IBM researcher R.M. Friedberg
Ref the posting earlier today - or was it yesterday . . .
I was intrigued by the reference to topics sounding quite a bit like genetic programming from so long ago.
The Freidberg papers are referenced in Koza bibliography (Genetic Programming Vol 1)
... and both papers - part I and part II are published in Evolutionary Computation, The Fossil Record - ISBN 0-7803-3481-7. pp 147 ...
Found on the Web . . .
Later, but still early on in the history of
computing, R.M. Friedberg invented a system,
called Herman, designed to learn
through random variations.4,5 He defined a
virtual assembler language space within
which these variations took place. He
tested the variations against a given task,
and the system preferred instructions that,
in a given location in the program, showed
better performance over less performing
instructions. Herman and its variants
showed limited success in adapting to the
prescribed task. Besides certain deficits in
the variation and selection loop, the most
limiting factor was computer time. Evaluating
programs one after the other to measure
their performance was very slow.
Authors Title Report Number Report Date B. Dunham , R. M. Friedberg (Columbia University) , J. H. North A Case History in Design Automation RC 3151 November 13, 1970
A Case History in Design Automation. November 1970.
Dunham, B. Friedberg, R. M. North, J. H.
20p. When people speculate as to the future capabilities of computers, they are not so much concerned with size of memory,
speed of access, and so forth. Rather, they are uncertain as to how well the computer can be expected to manage various
recalcitrant problems for which no obvious ready made modes of solution exist. How well for example, can a computer diagnose
medically, or analyze the contents of a photograph? The concern is more than academic, since the need for effective
computation is often great.
A Placement/ Wiring Method for the FET Static Logic Chip.
Proceedings of the Design Automation Seminar on Placement and Placement/ Wiring Techniques. Yorktown Heights, N.Y.
December 1970. p38-42.
Dunham, B. North, J. H.
This talk describes methods developed for doing circuit placement and wire routing on FET Static Logic chips. The nearly
completed programs, interfaced into CDS, are being written by Ken Sanderson and Ved Gupta of SDD and by Herb Roedig of
the Service Bureau Corporation. Two factors are paramount in determining wirability of these chips. First, the number of wires
passing through any circuit cell must not exceed the number of available wiring channels for that cell. Secondly, wiring
difficulties increase rapidly with the number of crossunders from one column to another. The early stages of this program are
designed to deal primarily with these two considerations.
For a more general discussion of the authors' problem solving philosophy, see A Case History in Design Automation, by B
Dunham, RM Friedberg, and JH North, IBM Research Report RC-3151 (November 1970). The nature of interchange
programs is studied in Design by Natural Selection, by B Dunham, D Fridshal, R Fridshal, and JH North, IBM Research Report
RC-476 (June 1961).
Ian Brackenbury FRSA, FBCS, Distinguished Engineer, IBM Corp,
Software Group - South Pacific region. "Teamwork over Networks"
Cell:+64-21-799-065 Fax:+64-6-879-9064 Tel:+64-6-879-9065