arXiv subjects
Publications and source records attributed to Vanja Stojanović.
How many sides does a tire have: two, three, or only one? The question is a joke only until we ask what the word ``side'' is supposed to measure. This article follows three reasonable answers into one small piece of topology: a loop on a tire can have no endpoints and still be a border of a chosen region. The point is not to find the official number of sides, but to see why the count changes when the model changes.
The Euler characteristic transform (ECT) is a computable topological signature for shapes embedded in Euclidean space. Existing ECT stability results control changes in the transform through deterministic geometric perturbation bounds, but many data-generating mechanisms produce perturbations that are random and unbounded in the worst case. In such settings a uniform deterministic bound may be unavailable or too pessimistic. This paper formulates a probabilistic stability-transfer principle for one-dimensional ECT stability: a deterministic stability estimate is combined with a tail bound for the induced geometric error to obtain a high-probability ECT error bound. We instantiate this principle using the stability theorem of Marsh and Beers for embeddings of finite one-dimensional CW complexes. The resulting consequences cover ambient displacement fields conditionally on admissibility and supremum-tail estimates, and finite landmark perturbations with reconstruction. For landmark perturbations, bounded, sub-Gaussian, sub-exponential, and polynomial-tail noise assumptions give explicit high-probability ECT stability bounds. The framework isolates a geometric stochastic component that arises in shape analysis pipelines across biological morphology, medical imaging, molecular structure analysis, and remote sensing.
The study of mutual visibility has traditionally focused on undirected graphs, asking for the maximum number of vertices that can communicate via shortest paths without intermediate interference from other set members. In this paper, we extend this concept to directed graphs, establishing fundamental results for several graph classes. We prove that for Directed Acyclic Graphs (DAGs), the mutual-visibility number $μ(D)$ is always 1, and for directed cycles of length $n\ge3$, it is strictly 2. In contrast, we demonstrate that tournaments can support arbitrarily large mutual-visibility sets; specifically, using properties of Paley tournaments, we show that $μ(T)$ grows linearly with the size of the tournament. On the algorithmic side, we show that while verifying a candidate set is polynomial-time solvable ($O(|S|(|V|+|A|))$), the problem of determining $μ(D)$ is NP-hard for general digraphs. We also analyze the impact of strong bridges and strongly connected components on the upper bounds of $μ(D)$.
The NP-complete mutual-visibility (MV) problem currently lacks empirical analysis on its practical behaviour despite theoretical studies. This paper addresses this gap by implementing and evaluating three distinct algorithms -- a direct random heuristic, a hypergraph-based approximation, and a genetic algorithm -- on diverse synthetic graph datasets, including those with analytically known $μ(G)$ values and general graph models. Our results demonstrate that for smaller graphs, the algorithms consistently achieve MV set sizes aligning with theoretical bounds. However, for larger instances, achieved solution sizes notably diverge from theoretical limits; this, combined with the absence of tight bounds, complicates absolute quality assessment. Nevertheless, validation on known optimal graphs showed the Genetic Algorithm and other heuristics empirically performing best among tested methods.
We investigate the cryptanalysis of affine ciphers using a hybrid neural network architecture that combines modular arithmetic-aware and statistical feature-based learning. Inspired by recent advances in interpretable neural networks for modular arithmetic and neural cryptanalysis of classical ciphers, our approach integrates a modular branch that processes raw ciphertext sequences and a statistical branch that leverages letter frequency features. Experiments on datasets derived from natural English text demonstrate that the hybrid model attains high key recovery accuracy for short and moderate ciphertexts, outperforming purely statistical approaches for the affine cipher. However, performance degrades for very long ciphertexts, highlighting challenges in model generalization.