[mlpack] Interested in writing documentation for Mlpack's ann module

Thu Apr 21 12:16:14 EDT 2016

Hello Abhinav,

That's really nice, I didn't thought the performance would be that good
(0.00358423 MSE) right out of the box. Would be interesting to see how the
performance changed using PCA. Also, you said you use pandas for preprocessing,
I guess to fill missing values?

Thanks again for all the work,
Marcus

> On 19 Apr 2016, at 12:10, Abhinav Gupta <abhinavgupta440 at gmail.com> wrote:
> 
> Hi  Marcus,
>    I'hv reduced the dimension of the dataset to 10 using PCA and have update the github wiki link and doc accordingly.
> However I'm not yet satisfied with the documentation of the code so I'll be looking into it. 
> 
> Please suggest if you have any changes in mind.
> 
> Thanks,
> Abhinav
> 
> On Mon, Apr 18, 2016 at 8:09 PM, Marcus Edel <marcus.edel at fu-berlin.de <mailto:marcus.edel at fu-berlin.de>> wrote:
> Hello,
> 
> sounds good, I'll take a look at the doc and the wiki page in the next days.
> 
> Also, I think it is a good idea, to reduce the dimension e.g. using PCA of the
> dataset before we train the network.
> 
> Thanks,
> Marcus
> 
>> On 18 Apr 2016, at 11:05, Abhinav Gupta <abhinavgupta440 at gmail.com <mailto:abhinavgupta440 at gmail.com>> wrote:
>> 
>> Hi Marcus,
>>    I'hv also created a wiki page on github.
>>    https://github.com/abhinvgpta/mlpack/wiki/Example-of-Feed-Forward-Network-using-mlpack <https://github.com/abhinvgpta/mlpack/wiki/Example-of-Feed-Forward-Network-using-mlpack>
>>    Please have a look.
>> 
>> Thanks,
>> Abhinav
>> 
>> On Mon, Apr 18, 2016 at 1:16 PM, Abhinav Gupta <abhinavgupta440 at gmail.com <mailto:abhinavgupta440 at gmail.com>> wrote:
>> Hi Marcus,
>>    I'hv preprocessed the Internet Advertisements dataset and implemented the example using FFN. I'hv created and shared a google doc with you where I'm adding some background information about the dataset and the method used.
>> 
>> I used the test case example.
>> If you get time can you please check my approach and let me know if I'm going in the wrong direction.
>> 
>> Thanks,
>> Abhinav
>> 
>> On Sat, Apr 16, 2016 at 2:52 AM, Abhinav Gupta <abhinavgupta440 at gmail.com <mailto:abhinavgupta440 at gmail.com>> wrote:
>> Hi Marcus,
>>   So I'll get started with the Internet Advertisements dataset and implement FFN for it, followed by Convolututional Network for it and FFN for Human Activity Recognition Using Smartphones dataset. I'm sharing a doc with you where I'll get started with the implementation of the first task (implementing FFN for Internet Advertisements dataset). 
>> 
>> As I'm new to the process please let me know if you feel that I should change my methodology at some points.
>> 
>> Thanks,
>> Abhinav
>> 
>> On Thu, Apr 14, 2016 at 11:11 PM, Marcus Edel <marcus.edel at fu-berlin.de <mailto:marcus.edel at fu-berlin.de>> wrote:
>> Hello Abhinav,
>> 
>> nice to hear from you. You found some really interesting dataset and except for
>> the "Anonymous Microsoft Web Data Data Set" which is a categorical dataset, we
>> can definitely use the datasets for some neat examples. I like that none of the
>> datasets looks like the other. So we could show e.g that a convolution neural
>> network is the preferred model for the Internet Advertisements Data Set as like
>> maybe a standard feed-forward network. I guess, we should start with one of the
>> datasets first and see how things go, what do you think?
>> 
>> Thanks again, for taking the time to look into this.
>> 
>> Thanks,
>> Marcus
>> 
>>> On 14 Apr 2016, at 00:58, Abhinav Gupta <abhinavgupta440 at gmail.com <mailto:abhinavgupta440 at gmail.com>> wrote:
>>> 
>>> Hi Marcus,
>>>   Sorry for the late response. 
>>> I'hv shortlisted these datasets, please have a look at them.
>>> http://archive.ics.uci.edu/ml/datasets/Internet+Advertisements <http://archive.ics.uci.edu/ml/datasets/Internet+Advertisements>
>>> http://archive.ics.uci.edu/ml/datasets/Human+Activity+Recognition+Using+Smartphones <http://archive.ics.uci.edu/ml/datasets/Human+Activity+Recognition+Using+Smartphones>
>>> http://archive.ics.uci.edu/ml/datasets/Anonymous+Microsoft+Web+Data <http://archive.ics.uci.edu/ml/datasets/Anonymous+Microsoft+Web+Data>
>>> http://archive.ics.uci.edu/ml/datasets/Tic-Tac-Toe+Endgame <http://archive.ics.uci.edu/ml/datasets/Tic-Tac-Toe+Endgame>
>>> 
>>> Please let me know if you find any one of them appropriate to mention in the example or if you have some comments regarding the datasets. In the meanwhile I'll be looking for more.
>>> 
>>> Thanks,
>>> Abhinav
>>> 
>>> On Mon, Apr 11, 2016 at 6:11 AM, Marcus Edel <marcus.edel at fu-berlin.de <mailto:marcus.edel at fu-berlin.de>> wrote:
>>> Hello Abhinav,
>>> 
>>> sounds great to me. The UCI repository is probably a good start:
>>> https://archive.ics.uci.edu/ml/datasets.html <https://archive.ics.uci.edu/ml/datasets.html> Also sharing a doc so that we can
>>> work together on it is a good idea. I look forward to hearing from you.
>>> 
>>> Thanks,
>>> Marcus
>>> 
>>>> On 10 Apr 2016, at 10:01, Abhinav Gupta <abhinavgupta440 at gmail.com <mailto:abhinavgupta440 at gmail.com>> wrote:
>>>> 
>>>> Hi Marcus,
>>>>     So I'll look for some datasets, make a list and will share them with you. Once we settle on the dataset I'll share a doc with the information about model, dataset and the example like you mentioned. And I think representing dataset with few neat images can be taken care of while choosing the dataset.
>>>> 
>>>> Please let me know if it sounds good to you.
>>>> 
>>>> Thanks,
>>>> Abhinav
>>>> 
>>>> On Sat, Apr 9, 2016 at 4:58 AM, Marcus Edel <marcus.edel at fu-berlin.de <mailto:marcus.edel at fu-berlin.de>> wrote:
>>>> Hello Abhinav,
>>>> 
>>>> I think it would be a good idea, to include some examples and to explain the
>>>> examples step by step; probably with some sort of background information about
>>>> the model and the dataset used. Maybe we can come up with some neat images, for
>>>> the dataset. If you like, you can start with that, don't feel obligated to do
>>>> so. Also, if you don't like the idea, that's also fine, in this case, we can
>>>> come up with other ideas.
>>>> 
>>>> Also, I'm not sure we should use the thyroid dataset for the examples, it's kind
>>>> of an uninteresting dataset, don't you think? Don't get me wrong, the dataset is
>>>> great and important, but maybe we can find something fancy.
>>>> 
>>>> Thanks,
>>>> Marcus
>>>> 
>>>>> On 07 Apr 2016, at 08:09, Abhinav Gupta <abhinavgupta440 at gmail.com <mailto:abhinavgupta440 at gmail.com>> wrote:
>>>>> 
>>>>> Hii Marcus,
>>>>>    Sure I would love to collaborate with you. I'hv started going through amf's documentation.
>>>>> 
>>>>> I'hv successfully run the test cases as independent examples like I mentioned in the issue: https://github.com/mlpack/mlpack/issues/562 <https://github.com/mlpack/mlpack/issues/562> If you like we can include those, or come up with more different examples.
>>>>> 
>>>>> Please let me know the procedure you would like me to follow, to collaborate in an efficient manner.
>>>>> 
>>>>> Thanks,
>>>>> Abhinav
>>>>> 
>>>>> 
>>>>> On Wed, Apr 6, 2016 at 3:57 AM, Marcus Edel <marcus.edel at fu-berlin.de <mailto:marcus.edel at fu-berlin.de>> wrote:
>>>>> Hello Abhinav,
>>>>> 
>>>>> sorry for the slow response. We think that the best documentation is written by
>>>>> the person who wrote the code. That doesn't mean, we don't appreciate any help
>>>>> with the documentation. I guess the person who wrote the code, has a different
>>>>> view on what might be helpful and what is trivial as another user. So, if you
>>>>> like we can combine both views and write a strong documentation, what do you
>>>>> think?
>>>>> 
>>>>> The documentation should be based on the already existing tutorials, you could
>>>>> take a look at the amf tutorial. I think, it would be nice to include some neat
>>>>> examples.
>>>>> 
>>>>> Thanks,
>>>>> Marcus
>>>>> 
>>>>> > On 03 Apr 2016, at 09:11, Abhinav Gupta <abhinavgupta440 at gmail.com <mailto:abhinavgupta440 at gmail.com>> wrote:
>>>>> >
>>>>> > Hi Ryan, Marcus ,
>>>>> >      I'm interested in writing a basic documentation on how to use mlpack to build neural networks along with few basic examples (derived mostly from the test cases). If someone is already working on this I would love to collaborate with him/her.
>>>>> > Also is there any standard procedure that you would like me to follow ?
>>>>> >
>>>>> > Thanks,
>>>>> > Abhinav
>>>>> 
>>>>> 
>>>> 
>>>> 
>>> 
>>> 
>> 
>> 
>> 
>> 
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack/attachments/20160421/89221ad8/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4972 bytes
Desc: not available
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack/attachments/20160421/89221ad8/attachment-0001.bin>