Could you me give more demos on simnets.

[CuriousCat-7]

I use your keras example in README.md. And get only 0.9 acc in MNIST, which is equal to random guess.

What's the best way to train the plain MNIST classifier using SimNet?

Could you give more demos like experiments in.
In Deep SimNet.

BTW, the keras example has some mistakes, fixed in there

Thank you,
Neo

[orsharir]

Hi,

We'll try to upload new examples based on our older works in the near future. Meanwhile, you can either follow the architecture we've discussed in the paper you've linked to, or examine the slightly different take on SimNets as a generative model that can be found in this repository (based on Caffe instead of TF): https://github.com/HUJI-Deep/Generative-ConvACs

Or

[CuriousCat-7]

@orsharir
Thank you for your reply.
I'm looking for your new examples. 😶