Research.

A ransomware-recognition dataset is used to compare different descriptions of execution: system calls, libraries and paths, hardware-counter signals and combinations of these views. Information Concordance and Selector Adjusted Information Concordance measure agreement between explanations, also accounting for feature-selection effects. In this comparison, system calls yield more stable explanations across models; other representations are more sensitive to processing choices. The finding concerns the benchmark studied and encourages joint evaluation of accuracy and the stability of the reasons underlying predictions.

The work was presented at ANNPR in October 2026 and published online in September 2026; AIR UniMI assigns 2027 to the bibliographic edition of the proceedings volume.

The volume reports a project carried out between June 2024 and March 2025, combining training, an exploratory survey with 72 participating municipalities and a prototype for automatically classifying municipal website pages. The responses reveal interest in automating repetitive tasks, data analysis and decision support, together with a demand for transparent models and human oversight. The findings concern the administrations that took part in the survey and offer a basis for designing pilot initiatives accompanied by training.

Hardware counters record aspects of processor activity during program execution. Sample entropy summarises their regularity over time; this information feeds probabilistic graphical models, which describe relationships between variables. Comparison with other classifiers allows the ability to distinguish ransomware to be evaluated, while the structure of the models helps examine the reasons for decisions. The result is an interpretable analytical methodology evaluated on the study’s data; operational deployment requires testing under the intended conditions of use.

The research, conducted within the MEIEC centre in collaboration with AnciLab, examines cybersecurity in Lombardy’s municipalities. The volume’s synopsis describes widespread awareness of the importance of security, alongside a need to strengthen practical protection capabilities, including against common attacks. The contribution helps explain the gap between concern about the issue and operational readiness, providing a basis for supporting local authorities.

The contribution uses CNEOS data on asteroid hazard classification to compare Markov random fields, Random Forests and Support Vector Machines. Physical knowledge of the problem makes it possible to examine both classification quality and interpretability. The results described in the abstract identify graphical models as a solution combining readable explanations with accuracy appropriate to the case studied. The subject is classification within the dataset considered; this record is not an impact forecast.

The authors use astronomical data and data constructed from physical laws, also adding noisy variables. They train several algorithms, including neural networks, graphical models, Random Forests and statistical classifiers, and examine their behaviour through interpretation techniques. Comparison with the law that generated the data allows the learned dependencies to be assessed against the phenomenon. The contribution thus offers a way to evaluate interpretability before addressing situations in which physical relationships are less well understood.

The study combines electron energy-loss spectroscopy measurements with theoretical calculations. This comparison associates certain signals with vibrations of carbon–hydrogen bonds at the edges, distinguishing features of their structure. Ribbon ends also produce specific signals whose intensity can indicate average length. The relationship between spectrum and structure therefore provides a tool for characterising materials that are difficult to describe at atomic scale through visible geometry alone.

The authors compare nanoribbons with the same aromatic core but different side-chain positions. The steric bulk of these groups changes the structural distortion and, consequently, the bandgap: the energy separation between the electronic states under consideration. Theoretical predictions are compared with absorption and photoluminescence excitation measurements. The work shows how shape control, together with chemical composition, can tune the response of these systems.

Raman spectroscopy is compared with ab initio calculations to understand which nanoribbon features correspond to the measured signals. The work examines the effects of width, edge geometry and chemical functionalisation. Certain low-frequency vibrations are particularly sensitive to the edges; the dependence of one spectral band on the excitation is also studied. These correspondences help identify nanoribbon structure through spectroscopic measurement.

The theoretical work compares pairs of identical nanoflakes at different relative displacements, as well as pairs differing in size or chemical groups. The calculations predict a redshift and broadening of the lowest excitations, extending the absorption range without strongly increasing intensity. Appropriately functionalised combinations also exhibit excitations involving charge transfer between fragments. These modelling results help explain the role of interlayer interactions.

The study derives formulae for two topological indices, the eccentric connectivity index and the Wiener index, in lattices made of nanotubes connected through schwarzitic junctions. It also examines the indices as the number of units becomes very large. The Wiener index offers theoretical indications of the lattices’ relative stability: the connection network thus becomes a compact way to reason about structural differences.

The chapter studies negatively curved carbon surfaces, or schwarzites, comparing topological arguments with ab initio calculations. It examines the approximations under which connectivity helps describe vibrational spectra and how defects change those spectra. For certain planar schwarzites, it also compares calculated stability with indications from the Wiener index. The relationship between a network model and a physical description is discussed alongside the assumptions that delimit its validity.

The two proposed forms are C28 clathrates and differ in how the structure repeats within the crystal: one is body-centred orthorhombic and the other simple cubic. Electronic properties and certain vibrations are calculated in detail for the former. The authors also discuss how units resembling small diamond fragments could assemble into the latter. These are hypothetical structures investigated with theoretical tools, illustrating a method for interpreting complex crystal lattices.

The chapter compares carbon-nanoribbon models with ordinary closure and with Möbius-type closure. It calculates topological invariants and studies their behaviour as the system grows, examining compactness and connections between distant parts of the network. These quantities are discussed in relation to the stability of the structures considered. The work concerns mathematical analysis of models in which twisting changes the lattice topology.

“Coloured” topology assigns parts of the structure properties derived from electronegativity and chemical hardness, which describe aspects of a molecule’s electronic response. The chapter develops these indices within conceptual density functional theory and applies them to polycyclic aromatic hydrocarbons. Comparing the topological description with energy-related quantities helps investigate structure–reactivity relationships. Here, “colour” means the chemical information assigned to the network.

The review considers disordered, sponge-like schwarzites and periodic structures, relating topology to stability and possible formation pathways. It also discusses electronic properties, electron–vibration interactions and atomic-dynamics simulations that follow the network’s evolution. The contribution provides an overview of the physics of these structures in the context of the studies available at the time of publication.