Modified | DDT |>| WED.MAR,990310,11:55-4 | CoSy/Home ; CoSy/Current © Coherent Systems Inc .

KUG:990225
My CoSyNotes
 
 Two talks on K:
7-8pm  "Experiences of a VB / Excel User Learning K"
8-9pm  "Genetic Programming in K, part 2"
Date:         Thursday, February 25, 1999

Location:     The Zurich Insurance Group  http://www.zurich.com
              1 Chase Manhattan Plaza,  39th Floor,  New York, NY 10005
              Wall St area, one block east of Broadway.
              Located on Pine Street,  between William St and Nassau St.
              Directions: http://www.objdev.org/directions/zurich.html

Cost:         $5 contribution http://www.objdev.org/contrib
              Free to Zurich Centre employees (our hosts) and ODG paid
              annual members
Register:     Required, please use http://www.objdev.org/register

Leader:       Stevan Apter, apter@panix.com
Group:        K SIG of ODG

7:00-7:50pm   "Experiences of a VB / Excel User Learning K"
              Speaker:  Jacob Moskowitz

              Jake is a Microsoft VB / Excel user who set out to learn K
              late last year.  Now you can benefit from his experiences,
              including some tribulations, as he sets out some of the
              more subtle hazards facing programmers whose minds have
              been softened by years of exposure to Fortran, Basic, and
              their relatives.  Various tips & tricks will be
              presented.  Stevan Apter, K-SIG group leader and an expert
              K programmer, will also be present to field difficult
              questions.

Bio:          Jacob Moskowitz is a Consultant at JP Morgan, specializing
              in Microsoft Excel and VB development for Fixed Income.

8:00-9:00 pm  "Genetic Programming, part 2"
              Speaker:  Ed Bierly

              This is a continuation of the 1/26/99 presentation:
                http://www.objdev.org/news/odg9901.html#k
              on Genetic Programming where Ed sent an ant out on the
              Santa Fe trail.  In this talk, Ed will go into greater
              detail on the K code used in this application, and field
              further questions.

Bio:       Ed Bierly is the head of the software development group at
              Millenium Partners, a fund manager.  He has been there for
              8 years.  Prior to that, he was a programmer at First
              Boston.  Ed has been programming in K for the last year
              and a half.

 NAME= "#CoSyNotes"
  ============================: THU.FEB,990225 :============================
 moskowitz-jacob@jpmorgan.com .
  vs Mathcad7 : KleinGordon quantum wave pakets .  ( faster thatn MACISMA
 summing 20 complex exponentials  complex magnitude .  MathCad bear to edit .
 K can use more mathematical syntax | anthonyCorso : not from APLer view . |
 Animation bug .
 K much much faster animation .
 w Excel/VB : K just a free DLL . : Fin Pricing : Months v Excercise price
 CRR Binomial pricing model . | 16 time intervals .
 (A thru C mid speed ) .
 different tree for each line .  | 1 tree % cell .
  Jacob  Uses Spaces for clarity . 
 Same code as cell by cell . just fed in row .
 About 5 lines of K
 Equivalent C code developed & maintained by 4-5 C programmers .
 Has K code stored in Excel sheet itself . has collapsible heirarchical
 outlining links like ThinkTank introduced . to store K code .
 Mathematica too slow .

|20:6|  Why to k ? Ed Bierly : GeneticProgramming
 Emulating S expressions from Lisp .  probabilistic so don`t get stuck .
 Elder Mead`s(?) Algorithm : 3 points on surface - take bisector of 2 best .
 bit string or integers
 GP code manipulates  S expressions ( nested lists )
 Book  : R Koza : Genetic Programming .
 Tony Corso : Redundant fns compensating in soln ?  Ya . Need to add pressure
  to be short and consise .
 Convergance to still much more complex fn than true : |  +/ x  }
  ( containing  ( Log  exp ) canceling terms  like Tony hypothesized . )
 could use  algebraic simplifier .
 SantaFe Trail Problem  |  Tony : GP is Fitness fns on tokens |
  Result oN 23rd generation .
 If children poorer , jiggles like simulated anealing .
 More difficult Altimus trail .
 Genetic algols used to find holes in RISC architecture .
 Read Danny Hillis Book ( 82 pages ) : Breed sorting algos : Pattern in Stone
 Massivly parallel searching .
 | -/.Area fo 2 boxes  }
 Population of 4000 random programs .
 Dialog memory leaks . Lisp  Dog slow  . C++ so painful .
 Linear speed up w parallel      | % e ^ 2  )
  x/ L0 w0 ,
 faith in Sort programs - survived thru so many generations  .
|20:56|


 CoSy NoteComputing Environment & Language ;
CoSy The
NoteComputer
 Current , WallSt , MotorBoard , Art
Feedback :
bob@cosy.com
NB : I reserve the right to post all communications I receive or generate to CoSy website for further reflection .