naresh1318 / adversarial_autoencoder Goto Github PK

Hi, dear author:
I found the first diagram, dense network structure in your readme file is not generate by TF. Hence i wonder which tools you used to draw it?

When I'm trying to run the supervised_adversarial_autoencoder.py I get the following uncaught exception:

`2019-02-24 17:00:07.426921: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2019-02-24 17:00:07.586 python[6976:8326575] -[NSApplication _setup:]: unrecognized selector sent to instance 0x7fd270bef700
2019-02-24 17:00:07.587 python[6976:8326575] *** Terminating app due to uncaught exception 'NSInvalidArgumentException', reason: '-[NSApplication _setup:]: unrecognized selector sent to instance 0x7fd270bef700'
*** First throw call stack:
(
0 CoreFoundation 0x00007fff4f351ecd __exceptionPreprocess + 256
1 libobjc.A.dylib 0x00007fff7b409720 objc_exception_throw + 48
2 CoreFoundation 0x00007fff4f3cf275 -[NSObject(NSObject) __retain_OA] + 0
3 CoreFoundation 0x00007fff4f2f3b40 forwarding + 1486
4 CoreFoundation 0x00007fff4f2f34e8 _CF_forwarding_prep_0 + 120
5 libtk8.6.dylib 0x000000012e3f131d TkpInit + 413
6 libtk8.6.dylib 0x000000012e34917e Initialize + 2622
7 _tkinter.cpython-36m-darwin.so 0x000000012e171a16 _tkinter_create + 1174
8 python 0x000000010e401068 _PyCFunction_FastCallDict + 200
9 python 0x000000010e4d661f call_function + 143
10 python 0x000000010e4d4175 _PyEval_EvalFrameDefault + 46837
11 python 0x000000010e4c78c9 _PyEval_EvalCodeWithName + 425
12 python 0x000000010e4d72cc _PyFunction_FastCallDict + 364
13 python 0x000000010e37ff80 _PyObject_FastCallDict + 320
14 python 0x000000010e3a75f8 method_call + 136
15 python 0x000000010e3875ce PyObject_Call + 62
16 python 0x000000010e4285b5 slot_tp_init + 117
17 python 0x000000010e42caf1 type_call + 241
18 python 0x000000010e37fef1 _PyObject_FastCallDict + 177
19 python 0x000000010e388137 _PyObject_FastCallKeywords + 327
20 python 0x000000010e4d6718 call_function + 392
21 python 0x000000010e4d4225 _PyEval_EvalFrameDefault + 47013
22 python 0x000000010e4d69dc fast_function + 188
23 python 0x000000010e4d667c call_function + 236
24 python 0x000000010e4d4175 _PyEval_EvalFrameDefault + 46837
25 python 0x000000010e4c78c9 _PyEval_EvalCodeWithName + 425
26 python 0x000000010e4d72cc _PyFunction_FastCallDict + 364
27 python 0x000000010e37ff80 _PyObject_FastCallDict + 320
28 python 0x000000010e3a75f8 method_call + 136
29 python 0x000000010e3875ce PyObject_Call + 62
30 python 0x000000010e4d4376 _PyEval_EvalFrameDefault + 47350
31 python 0x000000010e4c78c9 _PyEval_EvalCodeWithName + 425
32 python 0x000000010e4d6a8a fast_function + 362
33 python 0x000000010e4d667c call_function + 236
34 python 0x000000010e4d4175 _PyEval_EvalFrameDefault + 46837
35 python 0x000000010e4d69dc fast_function + 188
36 python 0x000000010e4d667c call_function + 236
37 python 0x000000010e4d4175 _PyEval_EvalFrameDefault + 46837
38 python 0x000000010e4c78c9 _PyEval_EvalCodeWithName + 425
39 python 0x000000010e4d6a8a fast_function + 362
40 python 0x000000010e4d667c call_function + 236
41 python 0x000000010e4d4175 _PyEval_EvalFrameDefault + 46837
42 python 0x000000010e4c78c9 _PyEval_EvalCodeWithName + 425
43 python 0x000000010e4d6a8a fast_function + 362
44 python 0x000000010e4d667c call_function + 236
45 python 0x000000010e4d4225 _PyEval_EvalFrameDefault + 47013
46 python 0x000000010e4c78c9 _PyEval_EvalCodeWithName + 425
47 python 0x000000010e4d6a8a fast_function + 362
48 python 0x000000010e4d667c call_function + 236
49 python 0x000000010e4d4225 _PyEval_EvalFrameDefault + 47013
50 python 0x000000010e4c78c9 _PyEval_EvalCodeWithName + 425
51 python 0x000000010e52055c PyRun_FileExFlags + 252
52 python 0x000000010e51fa34 PyRun_SimpleFileExFlags + 372
53 python 0x000000010e5467c6 Py_Main + 3734
54 python 0x000000010e377f59 main + 313
55 libdyld.dylib 0x00007fff7c4d7ed9 start + 1
56 ??? 0x0000000000000002 0x0 + 2
)
libc++abi.dylib: terminating with uncaught exception of type NSException

Process finished with exit code 134 (interrupted by signal 6: SIGABRT)`

Anyone please help me with this!!

Hi Naresh,

Really appreciate you taking the time to make the AAE tutorial. It is a great read!

I have a question regarding the implementation of generator_optimizer in the code. When I print en_var, I get the following list of variables

e_dense_1/weights:0
e_dense_1/bias:0
e_dense_2/weights:0
e_dense_2/bias:0
e_latent_variable/weights:0
e_latent_variable/bias:0
d_dense_1/weights:0
d_dense_1/bias:0
d_dense_2/weights:0
d_dense_2/bias:0

In your post, you mention that:

We’ll backprop only through the encoder weights, which causes the encoder to learn the required distribution and produce output which’ll have that distribution.

Do the decoder weights get updated as well?

The returned datetime was not supported by Windows. I replaced the line with folder_name = "\\{0}_{1}_{2}_{3}_{4}_{5}_autoencoder".format(int(time.time()), z_dim, learning_rate, batch_size, n_epochs, beta1)
as well as just changing the import from datetime to time.

Hi, I'm trying to implement the adversarial autoencoder myself. I found this repo and use it as a guide. It trains the autoencoder part but I think this line is a bug? Shouldn't the categorical discriminator train like the gaussian one? If it's not a bug could you comment on the training objective? Thanks!

Thank you for your code!
Sorry, I don't know much about TF.
When I tried the unsupervised experiment, I tried to improve the dimension of latent representation(like, 10, 128), but the generator was completely broken, and I failed to try to modify it. I wonder if you have tried, could you please provide some information? Thank you very much!

Good Morning!

I am trying to reproduce the very nice results you have in your tutorial but my training do not go as well when I execute python adversarial_autoencoder.py

Here are the losses I obtain:

I tried basically to decrease the learning rate without effect.
Can you help me?

Thanks in advance,

Pierre

Hi Naresh1318,

Thank you very much for the code. I ran your implementation and was able to get the reported results in README.md.
However my question is, have you been able to reach an error rate of 1.90 (±0:10)% for MNIST 100 labels (as reported in original paper)? Even after 5000 epochs the lowest I get is somewhat around 5%.

Hi Naresh

I am using Window 10, Tensorflow 1.8, Anaconda
When I have run autoencoder.py file than I am getting the error

ValueError: Variable e_dense_1/weights already exists, disallowed. Did you mean to set reuse=True in VarScope? Originally defined at

I am waiting you response

Thanks for the great tutorial :-)

How can I save the encoder model? I know how that can be done in Keras, but not in TF. Basically I would like to save the latent representation of the input, ex: (n,2) dim latent representation of the data.