Abstract:We investigate correlations in information carriers, e.g. texts and pieces of music, which are represented by strings of letters. For information carrying strings generated by one source (i.e. a novel or a piece of music) we find correlations on many length scales. The word distribution, the higher order entropies and the transinformation are calculated. The analogy to strings generated through symbolic dynamics by nonlinear systems in critical states is discussed.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is the computability problem of higher - order type functionals, especially how to define and understand the concept of "feasible" computation of these higher - order functions within polynomial time. Specifically, the article explores the definition and properties of basic feasible functionals, and verifies the rationality of these definitions through logical and Turing machine models.
### Detailed Interpretation
1. **Background and Motivation**
- The paper first points out that in lower - order types (such as ordinary functions \( \mathbb{N}^k \to \mathbb{N} \)), polynomial - time computation on Turing machines has been widely accepted as the correct formalization of feasible computation. However, for higher - order types (such as functionals \( (\mathbb{N}^\mathbb{N})^k \times \mathbb{N}^l \to \mathbb{N} \)), there is no similar consensus.
- K. Mehlhorn proposed a possible paradigm, extending Cobham's definition of feasible functions to second - order functionals. Subsequently, Townsend, Cook, and Kapron studied this class of functionals and introduced the concept of "basic feasible functionals".
2. **Main Contributions**
- **Recursion - Theoretic Properties**: The paper shows that basic feasible functionals have recursion - theoretic properties, which naturally generalize the corresponding properties of feasible functions. This provides further evidence for choosing the above - mentioned feasibility concept.
- **Improved Machine Representation**: The paper improves Kapron and Cook's results on machine representation, providing a new method to study the properties of basic feasible functionals using logical tools.
- **Extracting Programs from Proofs**: The author provides an example of how to extract a feasible "program" from a mathematical proof using non - feasible functionals.
3. **Technical Means**
- **Weak Fragment of Second - Order Arithmetic**: The author introduces a weak fragment of second - order arithmetic, where the second - order variables range over \( \mathbb{N}^\mathbb{N} \), and shows that this form can effectively characterize basic feasible functionals.
- **Logical Application**: Through logical tools, the author proves the definability of basic feasible functionals in certain formal theories, thereby verifying their rationality and effectiveness.
4. **Conclusion**
- The main result of the paper is that if a functional can be defined by basic feasible functionals using polynomial - bounded recursion, then it is itself a basic feasible functional. This conclusion is proved by combining logical and Turing machine model methods.
### Formula Presentation
- **Polynomial - Bounded Recursion**
Suppose the functional \( F(f, \vec{x}, y) \) satisfies the following conditions:
\[
F(f, \vec{x}, 0) = G(f, \vec{x})
\]
\[
F(f, \vec{x}, y) = H(f, \vec{x}, F(f, \vec{x}, \left\lfloor \frac{y}{2} \right\rfloor), y)
\]
\[
|F(f, \vec{x}, y)| \leq Q(|f|, |\vec{x}|, |y|)
\]
where \( Q \) is a second - order polynomial, and \( G \) and \( H \) are basic feasible functionals.
- **Second - Order Polynomial**
Let \( P(|f|, |\vec{x}|) \) and \( Q(|f|, |\vec{x}|) \) be two second - order polynomials, and \( G(f, \vec{x}) \) and \( H(f, \vec{x}, y, z) \) be basic feasible functionals. Suppose the functional \( F^*(f, \vec{x}, y) \) satisfies:
\[
F^*(f, \vec{x}, 0) = G(f, \vec{x})