What is some sample data for corr with your expected result.  Thanks

Linda

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Dan Bron
Sent: Tuesday, February 19, 2013 6:48 PM
To: [email protected]
Subject: Re: [Jprogramming] applying >1 gerunds to a set of items

Raul wrote:
>  I would be cautious about meta-arguments.  When we are entirely 
> abstract  it's easy to lose track of what we are talking about.

Agreed; I experienced this a bit while reading your reply.  Though of course
it's been some weeks since I thought about this, and a lot has happened
since then, so that could contribute as well.

I'm interested in metaprogramming in J: that is, writing programs which
produce programs.  I'm interested in how we reason about these, what uses
they have, what tools we have at our disposal to effect them, and how we
build trust in them.  My intuition, built up from years of working with J
and reading J- and APL-related materials, suggests that function composition
(i.e. tacit code) is at the heart of this topic.

So, following your line of thought:

>  I do not think they can replace concrete examples.

I am considering writing an Wiki essay on this topic, using statistical
correlation, expressed as a tacit verb, as the motivating example:

        corr =: (+/@:* % *&(+/)&.:*:)&(- +/%#)

It's got specimens of all the major compositions in J: fork, hook, atop,
compose, and under. And, if you know J, it's a clear improvement over the
standard mathematical notation, because you can literally see some of the
symmetries that underlie that formula: the left tine is the sum of the
product and right is the product of the sums.

Plus the whole thing is algebraically reduced.  What that means is, in
contrast to the SMN where you have Xs and Ys and X_BARs and Y_BARs scattered
all over the place, in corr it's clear that the very first thing we do is
standardize the variables, and thereafter all we ever deal with is the
delta, and so we are immune to changes in scale, units of measurement, etc.

Or, maybe I can just punt and wait for Pepe's surprise:

>  Something is cooking, stay tuned...

I took the F train to work today, and you can bet I'll be the first to board
whatever F Trains you get back on the tracks...

-Dan


----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to