FYI: 3 New GP Papers (fwd)

Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Mon, 11 Nov 1996 12:46:37 +0100 (MET)

Messages sorted by: [ date ][ thread ][ subject ][ author ]
Next message: J. D.: "Re: TWA 800: THE CAT IS OUT !!"
Previous message: David Musick: "Visual Languages"

Evolvable hardware is certainly the wave of the future. Notice that CAM=20
circuitry and evolvable hardware development streams are starting to=20
converge. I am very excited about this, can hardly wait until this stuff=20
hits the market. ( WYWIWYG, man ;)

---------- Forwarded message ----------
Date: Sat, 9 Nov 1996 12:22:41 -0800 (PST)
From: John R. Koza <koza@CS.Stanford.EDU>
To: CA@think.com
Subject: 3 New GP Papers=20

=FE7#=04=19=01=10=D0=17=A5=0F=D0=161=161=161=161=161=0E=16? =16_=16_=16_=16=
_
=16i
=16s=16sH=16=BBx=16_=173 =17S=1A=17{*=17=A5=161=01=01=17{=17m=0E=17{=17{=17=
=A5=17{=17{=17{=17{=17{=17{
THREE NEW PAPERS ON GENETIC PROGRAMMING
NOW AVAILABLE IN POST SCRIPT

------------------------------------------
These 3 papers are on evolutionary design of electrical circuits=20
using genetic programming and were presented recently at various=20
conferences in Japan. .=20
------------------------------------------
All 3 papers are available in Postscript via WWW under "Research=20
Publications" and "Recent Papers" on John Koza's home page at=20
Stanford University:
http://www-cs-faculty.stanford.edu/~koza/
------------------------------------------

"Design of a 96 Decibel operational amplifier and other problems=20
for which a computer program evolved by genetic pogramming is=20
competitive with human performance" =20

By=20
John R. Koza
David Andre
Forrest H Bennett III and
Martin A. Keane=20

Presented on October 5, 1996 in Ashikaga Japan and appearing in=20
Gen, Mitsuo and Zu, Weixuan (editors). Proceedings of l996 Japan-
China Joint International Workshop on Information Systems. =20
Ashikaga: Ashikaga Institute of Technology. Pages 30 - 49.=20

It would be desirable if computers could solve problems without=20
the need for a human to write the detailed programmatic steps. =20
That is, it would be desirable to have a domain-independent=20
automatic programming technique in which "What You Want Is What=20
You Get" ("WYWIWYG" =D0 pronounced "wow-eee-wig"). Genetic=20
programming is such a technique. This paper surveys three recent=20
examples of problems (one from the field of cellular automata and=20
two from the fields of molecular biology) in which genetic=20
programming evolved a computer program that produced results that=20
were slightly better than human performance for the same problem. =20
This paper then discusses a fourth problem in greater detail and=20
demonstrates that a design for a low-distortion 96 decibel op amp=20
(including both topology and component sizing) can be evolved=20
using genetic programming. The information that the user must=20
supply to genetic programming consists of the parts bin=20
(transistors, resistors, and capacitors) and the fitness measure=20
for the major operating characteristics of an op amp.=20
------------------------------------------

"Reuse, parameterized reuse, and hierarchical reuse of=20
substructures in evolving electrical circuits using genetic=20
programming"

By=20
John R. Koza
Forrest H Bennett III
David Andre and
Martin A. Keane=20

Presented October 8, 1996 in Tsukuba, Japan and appearing in=20
Proceedings of International Conference on Evolvable Systems: From=20
Biology to Hardware (ICES-96). Lecture Notes in Computer=20
Science, Volume ---. Berlin: Springer-Verlag.=20

Most practical electrical circuits contain modular substructures=20
that are repeatedly used to create the overall circuit. Genetic=20
programming with automatically defined functions and architecture-
altering operations successfully evolved a design for a two-band=20
crossover (woofer and tweeter) filter with a crossover frequency=20
of 2,512 Hz. Both the topology and the sizing (numerical values)=20
for each component of a the circuit were evolved. In the evolved=20
circuit, three different electrical substructures were used; one=20
was invoked five times; and one was invoked as part of a=20
hierarchy; and one substructure was invoked with different=20
numerical arguments so that different numerical component values=20
were assigned to the substructure's components. =20
------------------------------------------

"Evolution of a 60 Decibel op amp using genetic programming" =20

By=20
Forrest H Bennett III
John R. Koza
David Andre and
Martin A. Keane=20

Genetic programming was used to evolve both the topology and=20
sizing (numerical values) for each component of a low-distortion,=20
low-bias 60 decibel (1000-to-1) amplifier with good frequency=20
generalization. =20

=02uh=08=19=01=90=3D=E0/=D0=05=A0=05=A0=05=A0=05=A0=02=D0B=01=01=01=02=01=
=16=07Courier=01=10=D0=10=D3=FA
@=02=18=18=02=01=01=01=018=01j=01k=01=96=01=D8=02=1A=023=02^=02=A0=02=DF=02=
=F4=03=1E=03I=03J=03=8C=03=CB=03=F1=03=F2=03=F6=04=03=04=0F=04)=04:=04;=04|=
=04=BF=04=FC=059=05:=05{=05=BC=05=F9=06:=06u=06=B7=06=FA=076=07y=07=BD=07=
=FF=08B=08=81=08=C1=08=FA=09<=09s=09=9E=09=9F=09=D7
=14
!
"
&
3
I
Y
j
k
=A9
=ED=0B)=0BY=0BZ=0B=9C=0B=DD=0C =0Cb=0C=A4=0C=E8
+
m
=A7
=E3=0E%=0EW=0E=82=0E=83=0E=C2=0E=C3=0E=C7=0E=DD=0E=EA=0E=FA=0F=0B=0F=0C=0FJ=
=0F=8E=0F=CA=0F=FB=0F=FC=10:=10}=10=BD=10=CF=10=D0=FB=FB=FB=FB=FB=FB=FB=FB=
=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=
=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=
=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=
=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=FB=04=01$=90=0B`=05=FF=FF=0F
=18=16=14=14=FF=13=07=01=FF=07=02=01=03=DE=02=0F=D0=04=10=D0=09j=0F=D0=10=
=FF=FF=01=01=FF=FF=02=01=10=D3=09=01=10=D0
=02=03=04
=0E=0F=10=12=14=15=16=17!"=04?=04=89=05^=05k=05=95=05=B7=06=92=06=A6=06=DC=
=07=A3=07=BD=07=D1=08=BC=09=D3=09=F0
M
e=0B=9F=0F,=1F=FC!=8C=03HH=02=DA=02(=FF=E1=FF=E2=02=F9=02F=03G=05(=03=FC=02=
HH=02=DA=02(=01d=01=03=03=03=01'=0F=01=01h=08=19=01=90=3D=E0/=D0=05=A0=05=
=A0=05=A0=05=A0=02=D0R=01=01=01BH=1A=11-:LaserWriter 8.0
=01=16=07Courier=C0=01=01=01=19=01=19=06=80=80=01=19A=02+=03=80=04I(=02=01=
=C3=01=11

Next message: J. D.: "Re: TWA 800: THE CAT IS OUT !!"
Previous message: David Musick: "Visual Languages"