Usage of tf.variable_scope and tf.get_variable

Question

I am trying to follow a paper, which uses code as follows to define a convolutional layer:

def _conv(inputs, filters, kernel_size=1, strides=1, pad=\'VALID\', nam