Victor Erokhin

Existing challenges in modern AI architectures, including the von Neumann bottleneck and high energy demand, have led to increased research efforts towards the development of novel computing paradigms based on memristive technologies. This comprehensive review summarizes algorithmic approaches for training memristive neural networks (MNNs), including Memristive Spiking Neural Networks (MSNNs), integrating solutions implemented on hardware platforms with those developed in software. In contrast to existing reviews that consider algorithmic aspects within the broader context of hardware implementation, this review provides an algorithm-centric perspective, starting from training algorithms and analyzing different approaches to their adaptation to memristive hardware. The survey covers traditional machine learning algorithms such as backpropagation and its enhancements toward hardware-friendly solutions, as well as gradient-free methods, with particular emphasis on algorithms that address device non-idealities. We additionally provide an overview of bio-inspired learning rules, including Spike-Timing Dependent Plasticity (STDP) and reward-modulated STDP, which can be mimicked using memristive devices, along with reinforcement learning algorithms. Special attention is given to biologically plausible models of spiking neuron-astrocyte networks, and their direct implementation in memristive hardware, which significantly increase the efficiency of self-learning algorithms. Finally, this review presents an algorithm-centric perspective on open questions and challenges toward the large-scale implementation of memristive devices in AI systems.

   Objectives. The work set out to develop and substantiate a methodology for synthesizing high-performance and simultaneously verified implementations of elliptic curve cryptosystems. The challenge of ensuring mathematical correctness and engineering security of the code was overcome by integrating machine-checkable proofs with automated inspection.    Methods. In order to carry out formal verification within the Coq interactive environment ⁶ , the study employs ring ⁷ and nsatz8 tactics to automate algebraic proofs together with custom fold-based synthesis methods. For low-level security analysis, static range analysis and deductive verification are carried out within Frama-C ⁹ . The efficiency of the resulting solutions is evaluated through a comparative analysis of arithmetic complexity and execution time.    Results. The proposed approach is implemented by generating executable C code alongside a formal proof of its conformance to a simple mathematical specification to ensure the absence of algorithmic errors. Despite the rigor of the formal methods, the achieved performance approaches that of the best hand-optimized implementations. A key achievement consists in the development of a two-layer architecture for verified finite field arithmetic, which permits the extensive reuse of proofs across different curves and parameters. As part of the research, an explicit multiplication formula for the Curve25519 ¹⁰ curve in a 10-limb representation with weights 2 ^⌈25.5i⌉ was synthesized and formally verified for the first time. This formula is identical to the classic high-performance implementation by Bernstein and Schwabe.    Conclusions. An end-to-end Coq → Frama-C pipeline was successfully built and tested, providing an end-to-end correctness guarantee: from the mathematical specification of elliptic curves to optimized C code verified for overflows, array index out-of-bounds error, and timing variability. The feasibility of the proposed approach is confirmed by the final performance of the synthesized code, which falls within 7 % of the best manually optimized implementations on the target architecture.

Organic memristive devices are promising components for neuromorphic systems. Although based on solution-processable materials, their fabrication often involves complex, resource-intensive processes. Here, we report the fabrication of organic memristive devices using aerosol jet printing to deposit both the active channel based on proprietary polyaniline-based bioink and PEDOT:PSS electrodes. Polymers printing has been carried out both on rigid and flexible substrates, the latter with the aim of demonstrating a flexible device not subjected to films delamination upon bending. By optimizing printing parameters, we achieved devices exhibiting high ON/OFF current ratios exceeding 100 and rapid switching dynamics, with performance comparable on glass and Kapton supports. Morphological and electrical characterizations revealed that channel thickness and uniformity critically influence resistive switching behavior. These findings demonstrate that aerosol jet printing enables scalable, low-material-consumption production of flexible organic memristive devices suitable for neuromorphic applications, potentially facilitating their integration into complex, energy-efficient bio-inspired circuits.

This article presents a unified benchmark of energy consumption per synaptic event across five classes of neuro-computational substrates: graphics processing units, neural processing units, field-programmable gate arrays, digital spiking processors, and in-memory/memristive devices, with biological synapses used as reference bands. The main contribution is methodological as well as empirical. Methodologically, we introduce a common event-level metric that makes heterogeneous systems directly comparable despite major differences in architecture, coding scheme, and learning dynamics. Empirically, we combine measurements obtained by the authors with carefully normalized literature data to map the present energy landscape of artificial and biological neural computation. The results reveal three robust regimes. Dense von Neumann ANN implementations on GPUs, NPUs, and FPGA operate mainly in the $10^{5}$ to $10^{8}$ fJ range per synaptic event. Digital spiking processors reduce this requirement to about $10^{3}$ to $10^{4}$ fJ per event. Memristive and transistor-based artificial synapses span a much broader interval, from array-level values near $10^{6}$ fJ down to a few femtojoules in the most efficient device-level realizations. In particular, the best organic PANI memristive samples approach ~2 fJ per synaptic event, entering the biological vicinity and, in some cases, surpassing the rat reference range. Taken together, the benchmarked landscape spans up to ten orders of magnitude across the full set of systems considered in this study. This result clarifies where current AI hardware stands relative to biological efficiency, identifies event-driven and in-memory computation as the most promising routes toward sustainable AI, and provides a quantitatively grounded reference framework for future neuromorphic benchmarking. INDEX TERMS Artificial neural networks, energy consumption, energy efficiency, memristive devices, neuromorphic computing, spiking neural networks, sustainable AI.

The article analyzes the problems of academic emotionology as an applied branch of emotionology (the science of emotions) based on the use of sociomorphic neuromodeling, viewed through the prism of academic discourse in the educational system. For the first time, the proposed concept of sociomorphic neuromodeling is aimed not only at explaining, but also at applying a socially inherited mechanism of emotional response in a particular ethnic environment, which is reflected in the specifics of neuro-psychophysiological signals as a "deep" language. The main angle of the analysis is due to a non-discrete approach to assessing emotivity in communication and emotions in the speech and behavioral matrix of a modern native speaker. The authors consider the evidence of the proposed hypothesis verifying it with the work carried out on the emotivity analyzer of the speech and behavioral profile of a carrier a natural intelligence carrier and a "smart assistant" as a carrier of artificial intelligence. This representation allows us to undertake methodological holism towards the visual representation of emotions within NLP. The proposed "deep" neurocognitive designer of a "smart assistant" for social engineering rests on understanding neuromodulators activities as the neuro-psycholinguistic mechanisms of the human brain, which makes available the prospects for the development of the analyzer and the possibility of the proposed interface usage. The suggested refinement of the gradation of natural emotional intelligence and emotional artificial intelligence solves the issues of the general theory of emotions, which is an important step for further development of robotic humanoid systems based on the biomimetic concept. Authors' contribution. M. Talanov – development of a bioinspired "smart assistant" program, development of a mathematical model, writing part of the project; I.S. Karabulatova – general idea, development of the gradation of emotional intelligence, analysis of the history of the issue of psychometric research, writing a part, drafting instructions, editing; V. Erokhin – development of processes for biosensors and bioelectronics in the framework of neuromorphic computing; conducting experiments for bioinspired computing; clarifying terminology; J. Vallverdú – development of a project design using new intellectual solutions, development of a cognitive model, writing of a part of the project.

The paper examines how school websites are filled with information that is not required by law to be published but is highly demanded by the parent community, pupils, and teaching staff. Due to the fact that current solutions only deal with required information for school website publication, this approach has academic novelty. The analysis proved that there is not enough relevant information on school websites. Therefore, the development of assessment and monitoring automation of websites’ content is relevant, and they are developing it by creating methods that can be implemented in an automated program suite. The aim is to improve websites’ information content by implementing monitoring and assessment methods for automated program suite. To achieve the aim, the authors analyzed complex situations, developed information website content classification and created monitoring and assessment methods for school websites’ information content. The object of the research is school websites. The subject of the research is computer methods of text information processing. Methods of the research included an analysis of existing methods for assessment of information content on the website. The results revealed that developed methods enabled an effective assessment of information content on school websites with the use of an automated program suite as well.

The energy efficiency of Artificial Intelligence (AI) systems is a crucial and actual issue that may have an important impact on an ecological, economic and technological level. Spiking Neural Networks (SNNs) are strongly suggested as valid candidates able to overcome Artificial Neural Networks (ANNs) in this specific contest. In this study, the proposal involves the review and comparison of energy consumption of the popular Artificial Neural Network architectures implemented on the CPU and GPU hardware compared with Spiking Neural Networks implemented in specialized memristive hardware and biological neural network human brain. As a result, the energy efficiency of Spiking Neural Networks can be indicated from 5 to 8 orders of magnitude. Some Spiking Neural Networks solutions are proposed including continuous feedback‐driven self‐learning approaches inspired by biological Spiking Neural Networks as well as pure memristive solutions for Spiking Neural Networks.