ActivationFunctions.py 文件源码

def applyActivationFunction_PReLU( inputData, PreluActivations ) :
    """Parametric Rectified Linear Unit.
    It follows:
    `f(x) = alpha * x for x < 0`,
    `f(x) = x for x >= 0`,
    where `alpha` is a learned array with the same shape as x.

    - The input is a tensor of shape (batchSize, FeatMaps, xDim, yDim, zDim) """
    preluActivationsAsRow = PreluActivations.dimshuffle('x', 0, 'x', 'x', 'x')

    pos = T.maximum(0, inputData)
    neg = preluActivationsAsRow * (inputData - abs(inputData)) * 0.5
    output = pos + neg

    return (output)

# --- version 2 ---