Python backend.softmax方法代码示例

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def step(self, x, states):   
        h = states[0]
        # states[1] necessary?

        # equals K.dot(X, self._W1) + self._b2 with X.shape=[bs, T, input_dim]
        total_x_prod = states[-1]
        # comes from the constants (equals the input sequence)
        X = states[-2]
        
        # expand dims to add the vector which is only valid for this time step
        # to total_x_prod which is valid for all time steps
        hw = K.expand_dims(K.dot(h, self._W2), 1)
        additive_atn = total_x_prod + hw
        attention = K.softmax(K.dot(additive_atn, self._V), axis=1)
        x_weighted = K.sum(attention * X, [1])

        x = K.dot(K.concatenate([x, x_weighted], 1), self._W3) + self._b3
        
        h, new_states = self.layer.cell.call(x, states[:-2])
        
        return h, new_states 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def content_addressing(memory_t,  key_vector_t, key_strength_t):
    '''
    Focusing by content.
    :param memory_t: external memory.
    :param key_vector_t: key vector.
    :param key_strength_t: the strength of key.
    :return:
    '''
    # print("content addressing:")
    # print(">>memory_t")
    # print(key_vector_t)
    # print(">>key_vector_t")
    # print(key_vector_t)
    # print(">>key_strength_t")
    # print(key_strength_t)
    _weight_content_t = \
        key_strength_t * cosine_similarity_group(key_vector_t, memory_t)
    weight_content_t = softmax(_weight_content_t)
    # print("_weight_content_t")
    # print(_weight_content_t)
    return weight_content_t 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def labelembed_loss(out1, out2, tar, targets, tau = 2., alpha = 0.9, beta = 0.5, num_classes = 100):
    
    out2_prob = K.softmax(out2)
    tau2_prob = K.stop_gradient(K.softmax(out2 / tau))
    soft_tar = K.stop_gradient(K.softmax(tar))
    
    L_o1_y = K.sparse_categorical_crossentropy(output = K.softmax(out1), target = targets)
    
    pred = K.argmax(out2, axis = -1)
    mask = K.stop_gradient(K.cast(K.equal(pred, K.cast(targets, 'int64')), K.floatx()))
    L_o1_emb = -cross_entropy(out1, soft_tar)  # pylint: disable=invalid-unary-operand-type
    
    L_o2_y = K.sparse_categorical_crossentropy(output = out2_prob, target = targets)
    L_emb_o2 = -cross_entropy(tar, tau2_prob) * mask * (K.cast(K.shape(mask)[0], K.floatx())/(K.sum(mask)+1e-8))  # pylint: disable=invalid-unary-operand-type
    L_re = K.relu(K.sum(out2_prob * K.one_hot(K.cast(targets, 'int64'), num_classes), axis = -1) - alpha)
    
    return beta * L_o1_y + (1-beta) * L_o1_emb + L_o2_y + L_emb_o2 + L_re 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def _softmax(x, axis=-1, alpha=1):
    """
    building on keras implementation, allow alpha parameter

    Softmax activation function.
    # Arguments
        x : Tensor.
        axis: Integer, axis along which the softmax normalization is applied.
        alpha: a value to multiply all x
    # Returns
        Tensor, output of softmax transformation.
    # Raises
        ValueError: In case `dim(x) == 1`.
    """
    x = alpha * x
    ndim = K.ndim(x)
    if ndim == 2:
        return K.softmax(x)
    elif ndim > 2:
        e = K.exp(x - K.max(x, axis=axis, keepdims=True))
        s = K.sum(e, axis=axis, keepdims=True)
        return e / s
    else:
        raise ValueError('Cannot apply softmax to a tensor that is 1D') 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def mlp_v2():
    model = Sequential()
    model.add(Dense(2048, input_shape=(21099,)))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(BatchNormalization())
#     model.add(Dense(1024))
#     model.add(Activation('relu'))
#     model.add(Dropout(0.5))   
#     model.add(BatchNormalization())
    model.add(Dense(512))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))   
    model.add(BatchNormalization())
    model.add(Dense(128))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))    
    model.add(BatchNormalization())
    model.add(Dense(6))
    model.add(Activation('softmax'))

    model.compile(loss='categorical_crossentropy',
                  optimizer='nadam',
                  metrics=['accuracy'])
    return model 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def call(self, u_vecs):
        if self.share_weights:
            u_hat_vecs = K.conv1d(u_vecs, self.W)
        else:
            u_hat_vecs = K.local_conv1d(u_vecs, self.W, [1], [1])

        batch_size = K.shape(u_vecs)[0]
        input_num_capsule = K.shape(u_vecs)[1]
        u_hat_vecs = K.reshape(u_hat_vecs, (batch_size, input_num_capsule,
                                            self.num_capsule, self.dim_capsule))
        u_hat_vecs = K.permute_dimensions(u_hat_vecs, (0, 2, 1, 3))

        b = K.zeros_like(u_hat_vecs[:, :, :, 0])  # shape = [None, num_capsule, input_num_capsule]
        for i in range(self.routings):
            b = K.permute_dimensions(b, (0, 2, 1))  # shape = [None, input_num_capsule, num_capsule]
            c = K.softmax(b)
            c = K.permute_dimensions(c, (0, 2, 1))
            b = K.permute_dimensions(b, (0, 2, 1))
            outputs = self.activation(K.batch_dot(c, u_hat_vecs, [2, 2]))
            if i < self.routings - 1:
                b = K.batch_dot(outputs, u_hat_vecs, [2, 3])

        return outputs 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def CapsuleNet(n_capsule = 10, n_routings = 5, capsule_dim = 16,
     n_recurrent=100, dropout_rate=0.2, l2_penalty=0.0001):
    K.clear_session()

    inputs = Input(shape=(170,))
    x = Embedding(21099, 300,  trainable=True)(inputs)        
    x = SpatialDropout1D(dropout_rate)(x)
    x = Bidirectional(
        CuDNNGRU(n_recurrent, return_sequences=True,
                 kernel_regularizer=l2(l2_penalty),
                 recurrent_regularizer=l2(l2_penalty)))(x)
    x = PReLU()(x)
    x = Capsule(
        num_capsule=n_capsule, dim_capsule=capsule_dim,
        routings=n_routings, share_weights=True)(x)
    x = Flatten(name = 'concatenate')(x)
    x = Dropout(dropout_rate)(x)
#     fc = Dense(128, activation='sigmoid')(x)
    outputs = Dense(6, activation='softmax')(x)
    model = Model(inputs=inputs, outputs=outputs)
    model.compile(loss='categorical_crossentropy', optimizer='nadam', metrics=['accuracy'])
    return model 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def call(self, u_vecs):
        if self.share_weights:
            u_hat_vecs = K.conv1d(u_vecs, self.W)
        else:
            u_hat_vecs = K.local_conv1d(u_vecs, self.W, [1], [1])

        batch_size = K.shape(u_vecs)[0]
        input_num_capsule = K.shape(u_vecs)[1]
        u_hat_vecs = K.reshape(u_hat_vecs, (batch_size, input_num_capsule,
                                            self.num_capsule, self.dim_capsule))
        u_hat_vecs = K.permute_dimensions(u_hat_vecs, (0, 2, 1, 3))
        # final u_hat_vecs.shape = [None, num_capsule, input_num_capsule, dim_capsule]

        b = K.zeros_like(u_hat_vecs[:, :, :, 0])  # shape = [None, num_capsule, input_num_capsule]
        outputs = None
        for i in range(self.routings):
            b = K.permute_dimensions(b, (0, 2, 1))  # shape = [None, input_num_capsule, num_capsule]
            c = K.softmax(b)
            c = K.permute_dimensions(c, (0, 2, 1))
            b = K.permute_dimensions(b, (0, 2, 1))
            outputs = self.activation(K.batch_dot(c, u_hat_vecs, [2, 2]))
            if i < self.routings - 1:
                b = K.batch_dot(outputs, u_hat_vecs, [2, 3])

        return outputs 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def mask_attention_if_needed(self, dot_product):
        """
        Makes sure that (when enabled) each position
        (of a decoder's self-attention) cannot attend to subsequent positions.
        This is achieved by assigning -inf (or some large negative number)
        to all invalid connections. Later softmax will turn them into zeros.
        We need this to guarantee that decoder's predictions are based
        on what has happened before the position, not after.
        The method does nothing if masking is turned off.
        :param dot_product: scaled dot-product of Q and K after reshaping them
        to 3D tensors (batch * num_heads, rows, cols)
        """
        if not self.use_masking:
            return dot_product
        last_dims = K.int_shape(dot_product)[-2:]
        low_triangle_ones = (
            np.tril(np.ones(last_dims))
            # to ensure proper broadcasting
            .reshape((1,) + last_dims))
        inverse_low_triangle = 1 - low_triangle_ones
        close_to_negative_inf = -1e9
        result = (
            K.constant(low_triangle_ones, dtype=K.floatx()) * dot_product +
            K.constant(close_to_negative_inf * inverse_low_triangle))
        return result 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def softmax(x, axis=1):
    """Softmax activation function.
    # Arguments
        x : Tensor.
        axis: Integer, axis along which the softmax normalization is applied.
    # Returns
        Tensor, output of softmax transformation.
    # Raises
        ValueError: In case `dim(x) == 1`.
    """
    ndim = K.ndim(x)
    if ndim == 2:
        return K.softmax(x)
    elif ndim > 2:
        e = K.exp(x - K.max(x, axis=axis, keepdims=True))
        s = K.sum(e, axis=axis, keepdims=True)
        return e / s
    else:
        raise ValueError('Cannot apply softmax to a tensor that is 1D') 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def _get_weight_vector(self, M, w_tm1, k, beta, g, s, gamma):
#        M = tf.Print(M, [M, w_tm1, k], message='get weights beg1: ')
#        M = tf.Print(M, [beta, g, s, gamma], message='get weights beg2: ')
        # Content adressing, see Chapter 3.3.1:
        num = beta * _cosine_distance(M, k)
        w_c  = K.softmax(num) # It turns out that equation (5) is just softmax.
        # Location adressing, see Chapter 3.3.2:
        # Equation 7:
        w_g = (g * w_c) + (1-g)*w_tm1
        # C_s is the circular convolution
        #C_w = K.sum((self.C[None, :, :, :] * w_g[:, None, None, :]),axis=3)
        # Equation 8:
        # TODO: Explain
        C_s = K.sum(K.repeat_elements(self.C[None, :, :, :], self.batch_size, axis=0) * s[:,:,None,None], axis=1)
        w_tilda = K.batch_dot(C_s, w_g)
        # Equation 9:
        w_out = _renorm(w_tilda ** gamma)

        return w_out 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def softmax_activation(self, mem):
        """Softmax activation."""

        # spiking_samples = k.less_equal(k.random_uniform([self.config.getint(
        #     'simulation', 'batch_size'), 1]), 300 * self.dt / 1000.)
        # spiking_neurons = k.T.repeat(spiking_samples, 10, axis=1)
        # activ = k.T.nnet.softmax(mem)
        # max_activ = k.max(activ, axis=1, keepdims=True)
        # output_spikes = k.equal(activ, max_activ).astype(k.floatx())
        # output_spikes = k.T.set_subtensor(output_spikes[k.equal(
        #     spiking_neurons, 0).nonzero()], 0.)
        # new_and_reset_mem = k.T.set_subtensor(mem[spiking_neurons.nonzero()],
        #                                       0.)
        # self.add_update([(self.mem, new_and_reset_mem)])
        # return output_spikes

        return k.T.mul(k.less_equal(k.random_uniform(mem.shape),
                                    k.softmax(mem)), self.v_thresh) 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def set_reset_mem(self, mem, spikes):
        """
        Reset membrane potential ``mem`` array where ``spikes`` array is
        nonzero.
        """

        spike_idxs = k.T.nonzero(spikes)
        if (hasattr(self, 'activation_str') and
                self.activation_str == 'softmax'):
            new = mem.copy()  # k.T.set_subtensor(mem[spike_idxs], 0.)
        elif self.config.get('cell', 'reset') == 'Reset by subtraction':
            if self.payloads:  # Experimental.
                new = k.T.set_subtensor(mem[spike_idxs], 0.)
            else:
                pos_spike_idxs = k.T.nonzero(k.greater(spikes, 0))
                neg_spike_idxs = k.T.nonzero(k.less(spikes, 0))
                new = k.T.inc_subtensor(mem[pos_spike_idxs], -self.v_thresh)
                new = k.T.inc_subtensor(new[neg_spike_idxs], self.v_thresh)
        elif self.config.get('cell', 'reset') == 'Reset by modulo':
            new = k.T.set_subtensor(mem[spike_idxs],
                                    mem[spike_idxs] % self.v_thresh)
        else:  # self.config.get('cell', 'reset') == 'Reset to zero':
            new = k.T.set_subtensor(mem[spike_idxs], 0.)
        self.add_update([(self.mem, new)]) 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def call(self, u_vecs):
        if self.share_weights:
            u_hat_vecs = K.conv1d(u_vecs, self.W)
        else:
            u_hat_vecs = K.local_conv1d(u_vecs, self.W, [1], [1])

        batch_size = K.shape(u_vecs)[0]
        input_num_capsule = K.shape(u_vecs)[1]
        u_hat_vecs = K.reshape(u_hat_vecs, (batch_size, input_num_capsule,
                                            self.num_capsule, self.dim_capsule))    # noqa
        u_hat_vecs = K.permute_dimensions(u_hat_vecs, (0, 2, 1, 3))
        # final u_hat_vecs.shape = [None, num_capsule, input_num_capsule, dim_capsule]  # noqa

        b = K.zeros_like(u_hat_vecs[:, :, :, 0])  # shape = [None, num_capsule, input_num_capsule]  # noqa
        for i in range(self.routings):
            b = K.permute_dimensions(b, (0, 2, 1))  # shape = [None, input_num_capsule, num_capsule]    # noqa
            c = K.softmax(b)
            c = K.permute_dimensions(c, (0, 2, 1))
            b = K.permute_dimensions(b, (0, 2, 1))
            outputs = self.activation(tf.keras.backend.batch_dot(c, u_hat_vecs, [2, 2]))    # noqa
            if i < self.routings - 1:
                b = tf.keras.backend.batch_dot(outputs, u_hat_vecs, [2, 3])
        return outputs 

# 需要导入模块: from keras import backend [as 别名]
# 或者: from keras.backend import softmax [as 别名]
def call(self, x, mask=None):
        uit = K.tanh(K.dot(x, self.Ws1))
        ait = K.dot(uit, self.Ws2)
        ait = K.permute_dimensions(ait, (0, 2, 1))
        A = K.softmax(ait, axis=1)
        M = K.batch_dot(A, x)
        if self.punish:
            A_T = K.permute_dimensions(A, (0, 2, 1))
            tile_eye = K.tile(K.eye(self.weight_ws2), [self.batch_size, 1])
            tile_eye = K.reshape(
                tile_eye, shape=[-1, self.weight_ws2, self.weight_ws2])
            AA_T = K.batch_dot(A, A_T) - tile_eye
            P = K.l2_normalize(AA_T, axis=(1, 2))
            return M, P
        else:
            return M