Ordinal sums of the main classes of fuzzy negations and the natural negations of t-norms, t-conorms and fuzzy implications (Q2302900)

In traditional logic, every statement is either true or false. In real life, people are often not certain about their statements. Zadeh proposed to describe different degrees of certainty by numbers from the interval \([0,1]\); this is one of the main ideas behind \textit{fuzzy logic}. In traditional logic, the truth value of a composite statements obtained by using ``and'', ``or'', and ``not'' is uniquely determined by the truth values of the component statements. In general, this is not true for degrees of certainty. However, since we cannot solicit the expert's degree of confidence in \textit{all} composite statements, we thus need to be able to estimate, e.g., the degree of certainty in \(A\&B\) based on the degrees of confidence \(a\) and \(b\) in \(A\) and \(B\). This estimate \(T(a,b)\) is called an ``and''-operation, or, for historical reason, a \textit{t-norm}. It is known that every t-norm \(T\) that satisfies reasonable properties (such as commutativity and associativity) is an \textit{ordinal sum} of t-norms isomorphic to \(a\cdot b\), \(\min(a,b)\), or \(\max(a+b-1,0)\): namely, the interval \([0,1]\) can be represented as a union of disjoint subintervals on each of which \(T\) is isomorphic to one of the above three ``standard'' t-norms, and \(T(a,b)=\min(a,b)\) for \(a\), \(b\) from different subintervals. A similar representation is known for fuzzy ``or''-operations -- called \textit{t-conorms}. These known results, however, do not cover fuzzy negations \(N(a)\). This gap is being filled by this paper. Specifically, the authors analyze when the -- naturally defined -- ordinal sum of fuzzy negations satisfies different properties such as invertibility, the property that \(N(N(a))=a\) for all \(a\), etc.