Thank you Ralf and Kyle for your input! They sound really useful. I
will try them out.

Again thanks a lot for introducing two hilarious libraries!

-Shawn

On Mon, Apr 7, 2014 at 3:51 PM, Ralf Gunter <[email protected]> wrote:
> Ah, my apologies, I vaguely remembered reading about opencl somewhere
> in the theano docs...
>
> I stand corrected.
>
> 2014-04-07 14:42 GMT-05:00 Kyle Kastner <[email protected]>:
>> Also, I think pylearn2 and theano-nets only support CUDA cards (Nvidia),
>> since Teahno only supports CUDA or CPU to my knowledge. There is some
>> discussion of supporting an OpenCL backend
>> (https://groups.google.com/forum/#!msg/theano-dev/NhRUhWA6xzo/lYrIIOlw4D8J),
>> but I don't know whether that work was ever completed. Maybe someone else
>> has a better knowledge of this.
>>
>>
>> On Mon, Apr 7, 2014 at 2:39 PM, Kyle Kastner <[email protected]> wrote:
>>>
>>> You can also use the python interface to pylearn2, rather than the yaml.
>>> If you are interested in examples of the python interface for pylearn2, I
>>> have some examples (I greatly prefer the python interface, but to each their
>>> own):
>>>
>>>
>>> https://github.com/kastnerkyle/pylearn2-practice/blob/master/cifar10_train.py
>>> shows how to build a network and test using a pylearn2 builtin dataset
>>>
>>>
>>> https://github.com/kastnerkyle/kaggle-dogs-vs-cats/blob/master/kaggle_train.py
>>> This shows how to use scikit-learn's train-test split to create training
>>> and testing classes for new datasets in pylearn2 format. x and y are both
>>> 2D. Rows are samples, columns are features for x. y generally needs to be a
>>> "one hot" label matrix for classification, but will be a regression target
>>> for RMSE/regression tasks.
>>>
>>> In general, it is pretty easy to wrap your data into a pylearn2 compatible
>>> format, though doing "raw" input for convolutional nets can be tricky. I
>>> have an example of reading pngs from Kaggle's CIFAR10 competition into
>>> pylearn2 and using a convnet here:
>>> https://github.com/kastnerkyle/kaggle-cifar10
>>>
>>> All that being said, theano-nets *can* be easier to start with for those
>>> who are new to neural networks, as it is a little less roll-your-own.
>>>
>>> I have had success using both.
>>>
>>> Kyle
>>>
>>>
>>> On Mon, Apr 7, 2014 at 11:48 AM, Ralf Gunter <[email protected]> wrote:
>>>>
>>>> Two libraries[1,2] come to mind that can additionally support
>>>> accelerators through opencl. Just take note that it can take a bit to
>>>> familiarize yourself with pylearn2, especially because they seem
>>>> adamant in doing everything through yaml scripts.
>>>>
>>>> [1] -- https://github.com/lmjohns3/theano-nets (see e.g.
>>>> examples/xor-classfier.py)
>>>> [2] -- http://deeplearning.net/software/pylearn2
>>>>
>>>> 2014-04-07 11:18 GMT-05:00 Yuxiang Wang <[email protected]>:
>>>> > Dear all,
>>>> >
>>>> > I am not entirely sure whether this is the best place to post this,
>>>> > and please do excuse me if this is not the perfect list for this
>>>> > question.
>>>> >
>>>> > Is there any python packages for neural networks for regression
>>>> > (instead of classification)?
>>>> >
>>>> > Any help would be appreciated. Thanks!
>>>> >
>>>> > -Shawn
>>>> >
>>>> >
>>>> > ------------------------------------------------------------------------------
>>>> > Put Bad Developers to Shame
>>>> > Dominate Development with Jenkins Continuous Integration
>>>> > Continuously Automate Build, Test & Deployment
>>>> > Start a new project now. Try Jenkins in the cloud.
>>>> > http://p.sf.net/sfu/13600_Cloudbees_APR
>>>> > _______________________________________________
>>>> > Scikit-learn-general mailing list
>>>> > [email protected]
>>>> > https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>>
>>>>
>>>> ------------------------------------------------------------------------------
>>>> Put Bad Developers to Shame
>>>> Dominate Development with Jenkins Continuous Integration
>>>> Continuously Automate Build, Test & Deployment
>>>> Start a new project now. Try Jenkins in the cloud.
>>>> http://p.sf.net/sfu/13600_Cloudbees_APR
>>>> _______________________________________________
>>>> Scikit-learn-general mailing list
>>>> [email protected]
>>>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>
>>>
>>
>>
>> ------------------------------------------------------------------------------
>> Put Bad Developers to Shame
>> Dominate Development with Jenkins Continuous Integration
>> Continuously Automate Build, Test & Deployment
>> Start a new project now. Try Jenkins in the cloud.
>> http://p.sf.net/sfu/13600_Cloudbees
>> _______________________________________________
>> Scikit-learn-general mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>
>
> ------------------------------------------------------------------------------
> Put Bad Developers to Shame
> Dominate Development with Jenkins Continuous Integration
> Continuously Automate Build, Test & Deployment
> Start a new project now. Try Jenkins in the cloud.
> http://p.sf.net/sfu/13600_Cloudbees
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general



-- 
Yuxiang "Shawn" Wang
Gerling Research Lab
University of Virginia
[email protected]
+1 (434) 284-0836
https://sites.google.com/a/virginia.edu/yw5aj/

------------------------------------------------------------------------------
Put Bad Developers to Shame
Dominate Development with Jenkins Continuous Integration
Continuously Automate Build, Test & Deployment 
Start a new project now. Try Jenkins in the cloud.
http://p.sf.net/sfu/13600_Cloudbees
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to