Re: [Topic-models] Topic modeling w/ neural nets
Timmy Wilson
> curious about what you'd had in mind as the "recent resurgence / success of
> neural nets" - any good papers in the last 5 years or so?
Hi Anthony,
I've been thinking about this a lot lately.
I think this gets to an interesting/deeper question.
This paper does a good job of touching on this issue -- but there's
probably a lot more work in this area i haven't discovered yet::
'Letting structure emerge: connectionist and dynamical systems
approaches to cognition'
I'm interested in models that embrace/respect/admire nature + evolution.
Talk to any behavioral economist -- people don't make decisions based
on conditional probability analysis -- it's both simpler and more
complex/nuanced/beautiful than that.
How does nature work?
How did we get here?
How do we handle complexity?
I'm pretty sure the answers to these questions are the same (or
similar in spirit) to this answer:
How can we make machines more intelligent and useful to society?
On Mon, Dec 12, 2011 at 4:05 PM, Anthony Di Franco <di.f...@gmail.com> wrote:
> Hi Timmy, as someone who came from neural nets over to topic models, I'm
> curious about what you'd had in mind as the "recent resurgence / success of
> neural nets" - any good papers in the last 5 years or so?
>
> Anthony
>
> On Nov 9, 2011 4:33 AM, "Timmy Wilson" <tim...@smarttypes.org> wrote:
>>
>> Hi topic modelers,
>>
>> Given the recent resurgence/success of neural nets, i'm curious how
>> they'll do w/ topic modeling
>>
>> I did some exploring, and found Ruslan Salakhutdinov's -- 'Replicated
>> Softmax: an Undirected Topic Model'
>>
>> Ruslan details a RBM based topic model, and tests it's performance
>> using Annealed Importance Sampling
>>
>> I'm curious about two things:
>>
>> Is Annealed Importance Sampling the best method for general model
>> assessment? I found this thread which seems to imply it's not --
>>
>> https://lists.cs.princeton.edu/pipermail/topic-models/2010-March/000758.html
>>
>> Has anyone else gone down this path (topic modeling w/ neural nets)?
>> What are some of the hurdles to be aware of?
>>
>> Thanks,
>> Timmy Wilson
>> _______________________________________________
>> Topic-models mailing list
>> Topic-...@lists.cs.princeton.edu
>> https://lists.cs.princeton.edu/mailman/listinfo/topic-models