icluded in this 
arting from a 
rclass, and thus 
the graphical 
  
perattribute 
Tr 
perlink 
  
nsions. 
bed others, thus 
y. Ihe second 
amed embryo, 
yo looks like a 
th no realization; 
IS, composing a 
t a model, which 
each structure 
attributes and/or 
fic phenomenon 
the graphical 
med "GRAPH", 
al relationships 
d "nodes"), arcs 
and areas, as it 
:al themes. 
ype 
  
  
tructures. 
The third set of extensions consist of building new 
components, corresponding to classes of rules, 
classes of processes and classes of neurons. 
Rules are considered as objects of classes which 
have been previously included into a hyperclass 
named "RULE"; thus, they inherit of the attributes 
and links of this hyperclass, namely: -premis 
(boolean expression, or fuzzy expression), - 
consequent if yes (any set of executable statements), 
-consequent in no (likewise), —a preparing part 
named "PROLOG" (exec.statements), -a closing 
part named "EPILOG" (likewise), —an explanation 
which is purely textual, —various coefficients such 
as:  —askability, -editability, — -priority, - 
modifiability, —lauchability, —priority, -reliability, 
etc... Moreover, rules present relationships such as 
"actives" and "inhibits", because a rule may active 
another one which was not a priori activable, or on 
the contrary, an activated rule may inhib another 
rule which is launchable but not compatible with the 
first one. The figure 4 shows the graphical 
representation of the hyperclass RULE. 
askability hibit 
editability Eat S 
modifyability EZ es PROLOG 
  
     
  
lauchability C—4 
priority D— ET PREMISE 
reliability O— 
: en 
OOO FS 
EROR ege 
CONSEQUENT 
enc» €» IF NO 
C» CoD 
2 EPILOG 
explanation 
  
  
  
  
  
  
Fig. 4 : The hyperclass RULE and its components. 
The hyperclass RULE is itself included in the 
hyperclass "FACT", thus solving the problem of the 
so-called "metarules", a rule possibly appearing as a 
fact for another rule. Rules and facts are chained, 
according to the following graphical representation, 
shown of the figure 5. The facts are a.d.t. of any 
type, such as attributes of classes, links between 
classes, classes themselves, as shown on the figure 
5; they may be too objects or anything else. The 
expert system included in the HBDS kernel is not 
composed of a single inference engine, but 
composed of several ones which run forward and 
backward. These engines accept fuzzy facts and 
fuzzy rules. The engines are themselves included in 
the knowledge structure. The whole expert system is 
designed without the obsolete concepts of 
backtracking and pattern-matching (Bouille, 
1984a,b, 1988, 1991b). 
41 
  
class of 
rules 
  
  
d dst ttt ERE 
  
  
Fig. 5 : Graphical representation of the interaction of 
rules and facts. 
Discrete processes are considered as objects of 
classes which have been previously included into a 
hyperclass named "PROCESS"; thus they inherit of 
all the attributes of this hyperclass, namely : -a 
clock, dealing with the time at the scale of a possible 
simulation (nanoseconds as well as centuries or 
b.y.), -state (which may be active, passive, hold, 
terminated, -body, which is the algorithmic part 
comosed of any set of executable statements, 
working on a coroutine mode, —a local sequence 
counter, according to the term firstly introduced in 
the SIMULA 67 programming language. The figure 
6 shows the structure of the Hyperclass 
"PROCESS". 
  
   
  
  
  
  
  
( Active 
STATE Hold 
Passive 
aie 
CLOCK 
8 relire: 9 
PROCESS or — s 
CERT C BODY 
C» C»C» C» | (coroutine) 
CCS NS 
QUE C m C» 
(uu QE C 
elc... 
  
  
  
  
Fig.6 : The hyperclass PROCESS and its 
components. 
Neurons too are considered as objects of classes 
which have been previously included into a 
hyperclass named "NEURON"; thus they inherit of 
the following attributes: —minimum intensity for 
nA AA 
ET eee conne