Distributed machine learning has become increasingly vital as data sources continue to expand geographically. Traditional ensemble methods such as AdaBoost demonstrate impressive predictive capabilities but often require frequent synchronization across nodes, resulting in significant communication overhead. This paper introduces a novel paradigm of infrequent synchronization in which nodes perform multiple rounds of local AdaBoost before exchanging partial or complete model updates. The potential advantages include reduced communication costs, the ability to handle intermittent connectivity, and competitive accuracy compared to fully synchronized approaches. A real-world use case in the trucking industry is presented to demonstrate the feasibility and value of this new approach. The paper concludes by outlining future directions and the expected impact on communication-efficient distributed learning.

Medical image segmentation is a critical task in healthcare, particularly for disease detection and proper treatment planning. Deep learning models achieve high performance in medical image analysis. This paper presents the effectiveness of the new PDE-UNet architecture, inspired by the applications of partial differential equations (PDEs) in neural networks, to enhance medical image segmentation performance. Experiments were conducted on brain tumor MRI images from the BraTS2020 dataset and compared with the traditional UNet architecture.

This paper provides a comprehensive review of our previously obtained results on exactly solvable lattice models, with a primary focus on the Abelian sandpile model, dimer model, loop-erased random walks and their connections to the enumeration of spanning trees.

Speaker diarization is the process of partitioning an audio recording into segments corresponding to individual speakers. In this paper, we present a robust speaker diarization system and describe its architecture. We focus on discussing the key components necessary for building a strong diarization system, such as voice activity detection (VAD), speaker embedding, and clustering. Our system emerged as the winner in the Voxceleb Speaker Recognition Challenge (VoxSRC) 2023, a widely recognized competition for evaluating speaker diarization systems.

Dominant areas of computer science and computation systems are intensively linked to the hypercube-related studies and interpretations. This article presents some transformations and analytics for some example algorithms and Boolean domain problems. Our focus is on the methodology of complexity evaluation and integration of several types of postulations concerning special hypercube structures. Our primary goal is to demonstrate the usual formulas and analytics in this area, giving the necessary set of common formulas often used for complexity estimations and approximations. The basic example under considered is the Boolean minimization problem, in terms of the average complexity of the so-called reduced disjunctive normal form (also referred to as complete, prime irredundant, or Blake canonical form). In fact, combinatorial counterparts of the disjunctive normal form complexities are investigated in terms of sets of their maximal intervals. The results obtained compose the basis of logical separation classification algorithmic technology of pattern recognition. In fact, these considerations are not only general tools of minimization investigations of Boolean functions, but they also prove useful structures, models, and analytics for constraint logic programming, machine learning, decision policy optimization and other domains of computer science.

It has formerly been proved that there is an exponential speed-up in the number of lines of the quantified propositional sequent calculus over substitution Frege systems when considering proofs as trees. This paper shows that a linear proof of any quantifierfree tautology in quantified propositional sequent calculus can be transformed into a linear proof of the same tautology in a substitution Frege systems with no more than polynomially increasing proof lines and size.

The paper deals with the issue of determining the degree of fuzzy regularity of a crisp string. It is assumed that the concept of fuzzy regularity is formalized by a pattern given as a finite automaton with fuzzy properties of alphabet characters on transitions. Proceeding from this, we replace the problem of determining the degree of fuzzy regularity of a crisp string with the problem of determining the degree of belonging of such a string to the language of the corresponding automaton and propose an effective method for solving it using the dynamic programming approach. The solution to the considered problem makes it possible to fuzzify the set of strings in a given alphabet based on a pattern defining fuzzy regularity. This work is a continuation of the author’s previous works related to finding occurrences of a fuzzy pattern in the text. It may have applications in the field of pattern recognition, data clustering, bio-informatics, etc.

Sets of word tuples, accepted by multitape finite automata and a metric space for languages accepted by these automata, are considered. These languages are represented using the same notation as the known notation of regular expressions for languages accepted by one-tape automata. The only difference is the interpretation of the ”concatenation” operation in the notation. An algorithm is proposed for calculating the introduced distance between regular languages accepted by multitape finite automata.

Let D be a 2-strongly connected directed graph of order p ≥ 3. Suppose that d(x) ≥ p for every vertex x ∈ V (D) \ {x0}, where x0 is a vertex of D. In this paper, we show that if D is Hamiltonian or d(x0) > 2(p − 1)/5, then D contains a Hamiltonian path, in which the initial vertex dominates the terminal vertex.

To learn complex interactions between predicates and accurately estimate the cardinality of an SQL query, we develop a novel framework based on recursive tree-structured neural networks, which take into account the natural properties of logical expressions: compositionality and n-ary associativity. The proposed architecture is an extension of MSCN (multi-set convolutional network) for queries containing both conjunction and disjunction operators. The goal is to represent an arbitrary logical expression in a continuous vector space by combining sub-expression vectors according to the operator type. We compared the proposed approach with the histogram-based approach on the real-world dataset and showed that our approach significantly outperforms histograms with a large margin.