Hi, 
You can probably have a look at the implementation in Lasagne. 
https://github.com/Lasagne/Recipes/blob/master/examples/lstm_text_generation.py#L119-L156
 Instead 
of the Dense, you can use a convolution layer for your case

Cheers, 
Ramana

On Tuesday, June 20, 2017 at 6:42:39 AM UTC+5:30, Sunjeet Jena wrote:
>
> If you don't mind could you advice me how can I then possibly join a CNN 
> architecture with RNN and back propagate through time. Do I need to use 
> scan function  for the CNN also? If I have 't' frames to consider then if I 
> take the convolution outputs of all these 't' frames and then put them in a 
> RNN network using scan function will it work? 
>
> On Tuesday, 20 June 2017 03:45:46 UTC+5:30, nouiz wrote:
>>
>> Sadly, no.
>>
>> Fred
>>
>> Le lun. 19 juin 2017 06:39, Sunjeet Jena <[email protected]> a écrit :
>>
>>> Can we implement it using normal 'for' function of python? Like, saving 
>>> all the parameters in a list and then taking the gradient .
>>>
>>>
>>> On Monday, 19 June 2017 10:42:24 UTC+5:30, Jesse Livezey wrote:
>>>>
>>>> You can use the scan function to create RNN architectures.
>>>>
>>>> http://deeplearning.net/software/theano/library/scan.html
>>>>
>>>> On Sunday, June 18, 2017 at 4:13:44 PM UTC-7, Sunjeet Jena wrote:
>>>>>
>>>>> I am building a multi-layer RNN network and thus need a way to back 
>>>>> propagate through time in Theano. Does theano automatically knows how to 
>>>>> unfold the network as a feed forward network?
>>>>>
>>>> -- 
>>>
>>> --- 
>>> You received this message because you are subscribed to the Google 
>>> Groups "theano-users" group.
>>> To unsubscribe from this group and stop receiving emails from it, send 
>>> an email to [email protected].
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>

-- 

--- 
You received this message because you are subscribed to the Google Groups 
"theano-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to