Thus, (R←,R→)∘(X←,X→)∘(R←,R→)=(R←,R→), which completes the proof.
2. Dynamic fuzzy matrix HRL algorithm
Learning problems can be decomposed in accordance with their level so as to construct a bottom-up hierarchy. This is more conducive to qualitative and quantitative analysis. The traditional absolute value is used to show the uncertainty that does not accord with the subjective judgment and the characteristics of the object, so dynamic fuzzy theory is used to analyse the learning problem. We use the dynamic fuzzy matrix to represent the relationship rate between upper data and related data in this level. Assuming that upper element C and this level’s elements c1, c2, . . ., cn are related, then the dynamic fuzzy matrix (R←,R→) has the following form:
where (r←ij,r→ij)∈[0,1]×[←,→](i,j≤n) represents element ci and cj according to the relationship rate of element C and elements ci and cj contain the membership rates of the dynamic fuzzy relationship. According to the relationship between elements c1, c2, . . ., cn and upper element C, we construct the dynamic fuzzy matrix of each level as
Definition 6.9 For a dynamic fuzzy number M in dynamic fuzzy matrix (R←,R→) if and only if its membership rate (μ←M,μ→M)(x←,x→)∈[0,1]×[←,→] satisfies
then M is a triangular dynamic fuzzy number, written as M=((l←,l→),(m←,m→),(n→,n→)).Here,(l←,l→)≤(m←,m→)≤(n←,n→), and all M belong to [0, 1] × [←, →], where (l←,l→)and(n←,n→) denote the lower and upper bounds of M, respectively.
Consider two groups of triangular dynamic fuzzy number M1=((l←1,l→1),(m←1,m→1),(n←1,n→1))andM2=((l↼2,l⇀2),(m←2,m→2),(n←2,n→2)). Mathematical operations on these numbers are performed as follows: