rdx - Automatic Differentiation in R

[Chidambaram Annamalai]

Hi all,

I have submitted my proposal for Automatic Differentiation in R, titled rdx to the official GSoC appspot page, where, as mentors/admins, you can add comments and review my proposal. I would really appreciate feedback, deficiencies regarding the proposal either here (publicly) on the list or on the proposal page.

Regards,
Chillu

PS: Since the earlier revisions of my proposal lived in a wiki, I have converted the proposal to html and the rendering is not perfect.

[timgluz]

Ahoi,

i read your proposal - I like it, that's comprehensive . i'll take it
as model for my proposal.

But i have some questions - you mention that you used R for machine
learning task, did you miss workflow desinger(specially for data
preparation, cleansing), like RapidMiner Knime, Pasw modeller etc
already have? or you enjoyed R scripting? How big were your ML
projects? Were those school projects or yours? ( These are needed for
my ideas)

Good luck with this project - thats really challenging task !

Tim

On Mar 30, 1:21 am, Chidambaram Annamalai <quantumeli...@gmail.com>
wrote:
> Hi all,
>
> I have submitted<http://socghop.appspot.com/gsoc/student_proposal/show/google/gsoc2010...>my

[Chidambaram Annamalai]

On Thu, Apr 1, 2010 at 5:51 PM, timgluz <tim...@gmail.com> wrote:

Ahoi,

i read your proposal - I like it, that's comprehensive . i'll take it
as  model for my proposal.

But i have some questions  - you mention that you used R for machine
learning task, did you miss workflow desinger(specially for data
preparation, cleansing),  like RapidMiner Knime, Pasw modeller etc
already have?  or you enjoyed R scripting? How big were your ML
projects? Were those school projects or yours? ( These are needed for
my ideas)

I have no experience with professional data mining software. The tasks that you are referring to were simply assignments for my machine learning course which involved implementing algorithms that we had learned in class, testing them on datasets from the UCI repository, plotting the results graphically. They were not full fledged projects, but exercises designed to bring out the experimental flavor of the course. Specific examples include maximum margin SVMs, k-means clustering and PCA.

Regards,
Chillu

[Prof. John C Nash]

It may help timgluz to know about the report I wrote on free/libre data mining software
under contract for the Treasury Board Secretariat of the Canadian government. It is at

http://macnash.telfer.uottawa.ca/~nashjc/flbi-nis.pdf

and is Gnu Public documentation licensed as a condition of the contract. One of the main
tools we found and recommended was rattle, which is an R package (it is NOT trivial to
install or run -- see the R-wiki for some pointers on installation).

If anyone uses this report, may I request the courtesy of a citation; it would also be
interesting to know of such citations.

For the record, I'm the mentor of the ADinR project Chillu is proposing to attempt.

John Nash