What about Lisp-Stat,
http://lambda-the-ultimate.org/node/617 ,
the more-or-less defunct statistical package? Are not Ihaka and Lang
re-inventing a wheel that has already fallen off? Might efforts be
better directed at reviving it?
Disclaimer: I know little about Lisp-Stat other than that it is/was
Lisp-based.
In article
<····································@j35g2000yqh.googlegroups.com>,
·········@gmx.net wrote:
> What about Lisp-Stat,
>
> http://lambda-the-ultimate.org/node/617 ,
>
> the more-or-less defunct statistical package? Are not Ihaka and Lang
> re-inventing a wheel that has already fallen off? Might efforts be
> better directed at reviving it?
>
> Disclaimer: I know little about Lisp-Stat other than that it is/was
> Lisp-based.
The software is XLISP-STAT, which is based on XLisp
(an mostly abandoned dialect of Lisp with its own implementation).
XLisp was one of these Lisp implementations in C.
An interpreter (IIRC) and not very fast. It implements also
its own Lisp language (with some Common Lisp influence).
XLisp was quite successful MAAAAAANY years ago.
Now it is mostly dead. AutoLisp was once created based
on XLisp. David Betz later created a new XLisp implementation
based on Scheme - a different language.
These guys moved away from XLisp to languages like R.
Now they feel that R has not a bright future as
a implementation of an efficient programming language.
Going back to XLisp is no option, since it is no longer
used/maintained and also relatively 'slow'.
The authors of the paper argue that by using Common Lisp:
* they can write more code in Lisp and less code in C.
since Common Lisp implementations have a wider
range of performance (due to providing
optimizing, incremental, native code compilers)
and the implementations are mostly written in Common Lisp
themselves.
* they get the choice of several Common Lisp implementations that
are maintained, so they don't have to maintain
the language implementation.
* they get a standard language with features and extensions
that XLisp does not provide or in a more primitive form.
They don't have to maintain/invent their own programming
language.
Their goal is to get an interactive system and reasonable
performance at the same time - something that
several Common Lisp implementations can provide.
--
http://lispm.dyndns.org/
On Dec 5, 12:05 pm, Rainer Joswig <······@lisp.de> wrote:
> Their goal is to get an interactive system and reasonable
> performance at the same time - something that
> several Common Lisp implementations can provide.
>
> --http://lispm.dyndns.org/
See also Common Lisp Stat, which is an attempt to fast-forward
LispStat to Common Lisp, replace the hacky matrix stuff with a more
modern approach (lisp-matrix as well as rif and tamas' extensions to
that approach), and generally be a sensible platform for research in
statistical computing (as opposed to a sensible platform for
statistical computing, which is what R is right now -- but it's a bit
hellish for research in the area).
http://github.com/blindglobe for list of the current git repos and
dependencies (at least some of them).
There is an older set/mirror on repo.or.cz that I need to update.
If anyone is interested, I'd be happy to talk more. I'm also keeping
an R within CL bridge maintained (rif did the development) and am
generally keen on Common Lisp statistical stuff.
later...
A bit more about Common LispStat -- it's not entirely LispStat like,
as there are key problems with the LispStat approach, but the goal is
to create a DSL within lisp for statistical analysis (i.e. machine
learning, hypothesis discovery and testing, inference, regression
modeling, etc) which is reasonably general, plays well with other CL
packages, tries to be a toolkit and not a complete solution, and keeps
the syntax as much as possible (as well as playing with with CL data
structures when possible.
It's slowly making progress, it doesn't get much done today (again,
the goal for me is to create a research platform, not a functional
tool which works tommorow -- R does the latter for me), and that means
enjoying the combination of "what could I do if..." and "let's debug
to see what isn't there yet", slowly replicating what I do at work, in
a common lisp environment, as an on-going gap-analysis.
Anyway, I'd be happy to collab with others, it's just my tram-ride
commute hobby right now...
From: Tamas K Papp
Subject: Re: Back to the Future: Lisp as a Base for a Statistical Computing System
Date:
Message-ID: <6q4ok6Fat8b7U1@mid.individual.net>
On Sun, 07 Dec 2008 22:23:29 -0800, AJ Rossini wrote:
> It's slowly making progress, it doesn't get much done today (again, the
> goal for me is to create a research platform, not a functional tool
> which works tommorow -- R does the latter for me), and that means
I find that development is more rapid when I want functional tools, so I
switched to CL entirely for my statistical analysis. But as I mostly do
Bayesian stuff, I don't need a lot of libraries anyway. MCMC can be dead
slow on R since it is not vectorized.
> Anyway, I'd be happy to collab with others, it's just my tram-ride
> commute hobby right now...
I have an almost-complete 2d graphing library based on cl-cairo2 in the
works. I plan to clean it up and post it by February (I will be
superbusy until then).
Cheers,
Tamas
On 8 Dec 2008 14:17:42 GMT, <······@gmail.com> wrote:
>
> I have an almost-complete 2d graphing library based on cl-cairo2 in the
> works. I plan to clean it up and post it by February (I will be
> superbusy until then).
Wonderful news! I just stumbled across Cairo as an answer to my
plotting needs recently. It is also one of those laptop projects for
'away-from-office time.
Need an alpha tester?
--
"Most programmers use this on-line documentation nearly all of the
time, and thereby avoid the need to handle bulky manuals and perform
the translation from barbarous tongues." CMU CL User Manual
** Posted from http://www.teranews.com **