Quantum Inspired Computational Intelligence

5.4.2 Compression bound

The compression bound is the result of the approximation of the Kolmogorov theory [110] and, in particular, the minimum description length principle [111]. The compression bound theory [66, 84] states that, for a fixed quality of the learned model, the fewer data from ${\mathcal{D}}_n$ we need for learning the rule, the better will be its generalization performance [64, 173]. In particular, given a set of rules ${\mathcal{S}}^{\mathfrak{R}}$, it is possible to state the following theorem.

Theorem 28

([64, 66])

Let us consider a set of rules ${\mathcal{S}}^{\mathfrak{R}}$. If the criteria ...