scale=0.02, name=None):
    return K.variable(np.random.normal(loc=mean, scale=scale, size=shape), name=name)
Example #20
Source File: reinforcement.py    From detection-2016-nipsws with MIT License 5 votes vote down vote up 
def get_q_network(weights_path):
    model = Sequential()
    model.add(Dense(1024, init=lambda shape, name: normal(shape, scale=0.01, name=name), input_shape=(25112,)))
    model.add(Activation('relu'))
    model.add(Dropout(0.2))
    model.add(Dense(1024, init=lambda shape, name: normal(shape, scale=0.01, name=name)))
    model.add(Activation('relu'))
    model.add(Dropout(0.2))
    model.add(Dense(6, init=lambda shape, name: normal(shape, scale=0.01, name=name)))
    model.add(Activation('linear'))
    adam = Adam(lr=1e-6)
    model.compile(loss='mse', optimizer=adam)
    if weights_path != "0":
        model.load_weights(weights_path)
    return model
Example #21
Source File: model.py    From neural-style-keras with MIT License 5 votes vote down vote up 
def weights_init(shape, name=None, dim_ordering=None):
    return normal(shape, scale=0.01, name=name)
Example #22
Source File: NeuMF.py    From neural_collaborative_filtering with Apache License 2.0 5 votes vote down vote up 
def init_normal(shape, name=None):
    return initializations.normal(shape, scale=0.01, name=name)
Example #23
Source File: GMF.py    From neural_collaborative_filtering with Apache License 2.0 5 votes vote down vote up 
def init_normal(shape, name=None):
    return initializations.normal(shape, scale=0.01, name=name)
Example #24
Source File: MLP.py    From neural_collaborative_filtering with Apache License 2.0 5 votes vote down vote up 
def init_normal(shape, name=None):
    return initializations.normal(shape, scale=0.01, name=name)
Example #25
Source File: vaegan_svhn.py    From MachineLearning with Apache License 2.0 5 votes vote down vote up 
def fetch_next_batch(s):
    z = np.random.normal(0., 1., (batch_size, z_dim))  # normal dist for GAN
    x = s.train.next_batch(batch_size)
    return z, x[0]
Example #26
Source File: vaegan_cifar.py    From MachineLearning with Apache License 2.0 5 votes vote down vote up 
def fetch_next_batch(cifar):
    z = np.random.normal(0., 1., (batch_size, z_dim))  # normal dist for GAN
    x = cifar.train.next_batch(batch_size)
    return z, x[0]