CYBERNETI CS

AND
SYSTEMS
vol ume2
ROBERTTRAPPL
edi t or
Aust r i an Soci et y f or
Cyber net i c St udi es
Pr ef ace
Gener al Syst ems Met hodol ogy
Chai r per sons: C. J osl yn, USA, andG. J . Kl i r , USA
OnaNewSyst em
Theor y
andI t s NewPar adi gms
EZampa
AnI nt er nal Appr oacht o Syst ems Descr i pt i on andI t s Speci f i cat i on f or DEDSRel at i on Model l i ng
G. J uhas
Towar ds a Di f f er ent i abl e Bool eanDynami cs
F
. Bazso, E. Labos
The
Gener al i zedConcept of Hi er ar chy
M
. Kal ai dj i eva
Evol ut i onar y Hi er ar chi es of Ener gy Pr ocessi ngi n Nat ur e
P. Wmi war t er ,
C
. Cempel
Cyber net i c Communi cat i on andCont r ol wi t h Di scr et e El ement Model s
R. A. Gagl i ano
Adapt edSyst ems ver sus Syst ems i n Equi l i br i um
EPar r a- Lung
Obser vat i on,
Memor i zat i onand
Causal
Oper at or s
R. Val l ee
CONTENTS
Di scur si ve Syst ems Theor y:
J oi ni ng Har dandSof t Syst ems i n Anal yt i c Tool s
S. Leven, R. Smi t h
Bui l di ngI nt el l i gent Syst ems f r omGener al Syst ems Theor y
A. Casel l es
Use of
Li ngui st i c
Scal es
i nQual i met r yExami nat i on of Obj ect s
I . ETbm, O. VKr asko
NewDevel opment s i n Mat hemat i cal Syst ems Theor y
Chai r per sons: YRav, Fr ance, andFPi chl er , Aust r i a
I nt el l i gent Syst ems Combi ni ngReact i ve and Lear ni ng Capabi l i t i es
WJ acak, RBuchber ger , SSt i f t er
NewSyst em- Theor et i cal Met hods i nDesi gn of Di scr et e Aut omat i on Syst ems
E. Schni eder
Upper and
Lower Condi t i onal Possi bi l i t i es andNecessi t i es I nducedby a Mul t i val uedMappi ng
E. Tsi por kova, B. de Baet s, E. Ker r e
APet r i Net Appr oacht o Anal ysi s of Compl ex Di scr et e Event Syst ems andt o
Synt hesi s
of
t hei r Cont r ol
ECapkovi c
Model i ngof Di scr et e Dynamcal Syst ems by Neur al Net wor ks and
Genet i c Al gor i t hms
S. Dr ei sei t l
AnAut omat a- Theor yAppr oach t o Desi gn aFeedbackCont r ol Lawf or Aut onomous Agent s
I . Si er ocki
Emer gence of Space- Ti me St r uct ur es
f r om
Di f f usi ve Chaos
i nPear l - Ver hul st andLot ka- Vol t er r a Cel l ul ar
Aut omat a
D. M. Duboi s
Wal sh Hamoni zabl e
Pr ocesses i n Li near Syst emTheor y
Y. Endow
1 I ntroducti on
Evol uti onary
Hi erarchi es of Energy
processi ng i n Nature
Peter
Wi ni warter
Bordal i er
I nsti tute
41270 Boursay
France
f ax+33 5480 1937,
e- mai l : whri wartet@bordal i eri nsti tute. corn
Abstract
Types
of energy transf ormati on processes
are
proposed as a basi s
f or determi ni ng hi erarchi cal
l evel s i n nature
.
A
si mpl e and general model f or
energy
transf ormati on processors wi th l i mi ted
transf ormati on
capaci ty l eadi ng to the death of
the
processor i s proposed
. Thi s model can expl ai n the
stati sti cal
Pareto- Zi pf - Frechet- Wei bul l behavi or
whi ch i s observed f or symptoms of
energy
transf ormati on
at al l hi erarchi cal l evel s i n
the
uni verse
. Basedon the f ormal equi val ence between
trophi c webs of energy transf ormi ng
( Bi rth&Death)
processors andneural networks ageneral
f ramework
f or model i ng memory, adaptati on/ l earmi ng
and
evol uti on i s proposed
.
When we descri be the evol uti on
of the uni verse, f rom i ts
ori gi ns

up to soci o- economi c and
techni cal systems of
today' s manki nd, the concept of energy /
i nf ormati on
transf ormati on i s the onl y f eature
common to al l l evel s of
organi zati on anddescri pti on.
However, on the vari ous l evel s
we
f i nd
di f f erent types of
energy / i nf ormati on transf ormati on
processes, whi ch
determi ne the structure of systems
andcontrol the system' s
behavi or onthat l evel .
For exampl e
nucl ear energy i s governi ng the
transf ormati on processes goi ngonwi thi n massi ve stars l i ke
the sun,
whi l e geo- thermal energy i s i nvol vedand
shapes the
evol uti onof the earth- crust.
Howmany
l evel s of di f f erent energy / i nf ormati on
transf ormati on
canone observe, andhowcanthese l evel s be
orderedi na
coherent way?
The goal of thi s
paper i s to def i ne cl ear andexpl i ci t order
cri teri a f or a coherent
di sti ncti on of hi erarchi cal l evel s i n
nature.
Furthermore
we wi l l propose a general model f or a basi c
energy transf ormati on processor,
whi ch can be appl i ed to
any type of transf ormed
energy.
The behavi or of such an energy
transf ormati on processor,
descri bi ng i ts traj ectory f rom
bi rth to death, wi l l be l ookedat
f rom
an i nf ormati onal
poi nt
of
vi ewi n terms of bi nary state
transi ti on.
Thi s f ormal equi val ence wi th a
f ormal neuronwi l l al l owi s
to draw an anal ogy between trophi c webs of energy
transf ormati on processors andneural networks .
Based on thi s anal ogy we wi l l . outl i ne a conceptual
f ramework f or model l i ng adapti ve and evol uti onary
processes l eadi ngto the emergence of hi erarchi cal l evel s .
Czesl awCempel
Poznan Uni versi ty of Technol ogy
ul . Pi otrowo 3,
60- 965 Poznan, Pol and
f ax +48 6178 2307,
e- mai l : cempel @pozn1v. tup. edu. pl
2 Hi erarchi cal structure
of nature
Theconcept of
hi erarchy i s usedby many authors to descri be
andto
anal yze compl ex systems i n an ordered
f ashi on. The
order
cri teri a f or sl i ci ng nature i nto
hi erarchi cal or
organi zati onal l evel s are not al ways
cl earl y stated by the
di f f erent
authors, and i n most cases
they are onl y i mpl i ci t .
I n the f ol l owi ng we wi l l
anal yze some exampl es of
hi erarchi cal orderi ng.
Bi t we do not pretend to gi ve an
exhausti ve overvi ewon
l i terature coveri ng thi s topi c al ready
publ i shed. Summari zi ng one
can say, that there are as many
hi erarchi es of nature
proposed as there are authors .
We are convi nced,
that i t i s necessary to i ntroduce some
expl i ci t
order
cri teri a i n order to arri ve at a coherent approach
whencutti ngup
nature i nto hi erarchi cal l evel s .
Further on i n
thi s paper we wi l l show, that the type of
energy and
i nf ormati onprocessi ng canbe used as a common
order
cri teri on f or establ i shi ng " l evel s

i n al l types of
systems rangi ng
f rom astrophysi cal to soci al systems.
2. 1 Boul di ng :
control - ori ented hi erarchy
Ahi erarchy
of l evel s was f ormul ated by Boul di ng, [ 1956] , as
ci ted by
Bl anchard and Fabrycky [ 1990] , and i s summari zed
bel ow. The
i mpl i ci t order cri teri on usedhere by Boul di ng to
deri ve
l evel s 1 to 4 i s the type of control
mechani sm
governi ng the
system, whi l e the l evel s 5 to 9 are ordered
accordi ng to
i nf ormati on processi ng types wi thi n certai n
enti ti es .
Tabl e 1. Boul di ng' s hi erarchy of nature
2. 2 Mi l l er
: energy transf ormati on and
i nf ormati on
transf ormati on
I n hi s
magnumopus
e
Li vi ng Systems
,

Mi l l er [ 1978]
postul ates
7 hi erarchi cal l evel s as bel ow.
I t
i s the great meri t of Mi l l er to have
poi nted out that,
despi te the apparent di f f erences of
structures and processes
on
the di f f erent l evel s, sel f - si mi l ar
system concepts
can
be usedf or thedescri pti onof each l evel
.
Level Short descri pti on
9 unknowabl e no descri pti on
8 soci al cr
ani zati on
7 human sel f - consci ousness, symbol processi ng
6 ani mal
mobi l i ty, tel eol ogi cal behavi or, sel f -
awareness
5 pl ant geneti c- soci etal structure I
4 cel l
3 thermostat
rneti cSystems
2 cl ockwork
si tu l e dynami c stems
1 f ramework
stati c structures of the uni verse
Tabl e 2. Mi l l ers hi erarchy of l i vi ng Systems
Mi l l ers '

i ci t order cri teri on f or establ i shi ng
l evel s i s
modul ar decomposi ti on. An organ i s composed of
cel l s
.
An
organi smi s composed of organs etc. . . .
I t i s i mportant to note, that on al l l evel s
Mi l l er cl earl y
di sti ngui shes between
energy
transf ormati on
and
i nf ormati on
transf ormati on processes.
On
each
of
the seven l evel s, a system
can
be descri bed by
the
same
set
of 10 energy transf ormati on subsystems and 9
i nf ormati on transf ormati on subsystems .
The energy

transf ormati on subsystems are
Reproducer, Boundary, I ngestor, Di stri butor, Converter,
Producer, Storage, Extruder, Motor and Supporter.
The I nf ormati on processi ng subsystems are
I nput transducer, I nternal transducer, Channel and net,
Decoder, Associ ator, Memory, Deci der, Encoder,
and
Output
transducer .
Besi des of the great conceptual breakthrough wepoi nt out
two maj or l i mi tati ons of Mi l l er' s general theory
"

i ts l i mi tati on to e l i vi ng

systems ( i . e. the theory i s
not appl i ed to geo- chemi cal and astrophysi cal
systems)
and
"

the theory does not gi ve muchhi nts on howand why the
di f f erent l evel s have emerged.
2. 3 J antsch : phases i n the co- evol uti on of
macro- and mi cro processes
I t i s
the meri t of J antsch
[ 19801 to
have proposed
a
general
and
systemati c
approach
to
the
evol uti on of the uni verse
f rom
cosmi c symmetry brakes
up
to compl ex soci eti es.
The order cri teri on of
l aunch' s
hi erarchy are evol uti onary
phases emergi ng through pri mordi al " symmetry brakes

occuri ng wi thi n l ower l evel s .
Tabl e 3. J antsch' s evol uti onary hi erarchy of man
i n the Uni verse
Accordi ng to
J antsch

the

" dri vi ng

f orce

f or the
emergence of new
l evel s
shown expl i ci tl y i s the
co-
evol uti on
of mi croscopi c and macroscopi c systems.
On eachl evel mi croscopi c
enti ti es are descri bed
i n terms
o
f
mi cro- structures, mi cro- processes
and l ocal
control
mechani sms.
The mi croscopi c enti ti es evol ve wi thi n a
macroscopi c
" envi ronment* whi ch i tsel f i s descri bed i n terms
of
macro- structure, macro- processes
and
gl obal control
mechani sms
( see
tabl e 3. )
For exampl e on l evel 2,
the l evel of
prebi oti c co-
evol uti on, we observetwo
compl ementary trends
"

mi cro- evol uti on consi sts of
prokaryote
structures
undergoi ng
sel f - reproduci ng hypercycl e
processes, bei ng
control l ed
by genes.
macro- evol uti on i s
taki ng pl ace on Gai a ( the
macroscopi c
structure) undergoi ng bi o- energeti c
processes control l ed through synthesi s.
The above menti oned mi croscopi c systems co- evol ve
wi thi n the macroscopi c systems unti l a symmetry break
occurs , whi ch l eads to the emergenceof the next l evel 3, the
l evel
of
geneti c co- evol uti on:
" mi cro- evol uti on
consi sts
of
eukari ote structures
undergoi ng endosymbi oti c processes control l ed by
genotypes.
"

macro- evol uti on takes
pl ace i n ecosystems
undergoi ng
speci al i zati on
processes control l ed by chemi cal and
thermal
sel f stabi l i zati on of the bi osphere.
2. 4 Sal the : ecol ogi cal
and geneal ogi cal
hi erarchi es
Accordi ng to the ecol ogi st Sal the [ 19851, an evol vi ng
hi erarchi cal systems can be descri bed both i n terms of
ecol ogi cal or organi zati onal hi erarchy ( snapshot of
energy
transf ormati on hi erarchy) and of geneal ogi cal
hi erarchy
( i nf ormati on transf ormati on, ti me cross- secti on
of parent- chi l d rel ati ons) .
Tabl e 4. Sal the' s ecol ogi cal and geneal ogi cal
hi erarchi es of nature
The i mpl i ci t order cri teri on f or Sal the' s ecol ogi cal
hi erarchy i s spati al decomposi ti on at a gi ven
moment i n ti me. I t descri bes the di f f erent organi zati onal
l evel s wi thi n the energy transf ormati on processes of today' s
earth. Sal the' s geneal ogi cal hi erarchy on the other
hand i s based on the order cri teri on heredi ty ; i . e.
transmi ssi on of i nf ormati on i n evol uti onary ti me.
2
. 5
Organi zati onal l evel s of the central
nervous system
To showanother exampl e of hi erarchi cal l evel s i n
compl ex
adapti ve systems we menti on the organi zati onal
l evel s of the
central nervous system[ Sej nowski and Churchl and, 19921.
I n the exampl e bel ow, the order cri teri on to construct
the
hi erarchy i s expl i ci tl y stated! I t i s the spati al
di mensi on
whi ch corresponds to an observabl e physi cal
structure cal l ed
organi zati onal
l evel . Note that on each organi zati onal
l evel
energy / i nf ormati on transf ormati on processes
take pl ace,
Level
Short descri l i on
7
ranati onal stem
composed
of two or more soci eti es
6 soci ety l arge l i vi ng systemcomposed of
organi zati ons
5 o ani zati on systems wi th mul ti - echel on
deci des
4
rou
set of
si ngl e
o
ani sms
cal l ed members
3
ani sm mi ts usual bi ol o i cal sense
2
organ subsystems of organi sms, ani mal s and
ants
1
cel l l i vi ng organi zed protopl asm
Ecol ogi cal
hi erarchy
Geneal ogi cal
hi erarchy
7 total bi o here
6 surf ace of the earth 6 hi stori cal bi ota
5 bi o- geographi cal
regi onal
5 monophyl eti c l i neage
4 ecosystem 4 dame( set of repl i cati ng
i ndi vi dual s
3 anon 3 i ntegrated genotype phenotype
2
organsmi c 2
gene
I mol ecul ar 1 s
ci es
Level of Level of Phase of
mi cro- evol uti on macro- evoi uti on o .
evol uti on
7
sel f - i mage
man- i n- uni verse
i ntegrati ve
6 sel f - ref l ecti ve i deas, pl ans re- creati ve
mentati on ( soci o-
cul tural di mensi on
r -
5 ref l exi ve mentati on groups,
f ami l i es
i magi nati ve
( gestal t
on
4 organi sm/ organi srni c soci eti es wi th di vi si on epi geneti c
mentati on
of
l abor
3 cel l s
eukari ots
ecosystems
generi c
2 organel l es Gai a prebi oti c
roka sea
1 di ssi pati ve structures pri mal ocean chemi cal
( i ntracel l ul ar
processes)
0 l i near i rreversi bl e i norgani c chem. - physi cal
thennodynami cs physi cal
macrod cs
whi ch
ar e
descr i bed
( or not ) by a var i et y of di f f er ent
t heor et i cal
t ool s.
I t woul d be possi bl e t o ext end t he scope of t he above
hi er ar chy
zoomi ng- out t o t he di mensi ons of gr oups,
soci et i es and. supr anat i onal ent i t i es of Mi l l er , and zoomi ng-
i n t o t he
di mensi ons of at oms, nucl eons
and
subat omi c
par t i cl es.
2
. 6 Cempel
:
ener gy hi er ar chy
Accor di ng t o di ct i onar y def i ni t i on
" ener gy i s one of t he
char act er i st i c
scal ar quant i t i es descr i bi ng t he st at e of t he
syst em, bei ng t he equi val ent
t o t he possi bl e wor k,
associ at ed wi t h pr oper
f or mof mat t er and f ul f i l l i ng t he
t r ansf or mat i on and t he conser vat i on
pr i nci pl es .

Based on
t hi s def i ni t i on one of t he aut hor s has
pr oposed t he ener gy-
mat t er hi er ar chy as bel ow.
st r ong
i nt er act i on
pot ent i al
up
war ds
Tabl e 5. t he or gani zat i onal l evel s of t he cent r al
ner vous syst emand t hei r di mensi ons
t ype or ener gy
Y
unknown
>
ment al a
st r ong a
Tabl e
6. Cempel ' s ener gy hi er ar chy
weak
i nt er act i on
pot ent i al
downwar ds
Her e t he hi er ar chi cal l evel s

ar e or der ed accor di ng t o
t r ansf or mat i on pot ent i al
( const r uct i on and dest r uct i on
pot ent i al ) i n
associ at i on wi t h mat t er - ener gy ent i t i es
Al ower ener gy

l evel has t he
pot ent i al t o t r ansf or m
( dest r oy i n t he ext r eme
case) al l upper ener gy l evel s ; e. g.
t her mal ener gy can
dest r oy mechani cal , chemi cal ,
bi ochemi cal et c. ent i t i es .
Ahi gher ener gy
l evel has t he pot ent i al t o t r ansf or m
ent i t i es on l ower
mat er - ener gy l evel s onl y by weak
i nt ensi t y
and over l ong per i od of t i me; e. g. chemi cal
cor r osi on can
t r ansf or m t he shape and per f or mance
of
mechani cal ent i t i es .
Ahi gher
l evel can emer ge onl y when t he I owa l evel i s
compl et e,
so one can say t hat a hi gher l evel a f eeds
upon al ower l evel .
The
l evel s as or der ed above cor r espond al so t o t he
hi st or i cal
emer gence of par t i cul ar f or ms of mat t er - ener gy
dur i ng t he evol ut i on of t he
uni ver se.
The passage f r omone
l evel t o anot her , however r emai ns
unexpl ai ned.
2. 7 Wi ni war t er : evol ut i onar y
hi er ar chy of
ener gy / i nf or mat i on t r ansf or mat i on
Ref i ni ng t he above hi er ar chy,
and
t aki ng i nt o
consi der at i on
t he
maj or
aspect s of Mi l l er s and J ant sch' s
appr oaches we
pr opose ahi er ar chy of nat ur e as bel ow.
Her e
t he or der i ng
of t he l evel s i s based on t he
sequence of
emer gence of new
t ypes of ener gy /
i nf or mat i on
t r ansf or mat i on pr ocesses dur i ng t he
evol ut i on of
t he
uni ver se.
I n t he hi er ar chy bel oweach l evel i s char act er i zed
by a
macr oscopi c syst em composed of mi cr oscopi c
ener gy /
i nf or mat i on pr ocessor s whi chi nt er act under
t he i nf l uence of
t hei r own char act er i st i c f i el d. Thi s f i el d i s gener at ed by
t he
t ot al i t y of al l t he pr ocessor s i nt er act i ng wi t hi n t he syst em
.
For exampl e t he bi o- chemi cal f i el d consi st s of and r egul at es
t he i nt er act i on of mi cr oscopi c mol ecul es i n
t he macr oscopi c
ocean
as
a syst em
.
Tabl e 7.
Wi r uwar t er ' s s hi er ar chy of ener gy /
i nf or mat i on
syst ems, pr ocessor t ypes, and
associ at ed f i el ds of i nt er act i on
I nt er i evel
coupl i ng and t r ansf or mat i on :
Cl i mbi ng up t he
hi er ar chy we obser ve di mi ni shi ng
coupl i ng/ t r ansf or mat i on ener gi es bet ween
pr ocessor s
on each consecut i ve hi gher l evel . e. g. bi ndi ng ener gi es
bet weennucl eons ar e or der s of magni t udes hi gher t han
bi ndi ng ener gi es bet ween chemi cal el ement s et c.
I nt er i evel count i ng and t r ansf or mat i on
a) a
hi gher l evel can onl y emer ge and evol ve wi t hi n al l
pr evi ous l ower l evel s ; usual l y when t he f i el d i n t he l ower
l evel exceeds a cer t ai n t hr eshol d val ue ( bi f ur cat i on,
symmet r y
br eak) .
b) l ower l evel s can
dest r oy hi gher
l evel s, but
hi gher
l evel s can onl y const r ai n
l ower
l evel s . ( see Cempel ' s
ener gy hi er ar chy)
.
c) l ower
l evel
pr ocessor s
f eed i nt o ( or ar e
eat en
by)
hi gher l evel pr ocessor s .
Or gani zat i onal
l evel
char act er i st i c
di mensi on
8 cent r al
ner vous
syst em
I m
7 st ems 10 an
6 maps
I cm
5 net wor ks 1 mm
4 neur ons 100
3 mi cr oci r cui t s 10 pi n
1 mol ecul es
1
A
Macr oscopi c
Mi cr oscopi c
Newemer gi ng
Syst ems Ener gy/ Ener gy/ I nf or mat i on
I nf or mat i on Fi el ds
pr ocessor
t ype
15 syst ems
sci ence
f er al
met a- t heor et i cal
communi t y
t r ansdi sci pl i nar y
( met er - sci ent i st )
14
sci ent i f i c communt y
wr i t t en sci on
i s
&
symbol i c ( sci enr i st ) t echnol ogi cal
t heor i es,
r adi ms
13
al phabet i cal
wr i t t en symbol i c cor i umt r adi t i on /
communi t y
( l i t er at e i ndi vi dual ) hi st or y
12
l anguage anal
symbol i c or al t r adi t i on / cul t ur e
communi t y communi cat i on
( ver bal
communi cat i ng
i ndi vi dual )
11
cor nmuni t y
symbol i c soci al / ethol ogi cal
communi cat i an
( i ndi vi dual )
10
eco- ayst em
mul t i cel l ul ar
cent r al ner vous ( f ur l i ng)
ani sm
9 bi r o cel l eukar i ot e
genet i c/ exual
8
bi o- spher e / G ai a sel f - r epl i cat i ng
t r ophi c ( bi o- ener gy)
met abol i c uni cel l ul ar r
r oka or e)
7 at mos her e/ ocean macr o- mol ecul e bi o- chemi cal
6
ear t h- cr ust t ect oni c
mi cr o- st r uct ur e
geo- mechani cal
l at es
pl anet s physi co- chemi cal geo- t her mal
el ement
4 gal axy,
st ar s
nucl eus
el ect r omagnet i c + weak
nucl ear
3 pr ot o- gal axy hadr on st r ong
nucl ear
+
el ect r o- weak
2
pr ot o- uni ver se
quar k gr avi t at i on +gr and
uni f i ed
1 Pl anck- er a of bi g gr avi t on
super - gr and- uni f i ed
ban
I t
f ol l ows f rom t he above, t hat t he order
cri t eri on f or
hi erarchi cal l evel s used by
us
i s t he evol ut i onary
sequence of
syst ems wi t h new qual i t i es of energy
and i nf ormat i on
t ransf ormat i on . By new
qual i t i es we underst and
qual i t i es,
whi ch cannot be expl ai ned
i n t erms of t he qual i t i es
of t he
l ower l evel s .
3 Energy/
I nf ormat i on t ransf ormat i on
processors
( Bi rt h&Deat h processors)
I n recent papers

[ Compel , 1987
; Cempel , 1992; Cempel ,
1992
; Wi ni wart er, 1983;
and Wi ni wart er and Cempel , 1992]
we
have present ed and el aborat ed a t heory
of general i zed
energy and i nf ormat i on processors.
The model ( see f i g. 1

bel ow) i s
charact eri zed by energy
i nput ,
a l i mi t ed i nt ernal capaci t y
of t ransf ormat i on, upgraded
and degraded energy out put s
and some i mport ant
st at i st i cal
f eat ures .
Fi gure 1.
El ement ary Bi rt h&Deat h processor, t he
bui l di ng bl ock
of t he sel f - si mi l ar hi erarchi cal energy
t ransf ormat i on process i n operat i ng syst ems.
3. 1
Sympt omst at i st i cs over t he
l i f e- t i me of
a
si ngl e processor ( t i me cross- sect i on)
Observi ng one processor over i t s ent i re
l i f e- t i me f rombi rt h
( i nt ernal t i me 0=0) t o deat h ( breakdown t i me
0=0, ) , we
f i nd i t s asympt ot i c behavi or of t ransf ormed
energy. Hence
when
t he breakdown of t he processor i s
approached, t he
t ransf ormed
energy goes t o i nf i ni t y ( see f i gure 2a) .
But i n most cases we can nei t her di rect l y observe
nor
measure t hi s t ransf ormed energy, but onl y some
covari ant
sympt oms by means of some f i l t er Tas i l l ust rat ed i n
f i gure 1.
I f
we observe a gi ven sympt omof one
processor at regul ar
t i me
i nt erval s over i t s l i f e span we can make a cumul at i ve
probabi l i t y of sympt om occurrence. I t can be
shown
[ Compel , 1992] - anal yt i cal l y and experi ment al l y,
t hat such
a cumul at i ve
probabi l i t y i s of t he Zi pf - Paret o- Frechet -
Wei bul l t ype ( cal l ed
ZPFWdi st ri but i on hereaf t er) [ Paret o,
1897; Zi pf , 1948
; Wi ni wart er, 1983 and Cempel , 1992] .
Moreover, t here i s st ri ct equi val ence bet ween t he t ype
of
sympt oml i f e
curve observed and t he observed sympt om
probabi l i t y di st ri but i on
.
3. 2 Sympt om
st at i st i cs f or a popul at i on of
energy processors
( snapshot )
Observi ng a popul at i on
of energy t ransf ormat i on
processors,
whi ch bel ong t o t he same operat i ng
met s-
syst emand
whi ch are i n di f f erent advancement of processor
l i f e, we can
make a cumul at i ve probabi l i t y di st ri but i on of
sympt om
occurrence wi t hi n t he popul at i on .
I t can be shown al so
[ Compel , 1992; Compel and Nat ke,
1992],
t hat t he di st ri but i on
i s agai n a l ong- t ai l ed
di st ri but i on of t he ZPFWt ype as
above.
I n t he case
of processors whi ch are i dent i cal ( or si mi l ar i n
a f i rst
approxi mat i on) t he paramet ers def i ni ng t he
ZPFW
di st ri but i on are i dent i cal i n bot h
domai ns; i . e . ; i n t he
sympt om
l i f e domai n of a
si ngl e processor and t he sympt om
behavi or
i n t he popul at i on domai n. Theref ore
we can make
i nf erences about t he evol ut i on of a si ngl e
processor based
on paramet ers deri ved f rommeasurement s
of a popul at i on of
processors,
and vi ce versa.
4
" General i zed Li f e Sympt oms
of energy
processi ng syst ems
Empi ri cal evi dence f or PZFW
sympt omdi st ri but i ons
exi st s
on al most al l known
l evel s of energy t ransf ormi ng
syst ems
as seen bel owi n
Tabl e 8.
I n our t erms,
t he t abl e gi ves us an
overvi ew of syst ems
wi t h processor popul at i ons
f or whi ch PZFW sympt om
di st ri but i ons
are empi ri cal l y
observed [ Wi ni wart er, 1983 ;
Wi ni waner, 1985;
Wi ni l wart er,
1987
;
Wi ni wart er,
1995a;
andZi pf , 1948] . As can be
seen f romt he cross- ref erences t o
t he hi erarchi cal l evel s of t abl e
7, t he observat i ons of PZFW
regul ari t i es,

whi ch

we

cal l ed

general i zed

Li f e
Sympt oms
,

are f ound on
al l emergent l evel s of energy
t ransf ormat i on i n t he uni verse.
Tabl e 8. Empi ri cal l y
observed
general i zed Li f e
Sympt oms : Paret o- Zi pf
( PZFW) di st ri but i ons
f or
popul at i ons of
Bi rt h&Deat h processors
i n operat i ng
syst ems. For
a hi st ori cal overvi ew and
det ai l ed sources
see [ Wi ni waner and
Compel , 1992] .
I n manycases t hese
empi ri cal
observat i ons are f ormul at ed
even as general l aws
. e. g. t he l aw of
Gut enberg- Ri cht er
st at es, t hat
eart hquake i nt ensi t i es
f ol l ow a
Paret o
cress- ref .
operat i ng
Bi rt h&Deat h observed
r
Syst em
Processors sympt oms
t ab . 7 of one
t ransf ormat i on
14 comput er
syst em mi cro- processors
produced f i l es
14 t echni cal
machi nes of t he
produced
product i on syst em
same di ssi pat i on
one
13 economi c
syst em busi ness f i rms produced
/ count ry
cu rof i t
13 soci o- economi c
human i ndi vi dual s produced
i ncome
s t om
12
soci o- economi c
ci t i es
produced
syst em
popul at i ons
12
cul t ural syst em human brai ns
produced words
l anguage
l t bi o- soci al
syst em ani mal s of same
produced
/ col on
s ci es
l at i ons
10
ecosyst em bi ol ogi cal speci es
produced
l ot i ons
9 bi ot ope
bi ol ogi cal genera produced
speci es
8 bi osphere
bi ol ogi cal
f ami l i es
produced genera
7 ocean
mi cro- part i cl es produced
part i cl es
6 eart h
t ect oni c pl at es
produced
eart hquakes
5 sol ar
syst em, chemi cal
produced
pl anet s
compounds
can
nds
4 gal axy /
uni verse st ars produced
nucl ei /
t hem. l ement s
di st r i but i on
wi t h t he exponent y=
1, 5
f or al l r egi ons of t he
gl obe.
Accor di ng
t o our model , and i n accor dance wi t h t he
empi r i cal
obser vat i on st at ed i n t abl e S above,
we
can
post ul at e, t hat t he obser vat i on of sympt oms
f ol l owi ng
a
pZFW
di st r i but i on
may
be an
i ndi cat or f or ener gy
t r ansf or mat i on
syst ems of pr ocessor s under goi ng

i

d
deat h.
As a
consequence we suppose, t hat t he ener gy pr ocessi ng
wi t h
l i mi t ed t r ansf or mat i on capaci t y, as i n our model , i s a
f eat ur e common t o al l l evel s of evol ut i on.
Fur t her we can obser ve on al l l evel s t he or gani zat i on of
ener gy t r ansf or mat i on pr ocessor s i n t r ophi c hi er ar chi es or
web- l i ke st r uct ur es ( see f i gur e 3) . Byt r ophi c hi er ar chi es or
5
Co- t r ansf or mat i on
of
Ener gyand I nf or mat i on
So f ar we have consi der ed onl y t he ener gy t r ansf or mat i on
aspect of pr ocessor s dur i ng t hei r l i f et i me ( f r om bi r t h t o
deat h) . As i l l ust r at ed i n f i gur e 2a) bel ow, t hi s gener al i zed l i f e
pr ocess at a
gi ven
l evel can
be
descr i bed by a cont i nuos
f unct i on wi t h an asympt ot i c behavi or ( si ngul ar i t y) at t he
br eakdown of t he pr ocessor .
The same l i f e- pr ocess obser ved by
t he met asyst em
can
be
descr i bed as a bi nar y i nf or mat i on pr ocess .
As
l ong as t he
pr ocessor i s i n oper at i ng st at e ( l i f e) t he met asyst em
obser ves no change
or
0- st at e.
At
t he
br eakdown of t he
pr ocessor ( deat h)
t he
met asyst ems
obser ves
a
bi nar y st at e
t r ansi t i on t o 1- st at e.
Thi s
i s
i l l ust r at ed
i n
f i gur e 2b)
bel ow.
The i mpor t ant i dea put f or war d her e
i s
as f ol l ows
cont i nuos ener gy- t r ansf or mat i on
pr ocesses on one
l evel ar e equi val ent
t o
di scr et e

bi nar y

I nf or mat i on
st at e t r ansi t i ons as obser ved on t he met al evel .
I

b) amet a- vi ew. LEVELn a
t
t
t
t
oper at i ng
di scr et e l e I nf or mat i on t r ans-
f or mat i on;
bi nar y
st at e
t r ansi t i on
at deat h
dead
i i ,
Mebsyst em
t i me
t
a)
" I nt er nal
vi ew. LEVEL
n_
cont i nuous
ener gyt r ansf or mat l on
f r ombi r t h t o deat h
Syst emt i me B
Fi gur e
2. The equi val ence of a) cont i nuous ener gy
pr ocessi ng and b) bi nar y
i nf or mat i on t r ansmi ssi on t o
t he
met al evel .
5. 1
Neur al anal ogy expl ai ni ng ener gy /
i nf or mat i on
equi val ence
Based on
f i gur e
2
wecan dr awan anal ogy bet ween
an ener gy
t r ansf or mat i on
pr ocessor ( Bi r t h&Deat h pr ocessor ) and
a
f or mal neur on.
Bot h can be consi der ed as t hr eshol d

aut omat a wi t h
bi nar y st at e changes as expl ai ned bel ow
t i me 0, ) t he pr ocessor br eaks down,
whi ch i s per cei ved
byt he met asyst emas a bi nar yst at e
t r ansi t i on.
A f or mal neur on i nt egr at es wei ght ed
i nput pul ses,
compar es t hemwi t h a t hr eshol d val ue
and f i r es a bi nar y
si gnal t o t he. next l evel , when t hi s t hr eshol d
i s
r eached
.
Hence a t r ophi c webof ener gy t r ansf or mat i on pr ocessor s
evel
pr ocessor s
f eedi ng i nt o hi gh_
based on t he above anal ogy -
can

consi der ed

s a deni al
net wor k.
As i t was post ul at ed i n [ Wi ni war t er , 19956] t he web can be
descr i bed as a t r ophi c hi er ar chyof Bi r t h&Deat h pr ocessor s,
when l ooked
at
f r om an ener gy poi nt of vi ew and as a
net wor k
of
bi nar y t hr eshol d aut omat a, when l ooked at
f r om
an i nf or mat i onal
poi nt
of
vi ew.
A
net wor k
or web of bi nar y
t hr eshol d aut omat a i s al so cal l ed neur al net wor k.
An ener gy t r ansf or mat i on pr ocessor sums
up i nt er nal l y
st or ed t r ansf or mat i on ener gy up t o i t s
capaci t y
Edb
( see
f i gur e 1) . When t hi s capaci t y i s r eached, ( at
br eakdown-
5. 2 Memor y
Fi gur e.
3

Si mpl i f i ed r epr esent at i on of an oper at i ng
ener gy t r ansf or mat i on web
f or mal l y equi val ent t o a
neur al net wor k. I nt er pr et i ng
a
node
of
t he
gr aph as
ener gy pr ocessor , t he l i nk t o t he upper l evel
i s
const i t ut ed vi a upgr aded ener gyout put of t he
pr ocessor ,
whi l e t hedowngr aded ener gy out put
f r omt he pr ocessor
t o t he web const i t ues t he f eedback ( cal l ed
backpr opagat i on i n cer t ai n neur al net wor ks)
.
Based on t he neur al net wor k anal ogy a web of ener gy
pr ocessor s shoul d
.
r eveal
memor y f eat ur es as set of
oper at i ng/ non- oper at i ng st at es di st r i but ed over
t he ent i r e
web
"

I n
t he
case
of a neur al
net wor k, t he memor y consi st s of
t he t hr eshol d val ues, t he i nput
wei ght s
and
t he t opol ogy
of act i ve connect i ons
of
al l neur ons
i n
t he
net wor k
.
"

I n t he case of an ener gy / i nf or mat i on

t r ansf or mat i on
web, t he memor y consi st s
. of
t he capaci t i es
Edb or
br eakdown t i mes 0, of t he pr ocessor s and
t he t opol ogy
of
act i ve ( t r ophi c) connect i ons .
5. 3
Lear ni ng / adapt at i on
Agai n,
based on t he above anal ogy, t he l ear ni ng /
adapt at i on of ener gy t r ansf or mat i on webs or gani zed
hi er ar chi cal l y as i l l ust r at ed
i n f i gur e
3 coul d be model ed l i ke
t he l ear ni ng m f eedf or war d neur al net wor ks wi t h
backpr opagat i on.
t r ophi c webs we under st and f ood chai ns or ener gy
t r ansf or mat i on chai ns i n gener al , wher e l ower
l evel
pr ocessor s
f eed i nt o
( or ar e eat en by)
hi gher l evel
pr ocessor s
.
Froman energy
poi nt
of vi ew each
processor t ransf orms
i nput energy. Upgraded energy i s
f ed
f orward
t o t he next
hi gher l evel ( f roml ef t t o ri ght ) . Downgraded energy i s f ed
back t o t he l ower
l evel s ( f romri ght t o l ef t ) . Thi s f eedback
( backpropagat i on) modi f i es al so t he processor paramet ers of
a gi ven l evel f or
t he next operat i on/ t ransf ormat i on cycl e.
Froman i nf ormat i onal poi nt
of vi ew t he web of : energy
( Bi rt h&Deat h) processors can
be
vi ewed
as aneural net work.
Learni ng/ adapt at i on coul d be
underst ood here as repeat ed
processi ng cycl es of t he energy
t ransf ormat i on web. At each
t rophi c
cycl e t he f eedback modi f i es t he paramet ers of t he
next
generat i on
of newl y born processors i n t he web. When
t he f eedback and t he processor
repl acement pol i cy i s random,
t he webwoul d show chaot i c behavi or . However, when
t he
f eedback i s correl at ed t o t he overal l perf ormance
of t he web
( e. g. t he gradi ent descent met hod f or error
mi ni mi zat i on) , t he
web wi l l reveal
goal ori ent ed behavi or and l earni ng
.
5. 4 Evol ut i on of energy / i nf ormat i on
t ransf ormat i on
webs
Based on t he f act , t hat
cert ai n cl asses of neural net works
can sol ve opt i mi zat i on
probl ems ( based on t he t hermal
anneal i ng al gori t hm
f or exampl e) t he

di rect ed evol ut i on
of energy / i nf ormat i on
t ransf ormat i on webs coul d be
model ed i n a si mi l ar way. General a cool i ng
of t he web
dri ves i t t owards mi ni ma wi t hi n a
pot ent i al l andscape
[ Wi ri wart er,
19956] . Short
l ocal heat i ng, f ol l owed by
subsequent cool i ng can avoi d t rappi ng of t he web
i n l ocal
mi ni ma
of subopt i mi zat i on.
6 Concl usi on
1 . When speaki ng
about hi erarchi cal or organi zat i onal
l evel s i n nat ure,
we need mdef i ne
expl i ci t l y
t he
underl yi ng
order cri t eri on. Up t o now, t hi s was negl ect ed
by
most aut hors.
2. We propose
a consi st ent pri nci pl e f or est abl i shi ng
hi erarchi cal
l evel s i n nat ure, based
on energy
t ransf ormat i on process t ypes and t he
sequence of t hei r
occurrence duri ng t he evol ut i onof t he uni verse.
3. We present
a si mpl e model f or an energy / i nf ormat i on
processor, whi ch
i s general enough t o be appl i cabl e on
every hi erarchi cal l evel
i ndependent l y of t he nat ure
of
t ransf ormed energy.
These processors

are charact eri zed
by a l i mi t ed
t ransf ormat i on capaci t y

l eadi ng
t o
bi rt h&deat h behavi or .
4. The st at i st i cal behavi or of such
processors observed as
v
general i zed
l i f e sympt oms shows a
l ongt ai l ed
behavi or whi ch
can be descri bed mat hemat i cal l y by
Paret o- Zi pf - Frechet - Wei bul l ( PZFW) di st ri but i ons .
5. PZFWdi st ri but i ons
are empi ri cal l y observed on vi rt ual l y
al l hi erarchi cal l evel s of
energy t ransf ormat i on i n
nat ure.
6. From
an energy t ransf ormat i on poi nt of vi ew,
a
Bi rt h&Deat h processor cont i nuousl y t ransf orms
energy
up t o i t s breakdown
( deat h) . Froman i nf ormat i onal poi nt
of vi ewhowever
- as observed by t he met asyst ent
of t he
processor - a Bi rt h&Deat h processor
i s equi val ent t o a
bi nary t hreshol d aut omat on ( f ormal neuron)
wi t h t he t wo
possi bl e
st at es : 0- st at e ( operat i ng) or 1- st at e ( dead) .
7. Organi zat i on
of such processors i n a t rophi c web make
t hemf ormal l y equi val ent t o a
neural net work.
8. Hence
al l t he f eat ures of neural net works, l i ke memory,
adapt at i on/ l eami ng
and opt i mi zat i on can be l ooked at i n
an anal og way i n t rophi c
webs of energy / i nf ormat i on
t ransf ormat i on processors
.
The
above
approach may be t he basi s f or a general
t heory
expl ai ni ng
sel f - organi zat i on, sel f - l earni ng and evol ut i on
i n
nat ure.
Ref erences
[ Bl anchard and Fabrycky, 19901
B. S
. Bl anchard and
W. J . Fabrycky,
Syst ems Engi neeri ng and
Anal ysi s, p
. 9,
Prent i ce Hal l I nt ernat i onal , London,
1990.
[ Boul di ng, 1956] K. Boul di ng, General Syst ems
Theory : The
skel et on of Sci ence,
Management Sci ence, Apri l
1956.
[ Cempel , 19921 C. Cempel ,
Condi t i on evol ut i on
of
machi nes and i t s assessment
f rompassi ve
di agnost i c
experi ment . Mechani cal Syst ems
andSi gnal Processi ng,
5, 1991.
[ Cempel , 19871 C. Cempel , The wear
process and t he
machi nery vi brat i on - aTri boVi broAcust i cal
model , Bul l .
Pol i sh Academy of Sci ences, Techn
. Sc. ,
35,
N( 7- 8) :
347- 363, 1987.
[ Cempel , 1992] C. Cempel , Damage I ni t i at i on and
Evol ut i on
i n operat i ng Mechani cal Syst ems. Bul l . Pol i sh
Academy
of Sci ences, 40( 3) : 1- 14, 1992.
[ Cempel and Nat ke, 19921 C. Cempel and H
. G. Nat ke,
Damage Evol ut i on andDi agnosi s i n Operat i ng
Syst ems.
I n H. GNat ke et al . ( ed. ) , Proceedi ngs of
Saf et y
Eval uat i on
Based Syst em I dent i f i cat i on Approaches
Rel at edt oTi me- Vari ant and
Nonl i near St ruct ures, pp. 44-
61, Vi eweg, Braunschwei g, 1992.
( J ant sch, 1980] E. J ant sch, The
Sel f - organi zi ng Uni verse,
p. 224- 240, Pergamon Press,
Oxf ord, 1980.
[ Mi l l er, 1978] J . G. Mi l l er, Li vi ng
Syst ems, p. 1028- 1029,
McGrawHi l l , NewYork, 1978.
[ Paret o, 1897] V. Paret o, Cours
d' I conomi e pol i t i que. ( 1897)
Repri nt ed as a vol ume of Oeuvres
Compl t es, Droz,
Geneva, 1965.
[ Sal t he, 1985] S. N. Sal t he, Evol vi ng Hi erarchi cal
Syst ems,
p. 185, Col umbi a Uni versi t y Press, New
York, 1985.
[ Sej nowski and Churchl and, 19921 T. J . Sej nowski
and P. S .
Churchl and, Si l i con Brai ns. Byt e 10: 137- 146
, 1992.
[ Wi ni wart er, 1983] P.
Wi ni wart er,
The Genesi s Model
. Part B
Frequency di st ri but i ons of el ement s i n sel f - organi zed
syst ems. Spec. Sci ence and Technol ogy
6: 103- 112,
1983.
[ Wi ni wart er, 1985] P. Wi ni wart er, I so- dynami cs
of
popul at i on- si ze di st ri but i ons i n hi erarchi cal syst ems
. I n
B. Banat hy ( ed. )
Syst em
I nqui ri ng, Proceedi ngs
of t he
Soci et y f or General Syst ems Research, voLI pp. 87- 95,
I nt ersyst ems Publ i cat i ons, Seasi de Cal i f . , 1985.
[ Wi ni wart er, 1987] P. Wi ni wart er, Si ze- di st ri but i ons and
hi erarchi cal
cont rol i n bi ol ogi cal syst ems. I n B. Lck
( ed. ) , Bi ol ogi e
Thori que
pp. 289- 299, Edi t i ons
du
CNRS, Pari s, 1987.
[ Wi ni wart er and Cempel , . 1992] P. Wi ni wart er and C.
Cempel , Li f e Sympt oms : t he behavi or of open syst ems
wi t h l i mi t ed energy di ssi pat i on capaci t y and
evol ut i on.
Syst ems Research, 9( 4) : 9- 34, 1992.
[ Wi ni wart er, 1995a] P. Wi ni wart er and C. Cempel ,
Bi rt h&Deat h Processors 1 : t i me- energy equi val ence
i n
Syst emLi f e and t he f ract al di mensi on of Syst emTune i n
energy- t ransf ormi ng Hi erarchi es . I n B. Bergval l -
Darebom( ed. ) Proceedi ngs of t he I nt ernat i onal Soci et y
f or t he Syst ems Sci ence, pp. 73- 85, I nt ernat i onal Soci et y
f or t he Syst ems Sci ences, Loui svi l l e, Kent ucky, 1995.
[ Wi ni wart er, 19956] P. Wi ni wart er, Bi rt h&Deat h Processors
2
: model l i ng Syst emmemory, l earni ng and evol ut i on
based on t he f ormal equi val ence of energy t ransf ormi ng
hi erarchi es and neural net works . I n B. Bergval l - Darebom
( ed. ) Proceedi ngs
of t he
I nt ernat i onal Soci et y f or t he
Syst ems
Sci ence, pp. 86- 100,
I nt ernat i onal Soci et y f or
t he Syst ems Sci ences, Loui svi l l e, Kent ucky, 1995.
[ Zi pf , 1948] G. K. Zi pf , HumanBehavi or and t he
Pri nci pl e of
l east Ef f ort , Addi son Wesl ey, Cambri dge Mass . ,
1948.