Python keras.backend.square() Examples

def generate_pattern(layer_name, filter_index, size=150):
    # 过滤器可视化函数
    layer_output = model.get_layer(layer_name).output
    loss = K.mean(layer_output[:, :, :, filter_index])
    grads = K.gradients(loss, model.input)[0]
    grads /= (K.sqrt(K.mean(K.square(grads))) + 1e-5)
    iterate = K.function([model.input], [loss, grads])
    input_img_data = np.random.random((1, size, size, 3)) * 20 + 128.
    
    step = 1
    for _ in range(40):
        loss_value, grads_value = iterate([input_img_data])
        input_img_data += grads_value * step
    
    img = input_img_data[0]
    return deprocess_image(img) 

def smoothing(im, mode = None):
    # utility function to smooth an image
    if mode is None:
        return im
    elif mode == 'L2':
        # L2 norm
        return im / (np.sqrt(np.mean(np.square(im))) + K.epsilon())
    elif mode == 'GaussianBlur':
        # Gaussian Blurring with width of 3
        return filters.gaussian_filter(im,1/8)
    elif mode == 'Decay':
        # Decay regularization
        decay = 0.98
        return decay * im
    elif mode == 'Clip_weak':
        # Clip weak pixel regularization
        percentile = 1
        threshold = np.percentile(np.abs(im),percentile)
        im[np.where(np.abs(im) < threshold)] = 0
        return im
    else:
        # print error message
        print('Unknown smoothing parameter. No smoothing implemented.')
        return im 

def optimizer(self):
        a = K.placeholder(shape=(None,), dtype='int32')
        y = K.placeholder(shape=(None,), dtype='float32')

        py_x = self.model.output

        a_one_hot = K.one_hot(a, self.action_size)
        q_value = K.sum(py_x * a_one_hot, axis=1)
        error = K.abs(y - q_value)

        quadratic_part = K.clip(error, 0.0, 1.0)
        linear_part = error - quadratic_part
        loss = K.mean(0.5 * K.square(quadratic_part) + linear_part)

        optimizer = RMSprop(lr=0.00025, epsilon=0.01)
        updates = optimizer.get_updates(self.model.trainable_weights, [], loss)
        train = K.function([self.model.input, a, y], [loss], updates=updates)

        return train

    # approximate Q function using Convolution Neural Network
    # state is input and Q Value of each action is output of network 

def optimizer(self):
        a = K.placeholder(shape=(None, ), dtype='int32')
        y = K.placeholder(shape=(None, ), dtype='float32')

        py_x = self.model.output

        a_one_hot = K.one_hot(a, self.action_size)
        q_value = K.sum(py_x * a_one_hot, axis=1)
        error = K.abs(y - q_value)

        quadratic_part = K.clip(error, 0.0, 1.0)
        linear_part = error - quadratic_part
        loss = K.mean(0.5 * K.square(quadratic_part) + linear_part)

        optimizer = RMSprop(lr=0.00025, epsilon=0.01)
        updates = optimizer.get_updates(self.model.trainable_weights, [], loss)
        train = K.function([self.model.input, a, y], [loss], updates=updates)

        return train

    # approximate Q function using Convolution Neural Network
    # state is input and Q Value of each action is output of network
    # dueling network's Q Value is sum of advantages and state value 

def audio_discriminate_loss2(gamma=0.1,beta = 2*0.1,num_speaker=2):
    def loss_func(S_true,S_pred,gamma=gamma,beta=beta,num_speaker=num_speaker):
        sum_mtr = K.zeros_like(S_true[:,:,:,:,0])
        for i in range(num_speaker):
            sum_mtr += K.square(S_true[:,:,:,:,i]-S_pred[:,:,:,:,i])
            for j in range(num_speaker):
                if i != j:
                    sum_mtr -= gamma*(K.square(S_true[:,:,:,:,i]-S_pred[:,:,:,:,j]))

        for i in range(num_speaker):
            for j in range(i+1,num_speaker):
                #sum_mtr -= beta*K.square(S_pred[:,:,:,i]-S_pred[:,:,:,j])
                #sum_mtr += beta*K.square(S_true[:,:,:,:,i]-S_true[:,:,:,:,j])
                pass
        #sum = K.sum(K.maximum(K.flatten(sum_mtr),0))

        loss = K.mean(K.flatten(sum_mtr))

        return loss
    return loss_func 

def call(self, x, mask=None):
        if K.image_dim_ordering == "th":
            _, f, r, c = self.shape
        else:
            _, r, c, f = self.shape
        squared = K.square(x)
        pooled = K.pool2d(squared, (self.n, self.n), strides=(1, 1),
            padding="same", pool_mode="avg")
        if K.image_dim_ordering == "th":
            summed = K.sum(pooled, axis=1, keepdims=True)
            averaged = self.alpha * K.repeat_elements(summed, f, axis=1)
        else:
            summed = K.sum(pooled, axis=3, keepdims=True)
            averaged = self.alpha * K.repeat_elements(summed, f, axis=3)
        denom = K.pow(self.k + averaged, self.beta)
        return x / denom 

def get_weightnorm_params_and_grads(p, g):
    ps = K.get_variable_shape(p)

    # construct weight scaler: V_scaler = g/||V||
    V_scaler_shape = (ps[-1],)  # assumes we're using tensorflow!
    V_scaler = K.ones(V_scaler_shape)  # init to ones, so effective parameters don't change

    # get V parameters = ||V||/g * W
    norm_axes = [i for i in range(len(ps) - 1)]
    V = p / tf.reshape(V_scaler, [1] * len(norm_axes) + [-1])

    # split V_scaler into ||V|| and g parameters
    V_norm = tf.sqrt(tf.reduce_sum(tf.square(V), norm_axes))
    g_param = V_scaler * V_norm

    # get grad in V,g parameters
    grad_g = tf.reduce_sum(g * V, norm_axes) / V_norm
    grad_V = tf.reshape(V_scaler, [1] * len(norm_axes) + [-1]) * \
             (g - tf.reshape(grad_g / V_norm, [1] * len(norm_axes) + [-1]) * V)

    return V, V_norm, V_scaler, g_param, grad_g, grad_V 

def crosschannelnormalization(alpha = 1e-4, k=2, beta=0.75, n=5,**kwargs):
    """
    This is the function used for cross channel normalization in the original
    Alexnet
    """
    def f(X):
        b, ch, r, c = X.shape
        half = n // 2
        square = K.square(X)
        extra_channels = K.spatial_2d_padding(K.permute_dimensions(square, (0,2,3,1))
                                              , (0,half))
        extra_channels = K.permute_dimensions(extra_channels, (0,3,1,2))
        scale = k
        for i in range(n):
            scale += alpha * extra_channels[:,i:i+ch,:,:]
        scale = scale ** beta
        return X / scale

    return Lambda(f, output_shape=lambda input_shape:input_shape,**kwargs) 

def gradient_penalty_loss(self, y_true, y_pred, averaged_samples):
        """
        Computes gradient penalty based on prediction and weighted real / fake samples
        """
        gradients = K.gradients(y_pred, averaged_samples)[0]
        # compute the euclidean norm by squaring ...
        gradients_sqr = K.square(gradients)
        #   ... summing over the rows ...
        gradients_sqr_sum = K.sum(gradients_sqr,
                                  axis=np.arange(1, len(gradients_sqr.shape)))
        #   ... and sqrt
        gradient_l2_norm = K.sqrt(gradients_sqr_sum)
        # compute lambda * (1 - ||grad||)^2 still for each single sample
        gradient_penalty = K.square(1 - gradient_l2_norm)
        # return the mean as loss over all the batch samples
        return K.mean(gradient_penalty) 

def crosschannelnormalization(alpha=1e-4, k=2, beta=0.75, n=5, **kwargs):
    """
    This is the function used for cross channel normalization in the original
    Alexnet
    """

    def f(X):
        b, ch, r, c = X.shape
        half = n // 2
        square = K.square(X)
        extra_channels = K.spatial_2d_padding(K.permute_dimensions(square, (0, 2, 3, 1))
                                              , (0, half))
        extra_channels = K.permute_dimensions(extra_channels, (0, 3, 1, 2))
        scale = k
        for i in range(n):
            scale += alpha * extra_channels[:, i:i + ch, :, :]
        scale = scale ** beta
        return X / scale

    return Lambda(f, output_shape=lambda input_shape: input_shape, **kwargs) 

def crosschannelnormalization(alpha = 1e-4, k=2, beta=0.75, n=5,**kwargs):
    """
    This is the function used for cross channel normalization in the original
    Alexnet
    """
    def f(X):
        b, ch, r, c = X.shape
        half = n // 2
        square = K.square(X)
        extra_channels = K.spatial_2d_padding(K.permute_dimensions(square, (0,2,3,1))
                                              , (0,half))
        extra_channels = K.permute_dimensions(extra_channels, (0,3,1,2))
        scale = k
        for i in range(n):
            scale += alpha * extra_channels[:,i:i+ch,:,:]
        scale = scale ** beta
        return X / scale

    return Lambda(f, output_shape=lambda input_shape:input_shape,**kwargs) 

def __call__(self, loss):
        #if self.layer is None:
        #    raise Exception('Need to call `set_layer` on '
        #                    'ActivityRegularizer instance '
        #                    'before calling the instance.')
        regularized_loss = loss
        for i in range(len(self.layer.inbound_nodes)):
            output = self.layer.get_output_at(i)
            if self.l1:
                regularized_loss += K.sum(self.l1 * K.abs(output[:,:,:,1]))
            if self.l2:
                regularized_loss += K.sum(self.l2 * K.square(output[:,:,:,1]))
        return K.in_train_phase(regularized_loss, loss) 

def call(self, inputs, **kwargs):
        return K.sqrt(K.sum(K.square(inputs), -1)) 

def _to_normal2d(output_batch) -> ds.MultivariateNormalTriL:
    """
    :param output_batch: (n_samples, 5)
    :return
    """

    # mean of x and y
    x_mean = Lambda(lambda o: o[:, 0])(output_batch)
    y_mean = Lambda(lambda o: o[:, 1])(output_batch)

    # std of x and y
    # std is must be 0 or positive
    x_std = Lambda(lambda o: K.exp(o[:, 2]))(output_batch)
    y_std = Lambda(lambda o: K.exp(o[:, 3]))(output_batch)

    # correlation coefficient
    # correlation coefficient range is [-1, 1]
    cor = Lambda(lambda o: K.tanh(o[:, 4]))(output_batch)

    loc = Concatenate()([
        Lambda(lambda x_mean: K.expand_dims(x_mean, 1))(x_mean),
        Lambda(lambda y_mean: K.expand_dims(y_mean, 1))(y_mean)
    ])

    x_var = Lambda(lambda x_std: K.square(x_std))(x_std)
    y_var = Lambda(lambda y_std: K.square(y_std))(y_std)
    xy_cor = Multiply()([x_std, y_std, cor])

    cov = Lambda(lambda inputs: K.stack(inputs, axis=0))(
        [x_var, xy_cor, xy_cor, y_var])
    cov = Lambda(lambda cov: K.permute_dimensions(cov, (1, 0)))(cov)
    cov = Reshape((2, 2))(cov)

    scale_tril = Lambda(lambda cov: tf.cholesky(cov))(cov)
    mvn = ds.MultivariateNormalTriL(loc, scale_tril)

    return mvn 

def gradient_penalty_loss(self, y_true, y_pred, averaged_samples):
        """
        Computes gradient penalty based on prediction and weighted real / fake samples
        """
        gradients = K.gradients(y_pred, averaged_samples)[0]
        # compute the euclidean norm by squaring ...
        gradients_sqr = K.square(gradients)
        #   ... summing over the rows ...
        gradients_sqr_sum = K.sum(gradients_sqr, axis=np.arange(1, len(gradients_sqr.shape)))
        #   ... and sqrt
        gradient_l2_norm = K.sqrt(gradients_sqr_sum)
        # compute lambda * (1 - ||grad||)^2 still for each single sample
        gradient_penalty = K.square(1 - gradient_l2_norm)
        # return the mean as loss over all the batch samples
        return K.mean(gradient_penalty) 

def get_updates(self, loss, params):
        grads = self.get_gradients(loss, params)
        self.updates = [(self.iterations, self.iterations + 1)]

        t = self.iterations + 1
        lr_t = self.lr * K.sqrt(1. - K.pow(self.beta_2, t)) / (1. - K.pow(self.beta_1, t))

        ms = [K.variable(np.zeros(K.get_value(p).shape)) for p in params]
        vs = [K.variable(np.zeros(K.get_value(p).shape)) for p in params]
        gs = [K.variable(np.zeros(K.get_value(p).shape)) for p in params]
        self.weights = ms + vs

        for p, g, m, v, gg in zip(params, grads, ms, vs, gs):

            flag = K.equal(self.iterations % self.accum_iters, 0)
            flag = K.cast(flag, dtype='float32')

            gg_t = (1 - flag) * (gg + g)
            m_t = (self.beta_1 * m) + (1. - self.beta_1) * (gg + flag * g) / self.accum_iters
            v_t = (self.beta_2 * v) + (1. - self.beta_2) * K.square((gg + flag * g) / self.accum_iters)
            p_t = p - flag * lr_t * m_t / (K.sqrt(v_t) + self.epsilon)

            self.updates.append((m, flag * m_t + (1 - flag) * m))
            self.updates.append((v, flag * v_t + (1 - flag) * v))
            self.updates.append((gg, gg_t))

            new_p = p_t
            # apply constraints
            if getattr(p, 'constraint', None) is not None:
                c = p.constraints(new_p)
                new_p = c(new_p)
            self.updates.append((p, new_p))
        return self.updates 

def tv_loss(x):
    '''
    Total variation loss is used to keep the image locally coherent
    '''
    assert K.ndim(x) == 4
    a = K.square(x[:, :-1, :-1, :] - x[:, 1:, :-1, :])
    b = K.square(x[:, :-1, :-1, :] - x[:, :-1, 1:, :])
    return K.sum(a + b, axis=(1, 2, 3)) 

def style_loss(x, target, norm_by_channels=False):
    '''
    Style loss is the MSE between Gram matrices computed using activation maps.
    '''
    x_gram = gram_matrix(x, norm_by_channels=norm_by_channels)
    return K.mean(K.square(target - x_gram), axis=(1, 2)) 

def gradient_penalty_loss(y_true, y_pred, averaged_samples, gradient_penalty_weight):
    """
    This term is used for stabilizing the WGAN training.
    """
    gradients = K.gradients(y_pred, averaged_samples)[0]
    gradients_sqr = K.square(gradients)
    axes_for_sum = tuple(np.arange(1, len(gradients_sqr.shape)))
    gradients_sqr_sum = K.sum(gradients_sqr, axis=axes_for_sum)
    gradient_norm = K.sqrt(gradients_sqr_sum)
    gradient_penalty = gradient_penalty_weight * K.square(1 - gradient_norm)
    return K.mean(gradient_penalty) 

def vae_loss(self, x, x_decoded_mean):
        xent_loss = original_dim * metrics.binary_crossentropy(x, x_decoded_mean)
        kl_loss = - 0.5 * K.sum(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
        return K.mean(xent_loss + kl_loss) 

def vae_loss(self, x, x_decoded_mean_squash):
        x = K.flatten(x)
        x_decoded_mean_squash = K.flatten(x_decoded_mean_squash)
        xent_loss = img_rows * img_cols * metrics.binary_crossentropy(x, x_decoded_mean_squash)
        kl_loss = - 0.5 * K.mean(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
        return K.mean(xent_loss + kl_loss) 

def vae_loss(self, x, x_decoded_mean):
        xent_loss = original_dim * metrics.binary_crossentropy(x, x_decoded_mean)
        kl_loss = - 0.5 * K.sum(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
        return K.mean(xent_loss + kl_loss) 

def vae_loss(self, x, x_decoded_mean_squash):
        x = K.flatten(x)
        x_decoded_mean_squash = K.flatten(x_decoded_mean_squash)
        xent_loss = img_rows * img_cols * metrics.binary_crossentropy(x, x_decoded_mean_squash)
        kl_loss = - 0.5 * K.mean(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
        return K.mean(xent_loss + kl_loss) 

def vae_loss(self, x, x_decoded_mean):
        xent_loss = original_dim * metrics.binary_crossentropy(x, x_decoded_mean)
        kl_loss = - 0.5 * K.sum(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
        return K.mean(xent_loss + kl_loss) 

def vae_loss(self, x, x_decoded_mean_squash):
        x = K.flatten(x)
        x_decoded_mean_squash = K.flatten(x_decoded_mean_squash)
        xent_loss = img_rows * img_cols * metrics.binary_crossentropy(x, x_decoded_mean_squash)
        kl_loss = - 0.5 * K.mean(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
        return K.mean(xent_loss + kl_loss) 

def __call__(self, loss):
        #if self.layer is None:
        #    raise Exception('Need to call `set_layer` on '
        #                    'ActivityRegularizer instance '
        #                    'before calling the instance.')
        regularized_loss = loss
        for i in range(len(self.layer.inbound_nodes)):
            output = self.layer.get_output_at(i)
            if self.l1:
                regularized_loss += K.sum(self.l1 * K.abs(output[:,:,:,1]))
            if self.l2:
                regularized_loss += K.sum(self.l2 * K.square(output[:,:,:,1]))
        return K.in_train_phase(regularized_loss, loss) 

def build_backprop(model, loss):
    # Gradient of the input image with respect to the loss function
    gradients = K.gradients(loss, model.input)[0]
    # Normalize the gradients
    gradients /= (K.sqrt(K.mean(K.square(gradients))) + 1e-5)
    # Keras function to calculate the gradients and loss
    return K.function([model.input], [loss, gradients])


# Loss function that optimizes one class 

def build_backprop(model, loss):
    # Gradient of the input image with respect to the loss function
    gradients = K.gradients(loss, model.input)[0]
    # Normalize the gradients
    gradients /= (K.sqrt(K.mean(K.square(gradients))) + 1e-5)
    # Keras function to calculate the gradients and loss
    return K.function([model.input], [loss, gradients]) 

def build_backprop(model, loss):
    # Gradient of the input image with respect to the loss function
    gradients = K.gradients(loss, model.input)[0]
    # Normalize the gradients
    gradients /= (K.sqrt(K.mean(K.square(gradients))) + 1e-5)
    # Keras function to calculate the gradients and loss
    return K.function([model.input], [loss, gradients])


# Loss function that optimizes one class 

def build_backprop(model, loss):
    # Gradient of the input image with respect to the loss function
    gradients = K.gradients(loss, model.input)[0]
    # Normalize the gradients
    gradients /= (K.sqrt(K.mean(K.square(gradients))) + 1e-5)
    # Keras function to calculate the gradients and loss
    return K.function([model.input], [loss, gradients])