Do we need controller.compile?

[hoangcuong2011]

Hi Florian,

Thanks for your great code! I have a question though:
Do we really need this line of code (line 81):

controller.compile(loss='binary_crossentropy', optimizer=sgd, metrics = ['binary_accuracy'], sample_weight_mode="temporal")

from https://github.com/flomlo/ntm_keras/blob/master/main.py ?

I don't really think we need this. If we do need this, it would really confuse me. We already specify (line 43):

model.compile(loss='binary_crossentropy', optimizer=sgd, metrics = ['binary_accuracy'], sample_weight_mode="temporal")

from https://github.com/flomlo/ntm_keras/blob/master/model_ntm.py

controller is just a part of the NTM, and we should specific the ``compile" part only at NTM model. Am I correct?

I hope I made my question clear. Again, thanks a lot for your great work!

Best,

Cuong

[flomlo]

Hello and thank you for your interest! I think that back then, when I wrote the code, it was necessary as compiling the controller allowed to refer to something ... input/output of the controller, maybe? Or some internal state? I dont remember anymore, havent touched neither this code nor keras for almost a year. I am also not sure if compiling is "recursive" in this particular case. It could very well be however that this process in keras has changed. As I said, I haven't followed keras development. You may try it without this line, and if it works, I would merge the respective patch :) Otherwise: If you really need to understand that line perfectly I will have a look at it, but not right now, as my master thesis is due in 4 days :| Best, Florian

…

On 9/19/18 5:29 AM, Hoang Cuong wrote: Hi Florian, Thanks for your great code! I have a question though: Do we really need this line of code (line 81): controller.compile(loss='binary_crossentropy', optimizer=sgd, metrics = ['binary_accuracy'], sample_weight_mode="temporal") from https://github.com/flomlo/ntm_keras/blob/master/main.py ? I don't really think we need this. If we do need this, it would really confuse me. We already specify (line 43): |model.compile(loss='binary_crossentropy', optimizer=sgd, metrics = ['binary_accuracy'], sample_weight_mode="temporal") | from https://github.com/flomlo/ntm_keras/blob/master/model_ntm.py controller is just a part of the NTM, and we should specific the ``compile" part only at NTM model. Am I correct? I hope I made my question clear. Again, thanks a lot for your great work! Best, Cuong — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#7>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGaa-4CBx3mpmh1N3rznwL5qvluz1qrEks5ucbopgaJpZM4WvWNU>.

[hoangcuong2011]

Thanks Florian for your quick response!

I don't know what your master thesis is about. But given your deep understanding of this advanced topic, I know the thesis is really awesome! :-P

Good luck!

Best,

Cuong.