utils.py

import math
import tensorflow as tf
import numpy as np

def asinh(x, scale=5.):
    """Asinh transform."""
    f = np.vectorize(lambda y: math.asinh(y / scale))
    return f(x)

def sinh(x, scale=5.):
    """Reverse transform for asinh."""
    return scale * np.sinh(x)

def lrelu(x, leak=0.2, name="lrelu"):
    """Leaky ReLU activation."""
    return tf.maximum(x, leak * x)

def tbn(name):
    """Get the tensor in the default graph of the given name."""
    return tf.get_default_graph().get_tensor_by_name(name)

def obn(name):
    """Get the operation node in the default graph of the given name."""
    return tf.get_default_graph().get_operation_by_name(name)

def calculate_mmd(k1, k2, k12):
    """ Calculates MMD given kernels for batch1, batch2, and between batches """
    return k1.sum()/(k1.shape[0]*k1.shape[1]) + k2.sum()/(k2.shape[0]*k2.shape[1]) - 2*k12.sum()/(k12.shape[0]*k12.shape[1])