Ah, yes. Thanks Raul.
Perhaps this is a question for the programming list. I seem to always write
functions like this:
<bunch of local tacit definitions>
<global tacit definition that references above local definitions>
And then I flatten the final function using f. in the global definition.
This makes these big long tacit definitions more readable. I'm just
concerned with cluttering the namespace in a file with all those local
definitions. If I'm writing functions like this, should I put them each in
a separate file? Should I create a locale for every function? Just
wondering if this is good practice to write functions like this, and if so
how to deal with the locals. And since it's tacit, AFAIK I can't use the
locality of an explicit definition like 3 : 0. And 13 : 0 only works on one
line.
An example follows my signature in case that wasn't clear. There are two
functions, and they both locally try to define 'mu.' Here it isn't a
problem because it just gets redefined. It just creeps me out having the
local definitions accessible by both function definitions. I am comfortable
putting these in separate files. But it would be nice to know a way to put
them in the same file.
Thanks again!
Chris
NB. joint_prior mu;B;Q;L
mu =. 0&{::@:]
B =. 1&{::@:]
Q =. 2&{::@:]
L =. 3&{::@:]
Pa =. */@:(2&{.)@:$@:B -@:<. #@:[
mu_Pa =. Pa {. mu
B_Pa =. Pa {."1 ,./@:B
Q_Pa =. Pa {."1 Pa {. [
mu_W_t =. [ , mu + (B_Pa (+/ .*) mu_Pa)
s11 =. [
s12 =. B_Pa (+/ .*) (Pa {. [)
s21 =. |:@:s12
s22 =. Q + (B_Pa (+/ .*) Q_Pa) (+/ .*) |:@:B_Pa
sigma_W_t =. ((s11 , s12) ,. (s21 , s22)) f.
mu_W =. mu (mu_W_t ^: (<:@:L@:])) ]
sigma_W =. Q (sigma_W_t ^: (<:@:L@:])) ]
joint_prior =: (mu_W ; sigma_W) f.
NB. condition_LCG mu;sigma;yidx;yval
mu =. 0&{::
sigma =. 1&{::
yidx =. 2&{::
xidx =. -.@:yidx
yval =. 3&{::
mux =. -.@:yidx # mu
muy =. yidx # mu
sxx =. xidx #"1 xidx # sigma
sxy =. yidx #"1 xidx # sigma
syx =. xidx #"1 yidx # sigma
syy =. yidx #"1 yidx # sigma
mu_x_given_y =. mux + sxy (+/ .*) %.@:syy (+/ .*) (yval - muy)
sigma_x_given_y =. sxx - sxy (+/ .*) %.@:syy (+/ .*) syx
condition_CLG =: (mu_x_given_y ; sigma_x_given_y) f.
On Tue, Mar 23, 2010 at 12:12 PM, Raul Miller <[email protected]> wrote:
> You have probably long since realized this, but, in:
>
> s11 =. [
> s12 =. padB@:[ (+/ .*) [
> s21 =. |:@:s12
> s22 =. ] + ((padB@:[ (+/ .*) [) (+/ .*) |:@:padB@:[)
> absorbNode =: ((s11 , s12) ,. (s21 , s22)) f.
>
> you could have instead used
>
> s22 =. ] + s12 (+/ .*) |:@:padB@:[
>
> (and those parenthesis are unnecessary.)
>
> FYI,
>
> --
> Raul
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
