This article addresses lenges in enterprise financial management, including difficulties in processing multi-source data, limited adaptability to dynamic environments, and a lack of systematic integration in the decision-making process. To tackle these issues, a new intelligent optimization framework, named genetic algorithm-fuzzy logic-Transformer (GA-FL-Transformer), is proposed. First, the framework employs the Transformer architecture to achieve unified encoding and feature fusion across multiple sources of financial data, high-dimensional features with strong discriminative power. Subsequently, an attention-weight-guided co-evolutionary mechanism integrating genetic algorithm (GA) and fuzzy logic (FL) is designed. This mechanism incorporates the features and attention weights into chromosome encoding, fitness function formulation, and genetic operations, thereby enabling dynamic optimization of fuzzy rules and membership functions. Finally, an intelligent optimization framework that integrates perception, optimization, and decision-making is constructed, achieving closed-loop optimization from data to decision-making via a bidirectional flow mechanism and supporting continuous learning and system-wide self-adjustment. Results on financial datasets from Compustat and CRSP show that the proposed method outperforms competing models in financial optimization. Ablation experiments further validate the contributions of the Transformer-based feature extraction, genetic algorithm optimization, and fuzzy reasoning mechanism to the system’s performance. This study provides a crucial theoretical foundation for enterprises to construct intelligent financial decision-making systems.
We prove a kind of a pumping lemma for languages accepted by one-register alternating finite-memory automata. As a corollary, we obtain that the set of lengths of words in such languages is semi-linear.
We introduce a novel framework for simulating finite automata using representation-theoretic semidirect products and Fourier modules, achieving more efficient Transformer-based implementations.
Brent L. Winner, Theodore S. Switzer, Sean F. Keenan
et al.
Abstract Recognized limitations of fishery-dependent data under rapidly changing management regimes have resulted in significant effort during recent years to improve the availability of fishery-independent data in the southeastern United States. These fishery-independent surveys target numerous species and habitats with various sampling methods, including the use of trawls, longlines, gill nets, traps, and visual surveys. Although passively fished hooked gear (e.g., longlines) are often used to assess the abundance and life history of managed reef fishes, such types of gear are often limited by the habitats they can fish effectively and are species selective. To address these shortcomings, we developed and implemented an actively fished approach to provide fishery-independent data: the repetitive timed-drop hooked-gear method (RTD method). Despite the high degree of standardization applied to the RTD method, important questions remain as to whether active fishing imparts strong angler variability that may reduce the utility of survey data. Accordingly, we analyzed data from 2014 to 2018 to evaluate potential angler bias and how angler-associated variability compares to other factors often thought to be important predictors of reef fish abundance and community structure. During this study, 962 stations were sampled, representing a variety of artificial and natural reef habitats. In total, 5,770 fish were caught, representing 92 taxa. Sampling was conducted by 103 unique anglers, including 42 commercial or charter fishers and 61 scientists. Results from both population- and assemblage-level analyses found that most of the variability in the catch could be explained by hook size, habitat, water depth, and year. Angler type was rarely correlated with reef fish abundance or assemblages. Our analyses suggest that the RTD method is effective in gathering fishery-independent abundance and life history data for reef fishes in the eastern Gulf of Mexico and that the resulting data are not strongly biased by an angler effect.
In this abstract we propose a framework for explaining violations of safety properties in Software Defined Networks, using counterfactual causal reasoning.
We consider nondeterministic higher-order recursion schemes as recognizers of languages of finite words or finite trees. We propose a type system that allows to solve the simultaneous-unboundedness problem (SUP) for schemes, which asks, given a set of letters A and a scheme G, whether it is the case that for every number n the scheme accepts a word (a tree) in which every letter from A appears at least n times. Using this type system we prove that SUP is (m-1)-EXPTIME-complete for word-recognizing schemes of order m, and m-EXPTIME-complete for tree-recognizing schemes of order m. Moreover, we establish the reflection property for SUP: out of an input scheme G one can create its enhanced version that recognizes the same language but is aware of the answer to SUP.
Suppose that some polynomial $f$ with rational coefficients takes only natural values at natural numbers, i.e., $L=\{f(n)\mid n\in \mathbb N\}\subset\mathbb N$. We show that the base-$q$ representation of $L$ is a context-free language if and only if $f$ is linear, answering a question of Shallit. The proof is based on a new criterion for context-freeness, which is a combination of the Interchange lemma and a generalization of the Pumping lemma.
We introduce and study cellular automata whose cell spaces are left-homogeneous spaces. Examples of left-homogeneous spaces are spheres, Euclidean spaces, as well as hyperbolic spaces acted on by isometries; uniform tilings acted on by symmetries; vertex-transitive graphs, in particular, Cayley graphs, acted on by automorphisms; groups acting on themselves by multiplication; and integer lattices acted on by translations. For such automata and spaces, we prove, in particular, generalisations of topological and uniform variants of the Curtis-Hedlund-Lyndon theorem, of the Tarski-Følner theorem, and of the Garden-of-Eden theorem on the full shift and certain subshifts. Moreover, we introduce signal machines that can handle accumulations of events and using such machines we present a time-optimal quasi-solution of the firing mob synchronisation problem on finite and connected graphs.
This text is an extended version of the chapter 'Automata and rational expressions' in the AutoMathA Handbook that will appear soon, published by the European Science Foundation and edited by JeanEricPin.
Algorithms on grammars/transducers with context-free derivations: hypergraph reachability, shortest path, and inside-outside pruning of 'relatively useless' arcs that are unused by any near-shortest paths.