Hello Skand,

> Yes sure, I do agree that we need to define our base classes well so that they
> can be extended to the ssRBM. I am on it, it would take me a day or so to
> properly frame the api.

Sounds good, let us know if you have anything we can use as a basis for a
discussion.

> I had another question to ask, can we include even Deep Boltzmann Machines in
> this module itself apart from RBM and ssRBM? Here is the link for the paper :
> http://www.jmlr.org/proceedings/papers/v5/salakhutdinov09a/salakhutdinov09a.pdf

Sounds like a good idea, would be an insteresting and easy extension.

Thanks,
Marcus

> On 17 Mar 2017, at 12:16, Skand Vishwanath Peri <[email protected]> wrote:
> 
> Hey Marcus, 
>  
> However, we should
> think about the necessaries for additional features, so that we can add 
> features
> without redesigning the classes.
> 
> Yes sure, I do agree that we need to define our base classes well so that 
> they can be extended to the ssRBM. I am on it, it would take me a day or so 
> to properly frame the api. 
> 
> I had another question to ask, can we include even Deep Boltzmann Machines in 
> this module itself apart from RBM and ssRBM?
> Here is the link for the paper : 
> http://www.jmlr.org/proceedings/papers/v5/salakhutdinov09a/salakhutdinov09a.pdf
>  
> <http://www.jmlr.org/proceedings/papers/v5/salakhutdinov09a/salakhutdinov09a.pdf>
> 
> 
> Regards
> Skand Vishwanath Peri
> 
> 
> ᐧ
> 
> On Wed, Mar 15, 2017 at 8:14 PM, Marcus Edel <[email protected] 
> <mailto:[email protected]>> wrote:
> Hello Skand,
> 
> (cc mlpack list, please use the public mailing list)
> 
>> Yes I couldn't find an mlpack RBM implementation. I guess (if I am not 
>> wrong),
>> we have to start from scratch. In the previous few days I had gone through 
>> the
>> paper on spike and slab RBM (and also revised rbm). So can you guide me how 
>> to
>> start the method rbm? I would first like to implement the RBM class and then
>> extend to various modifications of it. What do you think?
> 
> Starting with the RBM and adding the improvements described in the "Spike and
> slab restricted boltzmann machine" paper sounds reasonable. However, we should
> think about the necessaries for additional features, so that we can add 
> features
> without redesigning the classes.
> 
> So I propose that we first brainstorm about the API, before implementing the
> model, does that sound reasonable?
> 
> Thanks,
> Marcus
> 
>> On 14 Mar 2017, at 10:12, Skand Vishwanath Peri <[email protected] 
>> <mailto:[email protected]>> wrote:
>> 
>> Hey Marcus, 
>>     
>> 
>> Do you mean you couldn't find an mlpack RBM implementation?
>> 
>> Yes I couldn't find an mlpack RBM implementation. I guess (if I am not 
>> wrong), we have to start from scratch. In the previous few days I had gone 
>> through the paper on spike and slab RBM (and also revised rbm). So can you 
>> guide me how to start the method rbm? I would first like to implement the 
>> RBM class and then extend to various modifications of it. What do you think?
>> 
>> Regards
>> Skand Vishwanath Peri
>> ᐧ
>> 
>> On Mon, Mar 6, 2017 at 8:23 PM, Marcus Edel <[email protected] 
>> <mailto:[email protected]>> wrote:
>> Hello Skand,
>> 
>> thanks for getting in touch.
>> 
>>> Among the topics given on the wiki page, I am interested in implemening 
>>> spike
>>> and slab RBM. I have done a course in Advance Machine Learning and am 
>>> familiar
>>> with the basics of Boltzmann Machine and RBM.  I wanted to ask you from 
>>> where
>>> should I start.
>> 
>> My recommendation is to start with the references given in the project
>> description, each paper also has some useful references that are really 
>> helpful.
>> In fact, to be successful at this project, you should have a good knowledge 
>> of
>> deep learning; i.e., you should be familiar with the way deep neural networks
>> are typically built and trained, and certainly you should be familiar with 
>> the
>> individual components that you plan to implement.
>> 
>>> I don't seem to find any previous implementation of RBM (Correct me if I am
>>> wrong. I did search through but didn't find it.)
>> 
>> Do you mean you couldn't find an mlpack RBM implementation?
>> 
>>> I have previously worked in Convolutional Neural Networks by doing a 
>>> project on
>>> detecting Distracted Driver and was successfully able to implement it.
>> 
>> Sounds interesting based on images/video? I wonder what dataset you used to
>> train the network, would be interesting to hear more about the project.
>> 
>>> I am currently going through the slab and spike rbm. It would be of great 
>>> help
>>> if you can guide me to start.
>> 
>> A good place to start is by working through the mlpack tutorials and making 
>> sure
>> you can get mlpack to compile and understand how to use it. If everything 
>> works
>> we can discuss some ideas how to improve the project or add new features.
>> 
>> Is that helpful? If you have any questions, feel free to ask.
>> 
>> Thanks,
>> Marcus
>> 
>>> On 6 Mar 2017, at 13:23, Skand Vishwanath Peri <[email protected] 
>>> <mailto:[email protected]>> wrote:
>>> 
>>> Hello, 
>>>        I am Skand Vishwanath Peri from India. I am currently in my pre - 
>>> final year of Undergraduate studies. I am interested in participating in 
>>> GSoC 17 with mlpack.
>>> 
>>> Among the topics given on the wiki page, I am interested in implemening 
>>> spike and slab RBM. I have done a course in Advance Machine Learning and am 
>>> familiar with the basics of Boltzmann Machine and RBM. 
>>> I wanted to ask you from where should I start. I don't seem to find any 
>>> previous implementation of RBM (Correct me if I am wrong. I did search 
>>> through but didn't find it.) 
>>> 
>>> I have previously worked in Convolutional Neural Networks by doing a 
>>> project on detecting Distracted Driver 
>>> <https://github.com/pvskand/DriverDetection> and was successfully able to 
>>> implement it.
>>> 
>>> I am currently going through the slab and spike rbm. It would be of great 
>>> help if you can guide me to start.
>>> 
>>> Regards
>>> Skand Vishwanath Peri
>>> ᐧ
>>> _______________________________________________
>>> mlpack mailing list
>>> [email protected] <mailto:[email protected]>
>>> http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack 
>>> <http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack>
>> 
> 
> 

_______________________________________________
mlpack mailing list
[email protected]
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Reply via email to