spark-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ameet Talwalkar <am...@eecs.berkeley.edu>
Subject Re: ML Algos
Date Sat, 17 Aug 2013 21:23:39 GMT
PMML is not on our near-term roadmap.  That being said, we'd welcome
contributions that extend our system to support it.


On Sat, Aug 17, 2013 at 12:32 PM, Paco Nathan <ceteri@gmail.com> wrote:

> +1 ADMM
>
> Also, in terms of commercial adoption, is support for PMML on the roadmap?
>
>
>
> On Sat, Aug 17, 2013 at 11:41 AM, Ameet Talwalkar <ameet@eecs.berkeley.edu
> > wrote:
>
>> Thanks for the clarification!  Yes, we aim to add ADMM to MLI in the
>> upcoming months.
>>
>>
>> On Sat, Aug 17, 2013 at 11:34 AM, Gowtham N <gowtham.n.mail@gmail.com>wrote:
>>
>>> Yes. It is the Stanford optimization algorithm.
>>>
>>>
>>> On Saturday, August 17, 2013, Nat wrote:
>>>
>>>> Hi,
>>>>
>>>> I am tempted to think it is from
>>>> http://www.stanford.edu/~boyd/papers/pdf/admm_slides.pdf
>>>>
>>>>
>>>> On Sat, Aug 17, 2013 at 10:26 AM, Ameet Talwalkar <
>>>> ameet@eecs.berkeley.edu> wrote:
>>>>
>>>>> Our upcoming release will include the Alternating Least Squares
>>>>> algorithm for Collaborative filtering. There are of course several other
>>>>> techniques for CF (e.g., SGD-based, trace-norm, neighborhood methods,
>>>>> divide-and-conquer approaches, etc.), and we hope that some of these
>>>>> methods will be added, with initial implementations facilitated by the
>>>>> high-level abstractions provided by the MLI.
>>>>>
>>>>> We are looking into adding neural network functionality (as noted in
>>>>> my previous email).
>>>>>
>>>>> Could you explain what you mean by "ADM" -- I'm not familiar with that
>>>>> acronym?
>>>>>
>>>>>
>>>>> On Fri, Aug 16, 2013 at 7:37 PM, Gowtham N <gowtham.n.mail@gmail.com>wrote:
>>>>>
>>>>>> Is anyone working on neural networks, ADM, Collaborative Filtering
>>>>>> etc?
>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>> --
>>> Gowtham Natarajan
>>>
>>
>>
>

Mime
View raw message