This achieved the maximum entropy principle
22 FRONTIERS OF EVOLUTIONARY COMPUTATION
a conditional probability computer would allow to compute all logical inferences, if we identify “from y follows z” by the condition The drawback of this proposal is that it needs units.Thus the computation is exponential in time and space.
| In both papers the assumption is made that all | are independent. This is |
|---|
Definition 2 A probabilistic statement that z is true given y is a conditional probability with “truth” value
| tation. Let | be the number of binary concepts. In addition let a number of |
|---|
But unique P-models are unrealistic. The specification has to set all of the
variables defining the distribution. Consequently, for incomplete specifications the missing information must be added by some automatic completion procedure. This is achieved by the maximum entropy principle. The entropy of a distribution is defined by
Maximum entropy principle: Find the maximal entropy distribution for
which satisfies the given marginals.
Then the maximum entropy solution gives, for instance
| cluster | on the unique path between |
|
|---|
