Driven by the rapid increase in variable renewable generation, ancillary services are an integral part of the transition to carbon-neutral and cost-effective energy systems. Ancillary services maintain continuous and discrete balance between generation and load aggregates through frequency and voltage control. As a primary source of renewable energy, the accurate prediction of solar photovoltaic generation is crucial for the effective operation of Frequency Control Ancillary Services (FCAS). Although Deep Learning (DL) algorithms have been used to improve solar power forecasting accuracy, the computational complexity of DL is a technical limitation for the fast response time expected in FCAS. This article presents a novel DL approach composed of sparsity-inducing techniques that reduce the computational complexity of solar power forecasting without loss of forecasting accuracy. This approach applies sparsity within the forecasting models providing a balance between computational efficiency and interpretability, since model capacity is concentrated onto a smaller set of influential connections. The approach is empirically evaluated on two real-world solar power forecasting datasets from diverse geographical regions, followed by performance comparisons with dense DL models. The results confirm and validate the proposed approach in providing fast and accurate forecasts for the decision-making needs of FCAS.

Generative AI has rapidly progressed from chatbots and assistants to agents across diverse applications and domains. Generative AI agents demonstrate sophisticated operation through autonomy, tool use and decision making with minimal human input. Despite these performance gains, agents are still impacted by the foundational limitations of Generative AI models. Among these, hallucinations are a major limitation that affects agent operation in real-world settings, leading to risk and loss. Several recent work aim to address hallucinations through methods such as retrieval-augmented generation and reflection prompting, however, these only provide partial improvements. An effective yet underexplored approach is in the observability data generated by an agent in its deployed and operational settings. Drawing on agent observability data, this paper proposes a dual memory knowledge graph approach that integrates Semantic and Observability Memory to address hallucinations in Generative AI agents. Semantic Memory provides organized domain knowledge for precise factual grounding. Observability Memory transforms logs, traces, and execution results into agent validated planning histories. Hallucinations are then addressed by grounded planning in verified past interactions with known, reliable outcomes. This approach is evaluated in a two-stage experimental setup aligned with its dual memory design. Observability memory is evaluated on the HotpotQA dataset to assess its impact on reasoning grounding, using metrics that capture both factual accuracy and reasoning hallucinations. The SM3-Text-to-Query benchmark and Synthea-based medical QA datasets are used to assess factual grounding of the semantic memory. Results from both experiments demonstrate reductions in hallucinations, with semantic memory for contextual grounding reducing factual hallucinations, and observability memory for reasoning grounding reducing faithfulness hallucinations.

PurposeThis study aims to examine how investors react to managerial emotion profiles portrayed during earnings conference calls. The authors argue that positive and negative emotion profiles of the call participants influence investor reactions, as emotions serve as social information cues.Design/methodology/approachDrawing on emotions as social information theory, this study hypothesizes that emotions influence investor behavior through two primary mechanisms: inferential processes and affective reactions. This study uses a sophisticated machine learning algorithm and a validated psychology dictionary to extract and identify managers' emotion profiles during earnings conference calls. This study uses large language models along with lexicons to create a hybrid artificial intelligence framework, with granular levels of emotion detection.FindingsThe evidence suggests that investors react positively (negatively) to the positive (negative) emotion profiles displayed by the conference participants, supporting the assertion of emotion contagion. The findings provide mixed evidence on whether earnings news conditions the informational value of managerial emotions.Practical implicationsThe findings suggest that emotion management is a strategic component of communication - positive emotional expression can foster investor confidence, while unchecked negative emotions may amplify concerns and erode market value.Originality/valueThis study provides new evidence on the impact of emotion profiles on investor behavior. Emotion profiles may reveal new insights about managerial confidence or uncertainties that are not immediately apparent in the content of the disclosure.

The rapid adoption of Artificial Intelligence (AI) systems has increased risks associated with their development, deployment, and operation. Provenance of artefacts across the AI lifecycle is critical for addressing these risks through traceability. Although existing blockchain-based provenance systems provide immutable ledgers, they are passive due to the lack of enforcement at runtime and computationally expensive when evaluation is performed on-chain. To address these gaps, we present a novel approach based on a Merkle-style directed acyclic graph (Merkle-DAG) with off-chain constraint evaluation and on-chain enforcement. The Merkle-DAG represents process-level provenance artefacts from the full AI lifecycle, including datasets, features, models, metrics, and human oversight records such as reviews and approvals. This graph is evaluated off-chain against constraints by Open Policy Agent, followed by on-chain enforcement by a smart contract. The proposed approach is empirically demonstrated on three lifecycle stages, development, validation, and approval, of an energy consumption forecasting task in a real-world microgrid setting.

This article introduces a family of multiclass linear perceptron classifiers with a multiplicative margin mechanism (MMPerc), as an alternative to standard margin-free and additive margin perceptrons. The multiplicative formulation enforces classification confidence by requiring the true class score to exceed that of competing classes by a specified fraction of itself rather than by a fixed additive threshold. This avoids dependence on score magnitudes arising from varied norms of data and class weight vectors. We propose several architectural and algorithmic variants of MMPerc, derive associated loss functions and mistake bounds for both linearly separable and nonseparable data, and analyze key design considerations, including bias, margin threshold selection, and training modes. Extensive experiments on synthetic and real data sets show that MMPerc classifiers typically outperform the standard perceptron, as well as classic baselines such as support vector machines and ridge classifiers. Owing to their simplicity, minimalistic design, and computational efficiency, MMPerc classifiers are promising candidates for conventional machine learning tasks, linear evaluation of deep neural networks, integration with hyperdimensional computing and vector symbolic architecture representations, and deployment in resource-constrained applications.

The increasing prevalence of renewable energy and the evolving nature of energy consumption have motivated the need for more complex and dynamic microgrid energy management systems. Recent advances in artificial intelligence (AI) address these challenges by learning, predicting, and optimizing based on the large volumes of data generated by microgrid systems and related operations. Drawing on this context, the article proposes a novel framework for multi-agent reinforcement learning (MARL) with clustering and forecasting for optimized energy sharing in a microgrid environment with renewables and battery storage integration. The framework consists of three components: first, a structure-adapting unsupervised learning approach for creating clusters of prosumer energy consumption and generation patterns; second, a time-series forecasting ensemble for predicting future behaviors of the prosumers; and third, a continuous internal auction with a MARL for optimized energy sharing within the microgrid that collectively leads to reduced dependence on external energy sources. The proposed framework is empirically evaluated in the microgrid setting of a large multi-campus tertiary education institution. The results of this evaluation include stabilization of mean reward gain between independent agent and multi-agent models, impact of forecasting on MARL across seasonal variation, performance gains of 10–15% of MARL against heuristics, and scalability of the framework against cost, stability, reward, and convergence metrics. These results confirm the effectiveness of this AI framework for optimized prosumer energy sharing in microgrids with renewables and battery storage integration. Received: 9 October 2025 | Revised: 14 January 2026 | Accepted: 25 March 2026 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available at https://github.com/CDAC-lab/UNICON. Author Contribution Statement Daswin De Silva: Conceptualization, Methodology, Formal analysis, Writing – original draft, Supervision, Funding acquisition. Thimal Kempitiya: Software, Validation, Formal analysis, Investigation, Data curation, Writing – original draft. Nuwan Madhusanka: Software, Validation, Resources, Data curation, Writing – original draft. Prabod Rathnayaka: Formal analysis, Investigation, Resources, Writing – original draft, Visualization. Nishan Mills: Methodology, Validation, Investigation, Writing – original draft, Supervision. Andrew Jennings: Conceptualization, Formal analysis, Resources, Writing – original draft, Supervision. Milos Manic: Conceptualization, Methodology, Validation, Writing – original draft, Supervision.

The growing demand for Artificial Intelligence (AI) capabilities in diverse application settings is having a proportional impact on computational processing and memory required for building and deploying such models. This has triggered a transition from large AI models to small models and energy-efficient machine learning techniques. Vector Symbolic Architectures (VSA) provide an effective framework for this transition to energy efficient learning by encoding and manipulating data in high-dimensional vector space. Particularly, sparse distributed representations in VSA significantly improve computational and memory efficiency of most machine learning algorithms. In this paper, we examine the parametrization of sparse distributed representations in VSA for vector data classification tasks. Focusing on dimensionality and density, the paper explores strategies for encoding vector data using Sparse Block Codes (SBC), Sparse Binary Distributed Representations (SBDR), and dense Holographic Reduced Representation (HRR). We use a compositional approach to encode vector data and systematically evaluate parameter choices to optimize classification performance. Through a detailed parametrization using 121 UCI classification datasets, we identify key design considerations for sparse distributed representations. These experiments demonstrate that the accuracy improves rapidly with increasing dimensionality followed by saturation. For SBC, moderate structured density is consistently high performing: 25% density yields the highest average test accuracy. For SBDR, moderate-density CDT settings have the highest performance, while extreme densities degrade performance. Under optimal settings, sparse representations match state-of-the-art dense performance: SBC achieves 80.98%, SBDR 80.13%, and dense HRR 81.02% mean accuracy over the 121 datasets, providing practical guidance on achieving performance comparable to state-of-the-art VSA methods, highlighting their effectiveness for efficient and scalable machine learning.

The accelerated development of Artificial Intelligence (AI) capabilities and systems is driving a paradigm shift in productivity, innovation and growth. Despite this generational opportunity, AI is fraught with significant challenges and risks. To address these challenges, responsible AI has emerged as a modus operandi that ensures protections while not stifling innovations. Responsible AI minimizes risks to people, society, and the environment. However, responsible AI principles and practice are impacted by ‘principle proliferation’ as they are diverse and distributed across the applications, stakeholders, risks, and downstream impact of AI systems. This article presents a systematic review of responsible AI principles and practice with the objectives of discovering the current state, the foundations and the need for responsible AI, followed by the principles of responsible AI, and translation of these principles into the responsible practice of AI. Starting with 22,711 relevant peer-reviewed articles from comprehensive bibliographic databases, the review filters through to 9700 at de-duplication, 5205 at abstract screening, 1230 at semantic screening and 553 at final full-text screening. The analysis of this final corpus is presented as six findings that contribute towards the increased understanding and informed implementation of responsible AI.

Solar photovoltaics are on track to becoming the largest renewable energy source by 2029. This means a rapid increase in the number of solar energy generation installations from residential roof-top systems to utility-scale power plants. The current industrial approaches towards predictive maintenance will be insufficient to manage and maintain the increasing numbers of such installations at peak performance. In this paper, we propose hyper-predictive maintenance as a novel approach based on Generative Artificial Intelligence (AI) agents for highly autonomous management of solar energy infrastructure. The proposed Agentic AI framework deploys multiple agents for baseline generation from solar installations, predictive model development, degradation estimation, degradation evaluation and predictive maintenance that combines baseline performance with contextual information to predict faults and potential causes. This framework is empirically evaluated in the real-world solar energy systems of a multi-campus tertiary education institution. The results of these experiments confirm the robust and accelerated performance of Generative AI agents for the hyper predictive maintenance of large-scale solar energy installations.

Graph Neural Networks (GNNs) are the most common approach for learning complex relational data represented using graph data structures. Although GNNs are effective at learning representations of both nodes and graphs for a given task, the learning process is computationally expensive and as such, time and energy-inefficient. This paper investigates this challenge within the context of recent work on untrained graph representations that only train the solver model. We present Graph Vector Function Architecture (GVFA), a novel alternative to learning graph representations in GNNs that is based on hyperdimensional computing (HDC) principles. GVFA is a general zero-shot approach for graph and node representations without learning. As such, our representations are not task-specific and the computational costs of constructing them is substantially lower compared to learning-based GNN. Empirically, we demonstrate the expressiveness and generalization properties of different GVFA configurations. Our experimental results demonstrate that GVFA outperforms several classic GNNs on their benchmark datasets in terms of classification accuracy for both graph and node classification tasks, while also yielding a substantial reduction in training time.

The rapid adoption of artificial intelligence (AI) systems, such as predictive AI, generative AI, and explainable AI, is in contrast to the slower development and uptake of robotic AI systems. Dynamic environments, sensory processing, mechanical movements, power management, and safety are inherent complexities of robotic intelligence capabilities that can be addressed using novel AI approaches. The current AI landscape is dominated by machine learning techniques, specifically deep learning algorithms, that have been effective in addressing some of these challenges. However, these algorithms are subject to computationally complex processing and operational needs such as high data dependency. In this paper, we propose a computation-efficient and data-efficient framework for robotic motion intelligence (RMI) based on vector symbolic architectures (VSAs) and blockchain-based smart contracts. The capabilities of VSAs are leveraged for computationally efficient learning and noise suppression during perception, motion, movement, and decision-making tasks. As a distributed ledger technology, smart contracts address data dependency through a decentralized, distributed, and secure transactions ledger that satisfies contractual conditions. An empirical evaluation of the framework confirms its value and contribution towards addressing the practical challenges of robotic motion intelligence by significantly reducing the learnable parameters by 10 times while preserving sufficient accuracy compared to existing deep learning solutions.

Telephone cancer information and support services (CISS) deliver essential evidence-based resources for people living with cancer. This research aimed to describe how attributes and levels were developed for a future discrete choice experiment to elicit preferences for operational characteristics of a CISS, focusing on Cancer Council Victoria’s service. Using a mixed-methods approach guided by the ISPOR checklist for conjoint analysis in healthcare, initial attributes were developed using an artificial intelligence framework to analyse CISS calls (January 2018−December 2021), focus groups with people with cancer and carers using the CISS (July−August 2022), and a systematic literature review of qualitative studies. A four-stage descriptive process guided attribute and level development. An expert panel of researchers (n = 10), a CISS staff member, a person with lived experience of cancer and a consumer-only panel (n = 7) met monthly to prioritise, refine and finalise attributes by consensus. Call data analysis (people with cancer n = 7701; carers n = 5500), six focus groups (people with cancer n = 10; carers n = 11) and a systematic literature review of qualitative studies generated 14 candidate attributes. The expert panels selected seven final attributes, each with three levels: follow-up call, operating hours, additional technology, operator type, operator consistency, call length and service fee. Transparent reporting of the discrete choice experiment design process is essential for credible interpretation. The four-stage approach enhanced the comprehensibility of the experiment, as multi-modal data ensured the selected attributes and levels accurately reflect CISS caller priorities, which may be applicable to other choice-based studies. Worldwide cancer impacts more than one in five people, creating a significant need for reliable and easily accessible information. One approach to meet this need is through freely available telephone cancer services delivered by trained professionals. While these services provide valuable benefits, such as access to evidence-based information, they also have limitations for consumers. These may include restricted operating hours and costs incurred by the organisation to sustain the service. We reviewed existing research and engaged with people living with cancer and carers who have experience using telephone cancer services to identify what service features are important. These insights will inform a broader survey exploring people’s preferences for telephone cancer services to determine which service features are most highly valued. The results will help the Cancer Council, governments, healthcare providers and other cancer charities running these services make decisions about the design of telephone support services to better meet the needs of people with cancer and their families.

Vibe coding is a recent Generative AI innovation where a human operator provides conversational natural language description of a problem and an AI assistant writes the code to develop a solution. It is more than collaborating or assisting with coding, as the entirety of coding and technical implementation is delegated to the AI agent. Vibe coding presents an immediate productivity boost for prototype development, as it addresses the growing complexity of development tools, increased demand for time-efficient prototype development, and the shortage of skills needed to address this demand. This paper presents the role and contribution of vibe coding for prototyping industrial systems. The paper begins with an introduction to vibe coding and vibe coding tools, followed by a workflow comparison of conventional prototyping and vibe coding worked through a sample scenario of an industrial system prototype. A further productivity comparison of the two approaches across diverse dimensions of cognitive load, development speed, code quality, debugging and iteration is also reported. The paper concludes by highlighting the opportunities and risks of vibe coding, and a selection of best practices that can accelerate the adoption of vibe coding for prototyping industrial systems.

Smart cities are Hyper-Connected Digital Environments (HCDEs) that transcend the boundaries of natural, human-made, social, virtual, and artificial environments. Human activities are no longer confined to a single environment as our presence and interactions are represented and interconnected across HCDEs. The data streams and repositories of HCDEs provide opportunities for the responsible application of Artificial Intelligence (AI) that generates unique insights into the constituent environments and the interplay across constituents. The translation of data into insights poses several complex challenges originating in data generation and then propagating through the computational layers to decision outcomes. To address these challenges, this article presents the design and development of a Hyper-Automated AI framework with Generative AI agents for sustainable smart cities. The framework is empirically evaluated in the living lab setting of a ‘University City of the Future’. The developed AI framework is grounded on the core capabilities of acquisition, preparation, orchestration, dissemination, and retrospection, with an independent cognitive engine for hyper-automation of these AI capabilities using Generative AI. Hyper-automation output feeds into a human-in-the-loop process prior to decision-making outcomes. More broadly, this framework aims to provide a validated pathway for university cities of the future to take up the role of prototypes that deliver evidence-based guidelines for the development and management of sustainable smart cities.

Sleep is increasingly acknowledged as a cornerstone of public health, with chronic sleep loss implicated in preventable injury and deaths. Obstructive sleep apnea (OSA) affects over one billion people worldwide but remains widely under-diagnosed due to dependence on polysomnography (PSG), an overnight, hospital-based intrusive procedure. As an adjunct to the clinical diagnosis of OSA, this paper presents a low-cost, smartphone-based Generative AI agent framework for sleep apnea detection and sleep coaching at the bedside. Powered by an on=device Generative AI model, the four agents of this framework include a classifier, an analyser, a visualiser, and a sleep coach. The key agent activities performed are sleep apnea detection, sleep data management, data analysis, and natural language sleep coaching. The framework was empirically evaluated on a subject-independent hold-out set drawn from a dataset of 500 clinician annotated clips collected from 10 clinically diagnosed OSA patients. Sleep apnea detection achieved an accuracy of 0.89, precision of 0.91, and recall of 0.88, with nightly Apnea–Hypopnea Index (AHI) estimates strongly correlated with PSG-based clinical scores. The framework was further assessed on the performance metrics of computation, latency, memory, and energy usage. The results of these experiments confirm the feasibility of the proposed framework for large-scale, low-cost OSA screening, with pathways for future work in federated learning, noise robustness, and broad clinical validation.

Artificial Intelligence (AI) is transforming the way we live and work. The disruptive impact and risks of Generative AI have accelerated the global transition from voluntary AI ethics guidelines to mandatory AI regulation. The European Union AI Act is the world’s first horizontal and standalone law governing AI that came into force in August 2024, just as other jurisdictions, countries and states, are navigating possible modes of regulation. Starting with the EU AI Act, most of the current regulatory effort follows a risk-based classification approach. While this is prescriptive and application-focused, it overlooks the complex circular impacts of AI and the inherent limitations of measurement of risk, overemphasis on high-risk classification, perceived trustworthiness of AI and the geopolitical power imbalance of AI. This article contributes an overview of the current landscape of AI regulation, followed by a detailed assessment of the limitations and potential means of addressing these limitations through a structuration theory approach. Summarily, this approach can be used to recognise AI systems as agents that actively participate in the duality of structure, and the subsequent shaping of society. It acknowledges the direct negotiation of agency granted to machines alongside their ability to determine an understanding from given inputs, which then qualifies AI as an active participant in the recursive structuration of society. This agentic view of AI in the structuration theory approach complements ongoing efforts to develop a comprehensive and balanced AI regulation.

Neuromorphic Computing surpasses conventional von Neumann architectures in terms of energy efficiency, parallelisation, scalability, and stochasticity. Given the inherent structure of neurons and synapses, neuromorphic computers can be directly implemented as spiking neural networks. Despite these advantages, neuromorphic computing applications are hitherto limited to benchmark datasets and empirical demonstrations. This is primarily due to the lack of a unifying computing framework that designates a middle-layer abstraction between the actual neuromorphic computing and the required application functionality. Drawing on the distributed vector representation of symbolic and numerical data structures and robust dual interface with diverse operational primitives, Vector Symbolic Architectures (VSA) have been positioned as a suitable candidate to address this middle-layer void. In this paper, we explore the potential of VSA as an intermediary abstraction layer to advance practical neuromorphic computing applications. We introduce a novel vectorised framework that efficiently processes parallel streams of spiking data by combining and computing them through VSA for real-time downstream learning tasks, leveraging spike latency encoding. Our implementation utilises containerised methods within Lava, an open-source framework for neuromorphic computing.

Data Center Networks (DCNs) play a critical role in enabling scalable, efficient, and reliable digital infrastructure. This scoping review presents recent advancements in optimization and management strategies for DCNs, with a particular focus on the application of Artificial Intelligence (AI) and Machine Learning (ML) techniques. This study follows a structured methodology aligned with the PRISMA-ScR protocol and extends it using Large Language Models (LLMs) to support thematic classification and relevance screening. The analysis identified seven principal research domains: optical networking, congestion control, flow scheduling, load balancing, Software-Defined Networking (SDN), fault tolerance, and energy-efficient resource management. Within these domains, this study contrasts traditional approaches with emerging AI-driven methods, highlighting their potential and limitations. Notably, the findings revealed a lack of integrated end-to-end AI/ML frameworks capable of addressing the multifaceted demands of modern DCN environments. Additionally, this review highlights the need for scalable, explainable, and context-aware solutions that align with evolving DCN requirements. By analyzing the current body of knowledge and highlighting critical research gaps, this study provides a comprehensive foundation for future investigations into intelligent DCN design and operation.