Thanks for the heads up. I am currently not using @, @: in the gerund train.

To quickly explain what I am attempting, I am trying to build a symbolic 
regression solver, i.e. "guess the function". ( 
https://en.wikipedia.org/wiki/Symbolic_regression)
The function to be guess in my original email was sin(x)+cos(x). My solver 
"guessed" correctly, using a genetic
algorithm, but had a lot of junk lef tin the final answer... I guess analogous 
to junk DNA. At the moment my "chromosomes"
are just gerund sequences, using no adverbs or conjunctions between them. 
The fitness function is 
measuring the closeness of applying to the input vars ( gerund sequence)`:6 to 
the target outputs.
It works well, as a proof of concept at the moment.
--------------------------------------------
On Tue, 3/27/18, Jose Mario Quintana <[email protected]> wrote:

 Subject: Re: [Jprogramming] Recognizing unused parts of a tacit verb train
 To: "Programming forum" <[email protected]>
 Date: Tuesday, March 27, 2018, 8:25 AM
 
 Simple gerunds are just a list of
 verbs, for example,
 
   
 u`v`w
 ┌─┬─┬─┐
 │u│v│w│
 └─┴─┴─┘
 
 However, gerunds can have complex boxed
 structures, for example,
 
 
    (u@:v) `''
 ┌──────────┐
 │┌──┬─────┐│
 ││@:│┌─┬─┐││
 ││  ││u│v│││
 ││  │└─┴─┘││
 │└──┴─────┘│
 └──────────┘
 
    takeafter f.`''
 ┌──────────────────────────────────────────────────────┐
 │┌─┬──────────────────────────────────────────────────┐│
 ││3│┌──┬─┬───────────────────────────────────────────┐││
 ││
 ││[:│]│┌─┬───────────────────────────────────────┐│││
 ││ ││  │
 ││3│┌────────────────────────────────┬──┬─┐││││
 ││ ││  │ ││
 ││┌─┬────────────────────────────┐│}.│]│││││
 ││ ││  │ ││
 │││3│┌─────────┬─┬──────────────┐││ 
 │ │││││
 ││ ││  │
 ││ │││
 ││┌─┬─────┐│+│┌─┬──────────┐│││ 
 │ │││││
 ││ ││  │
 ││ │││ │││@│┌─┬─┐││
 ││3│┌──┬──┬──┐││││  │
 │││││
 ││ ││  │ ││
 │││ │││ ││#│[│││ ││
 ││E.│i.│1:│││││  │ │││││
 ││ ││  │ ││ │││ │││
 │└─┴─┘││ ││
 │└──┴──┴──┘││││  │
 │││││
 ││ ││  │ ││
 │││ ││└─┴─────┘│
 │└─┴──────────┘│││  │
 │││││
 ││ ││  │ ││
 │││
 │└─────────┴─┴──────────────┘││ 
 │ │││││
 ││ ││  │
 ││
 ││└─┴────────────────────────────┘│ 
 │ │││││
 ││ ││  │
 ││
 │└────────────────────────────────┴──┴─┘││││
 ││ ││  │
 │└─┴───────────────────────────────────────┘│││
 ││
 │└──┴─┴───────────────────────────────────────────┘││
 │└─┴──────────────────────────────────────────────────┘│
 └──────────────────────────────────────────────────────┘
 
 So, if for a given tacit verb
 the aim to produce an equivalent tacit verb
 which is simpler, or has more a desirable form;
 then, one can certainly
 analyze and modify
 accordingly the gerund associated with the original verb
 (regardless how this original verb was
 produced) and use the (`:6) adverb
 to obtain
 the desired verb.  (Since the structures are recursive in
 nature,
 a natural solution is to employ a
 recursive verb for producing the gerund
 associated to the desired verb; anyway, that is
 what I do)
 
 One word of
 caution: one must be aware of potential intended side
 effects,
 for example, u [ v and u might not
 be equivalent (even if v does not
 produce an
 error).
 
 
 On
 Sun, Mar 25, 2018 at 7:47 PM, 'Jon Hough' via
 Programming <
 [email protected]>
 wrote:
 
 > Thanks, this
 makes sense. Actually it is even easier, hopefully. The
 verb
 > is formed by `:6 a sequence of
 > gerunds. I can figure out which gerunds
 terminate which subsequences of
 > the
 gerund train before  even
 > performing
 `:6.
 >
 > e.g +:`[`^
 > only +: is evaluated.
 >
 > Thanks,
 > Jon
 >
 --------------------------------------------
 > On Fri, 3/23/18, Henry Rich <[email protected]>
 wrote:
 >
 >  Subject:
 Re: [Jprogramming] Recognizing unused parts of a tacit verb
 train
 >  To: [email protected]
 >  Date: Friday, March 23, 2018, 6:19 PM
 >
 >  When dissect is
 constructing the
 >  display, gets to
 >
 >  (f [ h)
 >
 >  it knows that the
 execution of
 >  h is immaterial unless
 it encounters an
 >  error.  It displays
 only the f part.
 >
 > 
 It then grays out any part of
 >  the
 sentence that is not displayed.
 >
 >  Notice that it also grays out the
 leading ] and
 >  the ] in the @:],
 which
 >  are also not
 >  needed.
 >
 >  There is an option
 >  to display both sides of ][.
 >
 >  Using the
 transformations
 >
 > 
 (f [ h) -> f
 >  (] g@]) ->
 >  (] g)
 >  (] g) ->
 g@]
 >  (f@] g
 > 
 h@])-> (f g h)@]
 >
 >  you can
 >  turn it
 into
 >
 >  (2&o.
 +
 >  1&o.)@]
 >
 >  Henry Rich
 >
 >
 >
 >  On 3/23/2018 3:26 AM, 'Jon
 Hough' via
 >  Programming wrote:
 >  > Consider the
 > 
 following tacit verb
 >  > f1=: ]
 >  (2&o.@:] + 1&o.@:] [ 5&+@:]
 7&+@:] ([: <.
 >  ]) 10&*@:]
 7&*@:])
 >  >
 >  > It looks strange, and is pretty
 long. The
 >  interesting point is that
 only the left
 >  >
 >  section  ] (2&o.@:] + 1&o.@:] 
  is useful. The
 >  rest of it is not
 evaluated.
 >  > i.e. f1 is
 >  equivalent to sin(x) + cos(x).
 >  >
 >  > I
 noticed that the dissect tool recognizes
 >  this fact.
 >  >
 dissect '(] (2&o.@:]
 >  +
 1&o.@:] [ 5&+@:] 7&+@:] ([: <. ])
 >  10&*@:] 7&*@:])) 1'
 >  >
 >  >
 Everything but the sin(x) + cos(x) par tis
 >  greyed out, implying it is not
 evaluated.
 >  > My question is, how
 does dissect recognize
 >  this? And is
 it possible to extract a simplified
 > 
 > tacit verb from f1?
 >  >
 >  >
 >  >
 >  Thanks,
 >  >
 Jon
 >  >
 > 
 ----------------------------------------------------------------------
 >  > For information about J forums see
 http://www.jsoftware.com/forums.htm
 >
 >
 >  ---
 >  This email
 has been checked for viruses by
 > 
 AVG.
 >  http://www.avg.com
 >
 > 
 ----------------------------------------------------------------------
 >  For information about J forums see http://www.jsoftware.com/forums.htm
 >
 ----------------------------------------------------------------------
 > For information about J forums see http://www.jsoftware.com/forums.htm
 >
 ----------------------------------------------------------------------
 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