Re: [mlpack] Robotic Arm - GSoC Project Idea

[Vaibhav Jain]

Hey Marcus,
I submitted my proposal draft quite a while ago but haven't received any
inputs yet. I can understand that there are a lot of proposals coming
through at this point of time. But since final deadline line very near, can
you let me know what you think at once.
Thanks for your hard work.

Regards
-- Vaibhav Jain
___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Re: [mlpack] Paper Describing Saddle Points in LRSDP

[Ryan Curtin]

On Mon, Mar 26, 2018 at 11:06:19AM -0400, Abhijeet Krishnan wrote:
> Hi Ryan,
> 
> You had mentioned a tech note/paper by Sam Burer describing saddle
> points in LRSDP. I sent him an email regarding the same, and he
> believed this paper was being referred to. Is it the right one?

Hi Abhijeet,

Yeah, I am sorry, I should have reviewed my emails first.  I don't think
it was a paper that he himself wrote but he linked me to this paper:

https://link.springer.com/article/10.1007%2Fs10898-008-9328-4

There was no attachment in your email, so I am guessing that the papers
are the same.  Sorry for the confusion!

Thanks,

Ryan

-- 
Ryan Curtin| "The greater good..."
r...@ratml.org |   - Neighborhood Watch Alliance
___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

[mlpack] GSoC in Reinforcement Learning

[András Attila Sulyok]

Hi All,

I'm Attila Sulyok, second year MSc Computer Engineer student at the PPCU in
Budapest. I am also interested in participating the Google Summer of Code
this year, specifically developing the reinforcement learning modules of
mlpack.

One of my ideas is implementing the modification to the DQN algorithm
described here: [1] that uses (discretised) value distributions instead of
value functions. The trivial approach would be to implement it as a
separate algorithm (like QLearning) or modify the existing one, but I think
it's more general than that: it should be possible to use it with all
value-function-based algorithms. One idea is to hack it into a layer (not
sure if possible), the other is to extract the Q update part of the code
into a parameter, sort of like a loss function.

As I understand, the current state of the art algorithm for learning
continuous actions using value-functions is NAF [2], this may also benefit
from value distributions.

The third idea that I found is Hindsight Experience Replay [3], that wraps
a learning algorithm like DQN or NAF and creates additional goals to learn.

Would mlpack benefit from implementing these? Since the reinforcement
learning part is not large, they shouldn't require large modifications to
existing code.

I built the code and tested with some small algorithms; and one thing I
noticed (having only used keras-rl before) is the lack of metrics output
from training. Is that intentional? I've never used RL in the industry,
only for research (in my current thesis project), so I'm not quite sure
whether it would be useful. Same thing with the current state of the RL
agent not being visible.

Thanks,
Attila

[1]: https://arxiv.org/pdf/1707.06887.pdf
[2]: https://arxiv.org/abs/1603.00748.pdf
[3]: https://arxiv.org/pdf/1707.01495.pdf
___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Re: [mlpack] MVU Bug Fix GSOC

[LI Xuran]

Hi Ryan,


>I took a quick look at your proposal and I think it is relatively clear
>and sufficiently detailed.  I am not clear on exactly what you mean by
>"5.  it might also be useful to write an algorithm to pre-process the
>dataset to make it smoother and convex"---note that the LRSDP algorithm
>breaks the convexity of SDPs and it is a nonconvex optimization.


Thanks a lot for the feedback! Yeah, you are right about the LRSDP is a 
nonconvex optimization, what I intend to say is that during the process of 
sample generation, it might be helpful if we can have some constraints (say 
convex and closed? I am not too sure about the detail constraints yet but there 
is a theorem in the paper of local minima specifying it )  on the data to 
guarantee that an optimal solution can always be reached on the sample created, 
both by LRSDP and a dual solver.


I 've gone through the rest of papers and Posts about LRSDP on Github during 
the past few days, as I am interested in what effort have been made to debug 
it. Do you think the following  ideas would be helpful to the debug project:

  1.  refactor the SDP class to allow detail constraints specified in input.
  2.  create variable template to specify linear/sparse/dense constraints on 
input or A(constraint matrix) and support evaluation of Tr(A_i * UU^T)


I didn't add these two to my proposal, but if you think implementing those 
would be useful than I can also look into it and take it as a part of the 
project. Maybe I can try to approach it during the community bound.

Thanks again and Best Regards.
Daniel Li




The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Re: [mlpack] Fix MVU+LRSDP in GSoC 2018

[Kaiqiang Xu]

Hello, Ryan

Thanks for your reply. I modified my proposal based on your suggestion and
knowledges obtained from papers and codebase.

Best regards,
Kaiqiang

On Tue, Mar 27, 2018 at 1:09 AM, Ryan Curtin  wrote:

> On Sat, Mar 24, 2018 at 07:05:05AM +0800, kaiqiang Xu wrote:
> > Hi, Ryan
> >
> > Sorry to borther you directly by mistake in last email.
> > I modified proposal and emphasize approaches. Now it has been submited to
> > GSoC. Can you check it and give me some feedback  if available?
>
> Hi Kaiqiang,
>
> I took a quick look at the proposal; it seems me to like step 2 might be
> a little redundant.  You have written:
>
>   "2. Check the correctness of implementation of LRSDP.  Design several
>   special simple/hard problem, such as the Lovasz theta SDP, the
>   maximum cut SDP relaxation and etc."
>
> but I thought I should point out that we already have these in mlpack;
> see src/mlpack/tests/lrsdp_test.cpp.  So maybe this will help you save
> some time since you will not need to implement that yourself.
>
> Thanks,
>
> Ryan
>
> --
> Ryan Curtin| "You're lucky."
> r...@ratml.org |  - Giles
>
___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Re: [mlpack] gsoc proposal

[Ryan Curtin]

On Mon, Mar 26, 2018 at 10:52:05AM +0300, Артём Лян wrote:
> Hello mlpack mentors. 
> My name is Lyan Artyom. 
> Could you please review my proposal, that uploaded as draft at 
> summerofcode.withgoogle.com
> Thanks in advance. 

Hi Artyom,

Thanks for submitting a proposal.  I took a look at it.  I would suggest
that our expectations for strong proposals are that they have some more
detail; basically, I see a detailed timeline (that is nice, thank you
for that) but not so much detail on how exactly the JNI bindings would
fit into the existing automatic bindings section.

It looks like you have left some time already in the timeline for
studying the existing automatic binding implementations, but most
proposals for GSoC submitted to mlpack will already have done a study of
the existing code and will have an in-depth plan for how the work will
be implemented, so you might consider doing the same.

I hope this is helpful.

Thanks!

Ryan

-- 
Ryan Curtin| "Not like this...  not like this."
r...@ratml.org |   - Switch
___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Re: [mlpack] GSOC

[Ryan Curtin]

On Sun, Mar 25, 2018 at 08:40:37PM -, aditya mitkari wrote:
> Can the PCA API code be considered for parallelization using openMP?

Hi Aditya,

I think in many cases you can just use OpenBLAS and this will
parallelize most of the operation of PCA.  But if you see that there are
still parts that OpenBLAS will not parallelize, then I would agree that
the PCA code would be useful to apply OpenMP to.

Thanks,

Ryan

-- 
Ryan Curtin| "We need some time for some things to happen!"
r...@ratml.org |   - Bells
___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Re: [mlpack] Fix MVU+LRSDP in GSoC 2018

[Ryan Curtin]

On Sat, Mar 24, 2018 at 07:05:05AM +0800, kaiqiang Xu wrote:
> Hi, Ryan
> 
> Sorry to borther you directly by mistake in last email.
> I modified proposal and emphasize approaches. Now it has been submited to
> GSoC. Can you check it and give me some feedback  if available?

Hi Kaiqiang,

I took a quick look at the proposal; it seems me to like step 2 might be
a little redundant.  You have written:

  "2. Check the correctness of implementation of LRSDP.  Design several
  special simple/hard problem, such as the Lovasz theta SDP, the
  maximum cut SDP relaxation and etc."

but I thought I should point out that we already have these in mlpack;
see src/mlpack/tests/lrsdp_test.cpp.  So maybe this will help you save
some time since you will not need to implement that yourself.

Thanks,

Ryan

-- 
Ryan Curtin| "You're lucky."
r...@ratml.org |  - Giles
___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

Re: [mlpack] MVU Bug Fix GSOC

[Ryan Curtin]

On Thu, Mar 22, 2018 at 08:47:58PM +, LI Xuran wrote:
> Hi Ryan,
> 
> Thanks a lot for your kind advice. As you suggested, I've added
> details to my new ideas and integrated them into my proposal.  I 've
> submitted a draft proposal to GSoC website just now, to detail my
> current plan on MVU testing. Is it possible that I might have some
> feedback on my draft proposal before the deadline? I really want to
> make a proposal of great quality and feasibility on this subject, so
> that The implementation can be tractable and under control.

Hi Daniel,

I took a quick look at your proposal and I think it is relatively clear
and sufficiently detailed.  I am not clear on exactly what you mean by
"5.  it might also be useful to write an algorithm to pre-process the
dataset to make it smoother and convex"---note that the LRSDP algorithm
breaks the convexity of SDPs and it is a nonconvex optimization.

Thanks,

Ryan

-- 
Ryan Curtin| "Hungry."
r...@ratml.org |   - Sphinx
___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

[mlpack] Paper Describing Saddle Points in LRSDP

[Abhijeet Krishnan]

Hi Ryan,

You had mentioned a tech note/paper by Sam Burer describing saddle points in 
LRSDP. I sent him an email regarding the same, and he believed this paper was 
being referred to. Is it the right one?

Regards,
Abhijeet Krishnan

___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack

[mlpack] gsoc proposal

[Артём Лян]

Hello mlpack mentors. 
My name is Lyan Artyom. 
Could you please review my proposal, that uploaded as draft at 
summerofcode.withgoogle.com
Thanks in advance. 


-- 
All the best, 
Artyom Lyan___
mlpack mailing list
mlpack@lists.mlpack.org
http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack