Baris Kayalibay

Thank you for the kind words. One can use eq. 7 of our paper to calculate the receptive field of any layer in the network: ```python rec_field_0 = 1 rec_field_next...

It seems to me that the problem is related to bilinear upsampling. You can try the following to work around this issue: In model_defs.py, lines 421-428: ```python {'i':55, 'type': 'skip',...

I'm not sure if 4GB would be enough for that cnn with that input size. You can try running the code with fewer features or on inputs that have smaller...

Sorry, the code is based on Theano. It won't work with tensorflow.

The result of `ys = tf.stack([Exponential(1. / bw), Exponential(1. / bw)])` is not a random variable anymore but a `tf.Tensor`. KLqp only calculates the log probability of an item in...

So calling `ys_inference = ed.ReparameterizationKLqp({bw: qbandwidth}, {ys: numpy.array([100., 100.])})` works? I'm surprised by that since it should have the same problem as plain `ed.KLqp` due to [this if-statement](https://github.com/blei-lab/edward/blob/352596c3bc77e1cb5c01f77ac3283a53c74dbf21/edward/inferences/klqp.py#L683). So if...