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.

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.

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.

Thumb tip force was estimated from sEMG with an element description method (EDM). At the modeling, a hysteresis between sEMG and force was considered. An EDM is one of a system identification method and it can generate a non-black box model. Also, it is possible to set elements and structure of the model in an EDM modeling. The elements and functions was set to conduct a hysteresis modeling in this modeling. If the model is not a black box, the extensibility improves because it is possible to devise it. Moreover, a mechanical system needs a non-black box model because it is possible to adjust in case of emergency. In this paper, a novel model for the estimation with the EDM was proposed. Then, the estimation accuracy was compared with its of a conventional method. A long short-term memory (LSTM) was implemented as a conventional method in this estimation. In this proposal, the relationship was analyzed in stages which are that the force in increasing and decreasing. The functions were decided from the results of this analysis. At the force increasing and decreasing, the functions were set with an exponential and a hyperbolic tangent, respectively. As the result, the proposed method could improve the accuracy. From the results, the proposed method could estimate the force from sEMG than the conventional method. In addition, the calculation process of the model was clear.

Large Language Models (LLMs) are leading the Generative Artificial Intelligence transformation in natural language understanding. Beyond language understanding, LLMs have demonstrated capabilities in reasoning tasks, including commonsense, logical, and mathematical reasoning. However, their proficiency in causal understanding has been limited due to the complex nature of causal reasoning. Several recent studies have discussed the role of external causal models for improved causal understanding. Building on the success of Retrieval-Augmented Generation (RAG) for factual reasoning in LLMs, this paper introduces a novel approach that utilizes Causal Graphs as external sources for establishing causal relationships between complex vectors. This method is empirically evaluated using two benchmark datasets across the metrics of Context Relevance, Answer Relevance, and Grounding, in its ability to retrieve relevant context with causal alignment. The retrieval effectiveness is further compared with traditional RAG methods that are based on semantic proximity.

According to the World Health Organization (WHO), the population of people over 80 years of age is projected to triple by 2050, due to advances in quality of life, as well as continued progress in medicine and technology. With this context in mind, this article focuses on conducting a comprehensive study on conventional walkers, which represent a fundamental tool to improve the mobility of elderly people. The main objective of this study is to establish the foundations for the design of walkers, with the purpose of identifying the functional and end user requirements under the WalkIES project funded by IEEE’s Industrial Electronics Society (IES). In this initial phase, we seek to define the ergonomic design of a conventional walker, evaluating its advantages and disadvantages, in addition to establishing the standards that these devices must meet. Based on the results obtained, the specific characteristics of the WalkIES walker will be determined. Additionally, international manufacturing standards for walkers are examined and detailed, taking into account the specific features necessary to ensure the effectiveness and safety of these devices for elderly people. Special attention is paid to the individual characteristics of each component that makes up the walkers, with the aim of optimizing their functionality and comfort for the user. In summary, the study of walkers and the establishment of standards are essential to address the mobility needs of the elderly population and those who require assistance to walk, guaranteeing their safety, comfort, and quality of life, as well as to define the requirements of design and manufacturing for robotic walker WalkIES.

In this paper, the structure of a model to estimate human motion from surface-electromyography (sEMG) was defined mathematically and physically. Especially, an infinite impulse response (IIR) filter and an element description method were focused to generate the model in this paper. Human motion can be estimated from sEMG. However, the relationship between sEMG and human motion has a complex characteristics such as a non-linear relationship. The cause of the above complex relationship can be thought the muscle contraction velocity. Then, the IIR filter was applied to consider about the previous sEMG value in the estimation. In addition, some conventional models have been a black-box model and it is not easy to interpret the model mathematically and physically. From the above, an element description method (EDM) was applied to generate the model in this paper. EDM is one of the system identification method and it is easy to interpret the model mathematically and physically because EDM can generate the block diagram of the model.In this way, the proposed method could solve two conventional methods about the muscle contraction velocity and the black box model. Then, it was possible to define the structure of the model by the above in this paper. In this paper, the models about a wrist and a fingertip motion were generated and considered about the structure. As the results, the structure of models was defined with a hyperbolic tangent function. It will be possible to apply the model to various motions by just identifying the parameters about the relationship between each joint and muscle because the amount of muscles and range of sEMG are deferent from each muscle.

Motivated by recent innovations in biologically inspired neuromorphic hardware, this article presents a novel unsupervised machine learning algorithm named Hyperseed that draws on the principles of vector symbolic architectures (VSAs) for fast learning of a topology preserving feature map of unlabeled data. It relies on two major operations of VSA, binding and bundling. The algorithmic part of Hyperseed is expressed within the Fourier holographic reduced representations (FHRR) model, which is specifically suited for implementation on spiking neuromorphic hardware. The two primary contributions of the Hyperseed algorithm are few-shot learning and a learning rule based on single vector operation. These properties are empirically evaluated on synthetic datasets and on illustrative benchmark use cases, IRIS classification, and a language identification task using the n -gram statistics. The results of these experiments confirm the capabilities of Hyperseed and its applications in neuromorphic hardware.

Machine perception of emotions is integral to the development of human-centric Artificial Intelligence (AI) in sustainable industrial applications. Human expressions of emotions are not always direct. Word embeddings are mature techniques that can extract the semantics of such indirect expressions from text data. However, they are not primed to extract emotions. In this paper, we propose a novel approach that generates robust word embeddings for implicit and explicit expressions of emotion. This approach consists of two techniques, mask and rogue, we evaluate both techniques on two benchmark datasets for emotion classification. Our results confirm the effectiveness of the proposed approach in extracting emotions from diverse contexts. We have shared the emotion word embedding for public use.

Idioms are figures of speech that contradict the principle of compositionality. This disposition of idioms can misdirect Natural Language Processing (NLP) techniques, which mostly focus on the literal meaning of terms. In this paper, we propose a novel idiom detection model that distinguishes between literal and idiomatic expressions. It utilizes a token classification approach to fine-tune BERT(Bidirectional Encoder Representations from Transformers). It is empirically evaluated on four idiom datasets, yielding an accuracy of more than 0.94. This model adds to the robustness and diversity of NLP techniques available to process and understand increasing magnitudes of free-form text and speech. Furthermore, the social value of this model is in enabling non-native speakers to comprehend the nuances of a foreign language.

Healthcare systems and services are rapidly transitioning towards a patient-centred care approach. As an allied health profession that is highly individualised, physiotherapy, requires technological innovations to advance into this space. Virtual Reality (VR) technologies have been explored and applied successfully towards achieving this transition, specifically in single-player and multi-player gaming configurations. However, most VR games are not customised to the individuals and not appropriate for the rehabilitation of patients with unique needs. Given the large volumes of data generated by the patient during an engagement with a VR game, Artificial Intelligence (AI) can be leveraged to deliver personalised physiotherapy rehabilitation. This is a work-in-progress paper that presents the design and development of a framework for personalised physiotherapy rehabilitation using AI and VR, in the setting of a single-player VR game. The completed framework will be trialled in a public hospital setting to evaluate its effectiveness in providing patient-centred physiotherapy rehabilitation for diverse patient cohorts.

Microgrids and energy platforms have become increasingly intricate in nature. This is especially true in platforms adopted in large multi functional complexes which are spread across geographies of varying climactic conditions. Many of these complexes have evolved over time as organisations have grown and diversified, this has resulted in a mix of infrastructure, technology, networks, systems and equipment woven in an inextricable web. Often managing such complex hierarchies involve interacting with heterogeneous management platforms from various vendors and built on different platforms each providing the solution to a piece of the ’energy puzzle’ of the organisation. The complexities involved in managing such a network of systems are manifold and require extensive resourcing and expertise. Sustainability and Net Zero carbon emission initiatives that aim to achieve international and national targets add a further layer of complexity to the task at hand. In this paper, we propose a Cloud Edge architecture that leverages Artificial Intelligence (AI) and data analytics for microgrid energy optimisation and net zero carbon emissions. This architecture provides an intelligent and cohesive abstraction to assist in cataloging, unifying and managing the complexities of microgrids and enabling sustainable management of energy. The proposed architecture has been operationalised as the energy management and optimisation platform at a multi-campus, multi-functional tertiary education institution. Empirical evaluations conducted on this deployment have generated results that confirm the function and effectiveness of this architecture in addressing the emerging and evolving challenges of microgrid energy optimisation and net zero carbon emissions.

Hyperdimensional (HD) Computing leverages random high dimensional vectors (>10000 dimensions) known as hypervectors for data representation. This high dimensional feature representation is inherently redundant which results in increased robustness against noise and it also enables the use of a computationally simple operations for all vector functions. These two properties of hypervectors have led to energy efficient and fast learning capabilities in numerous Artificial Intelligence (AI) applications. Despite the increasing number of such AI HD applications, their susceptibility to adversarial attacks has not been explored, specifically in the text domain. To the best of our knowledge, this is the first research endeavour to evaluate the adversarial robustness of HD text classifiers and report on their vulnerability to such attacks. In this paper, we designed and developed n-grams based HD computing text classifiers for two primary applications of HD computing; language recognition and text classification, and then performed a set of character level and word level grey-box adversarial attacks, where an attacker’s goal is to mislead the target HD computing classifier to produce false prediction labels while keeping added perturbation noise as low as possible. Our results show that adversarial examples generated by the attacks can mislead the HD computing classifiers to produce incorrect prediction labels. However, HD computing classifiers show a higher degree of adversarial robustness in language recognition compared to text classification tasks. The robustness of HD computing classifiers against character-level attacks is significantly higher compared to word-level attacks and has the highest accuracy compared to deep learning-based classifiers. Finally, we evaluate the effectiveness of adversarial training as a possible defense strategy against adversarial attacks in HD computing text classifiers.

In this paper we introduce UNICON, a large-scale open dataset on UNIversity CONsumption of utilities, electricity, gas and water. This dataset is publicly released as part of La Trobe University’s commitment to Net Zero Carbon Emissions by 2029, for which we are building the La Trobe Energy AI/Analytics Platform (LEAP) that leverages Artificial Intelligence (AI) and Data Analytics to analyse, predict and optimize the consumption, generation and utilization of electricity, renewables, gas and water resources. UNICON contains consumption data for La Trobe’s five campuses in geographically distributed regions, across four years, 2018-2021 inclusive. This includes the COVID-19 global pandemic timeline of university shutdown and work from home measures that led to a significant decrease in the consumption of utilities. The consumption data consists of smart electricity meter readings at 15-minute granularity, gas meter readings at hourly intervals and water meter readings at 15-minute intervals. UNICON also contains weather data from the closest weather station to each campus, collected at two-speed latency of 1 minute and 10 minutes. The dataset is annotated with internal events of significance, such as energy conservation measures (ECMs) and other measurement and validation (M&V) activities conducted as part of LEAP optimization. To the best of our knowledge, this is the first large-scale, comprehensive, open dataset for the three main utilities, electricity, gas, and water consumption in a multi-campus university setting. A high granularity data dictionary and technical validation of the dataset for consumption trends, baseline modelling and forecasting are further contributions of this article that will enable interested research scientists, academics, industry practitioners, sustainability and energy consultants to experiment and evaluate their AI algorithms, models, forecasts, as well as inform the development of energy benchmarks, guidelines and much needed data-driven energy policies.

Safe and effective vaccination is leading the recovery from the COVID-19 pandemic. Despite the urgency of full vaccination that prevents serious illness, a state of vaccine messaging augmented by disinformation campaigns in online social networks has emerged. Several studies have established a link between social media activity and vaccine messaging, and most platforms are actively removing vaccine disinformation. The objective of this study is to apply a validated Artificial Intelligence (AI) framework to extract, analyze and synthesize themes, emotions and emotion transitions associated with COVID-19 vaccine messaging in online social networks. We applied the framework on approximately 400,000 COVID-19 vaccine-related posts and conversations on two social media platforms, Twitter and Reddit, from March 2020 to September 2021. The results of this study are threefold, firstly, the discovery of a minority of implied anti-vaccine themes on infertility, microchips, gene editing and fetal cells that have remained undetected. Secondly, the discovery of six themes that capture a majority of the vaccine messaging namely, social lockdown measures, frontline healthcare providers, side effects, vaccine distribution, breakthrough infections and vaccine efficacy on variants. Thirdly, the variety and intensity of emotions expressed since the start of the pandemic, and comparatively negative emotions being expressed in recent months. We anticipate the findings of our study will contribute towards improved vaccine messaging as the world returns to a new normal from the COVID-19 pandemic.

Recent advancements in Artificial intelligence (AI) have led to its widespread adoption in healthcare applications and services. The global pandemic has further hastened the integration of AI into telehealth services, such as service quality improvement and new models of care. This paper focuses on the service quality improvement of telehealth, specifically, cancer information and telephone support services, in terms of its human system interaction. We present an AI framework for the detection of patient emotions and emotion transitions based on call recordings of telehealth services. The call recordings are typically a conversation between the caller (patient or carer) and the healthcare practitioner (nurse or counsellor). All primary emotions are expressed across diverse topics during these calls. It is anticipated that the caller emotion state improves during the call due to the information and support received. The proposed AI framework is designed to detect emotions expressed at all stages of the call and based on these expressions, formulate the emotion transition during the call. We have evaluated the proposed AI framework on a large real-world dataset of 60,000 call recordings of cancer information and telephone support services provided by Cancer Council Victoria, Australia. The results confirm the effectiveness of this AI framework in detection of emotions and emotion transitions during the provision of telehealth services.

Time series forecasting is a critical requirement for the optimal operation of energy grids, systems, and platforms, where the forecasting challenge itself can span across energy consumption, renewables generation, and energy utilisation. Artificial Intelligence (AI) algorithms and models have been leveraged to predict these time series forecasts with increasing levels of accuracy. In contrast to local models that are developed separately for each time series, Global Models, which are trained across many sets of time series drawing on characteristics of ’relatedness’, have produced more accurate forecasts. In this paper, we propose a transformer architecture based global model as a generalist forecaster of energy time series data, where we frame a sequence forecasting model and represent numerical values of the corresponding time series as vector embeddings in this model. We evaluate this transformer architecture based global model on real-world time-series energy data generated by the La Trobe Energy AI platform (LEAP), a functional and operational microgrid deployed in the multicampus tertiary education setting of La Trobe University, Australia. The results of these experiments confirm that the proposed generalist forecasting approach outperforms specialist local models trained on individual time series. We also demonstrate the ability of this approach to forecast dissimilar time series from the same model.

The industrial advancement of human society has been fundamentally driven by diverse ‘systems’ that facilitate ‘human interaction’ within physical, digital, virtual, social and artificial environments, and upon the hyper-connected layers of system-system interactions across these environments. The research and practice of Human System Interaction (HSI) has undergone exponential development due to the enhanced capabilities, increased efficiencies and decreased costs of digitalization. Primarily driven by its unique capacity for information persistence, digitalization is now leading us into a nexus of transition where HSI is being transformed by Artificial Intelligence (AI). AI has leveraged the data and information amassed by digitalization to learn, reason, predict, optimize and thereby augment both human-system interaction and system-system interaction, within and across all hyper-connected environments noted above. In this paper, we review this evolution of HSI and contribute towards its future directions by articulating the AI transformation strategy for this nexus of transition into a Human-AI-System Interaction. The paper begins with a review of HSI that focuses on developments in the past 15 years, followed by the AI transformation strategy which comprises of the primary configurations for Human-AI-System Interaction, the current capabilities of AI, a lifecycle approach for the design, development and deployment of an AI solution and the ethical implications of AI in HSI.

The increasing use of Vector Symbolic Architectures (VSA) in machine learning has contributed towards en-ergy efficient computation, short training cycles and improved performance. A further advancement of VSA is to leverage sparse representations, where the VSA-encoded hypervectors are sparsified to represent receptive field properties when encoding sensory inputs. The hyperseed algorithm is an unsupervised machine learning algorithm based on VSA for fast learning a topology preserving feature map of unlabelled data. In this paper, we implement two methods of sparse block-codes on the hyperseed algorithm, they are selecting the maximum element of each block and selecting a random element of each block as the nonzero element. Finally, the sparsified hyperseed algorithm is empirically evaluated for performance using three distinct bench-mark datasets, Iris classification, classification and visualisation of synthetic datasets from the Fundamental Clustering Problems Suite and language classification using n-gram statistics.

Sequence-based methods for visual place recognition (VPR) have great importance due to their ability of additional information capture through the sequences compared to single image comparison. Vector symbolic architecture (VSA) started to gain attention within these methods due to the unique capabilities for representing variable-length sequences using single high-dimensional vectors. But the effect of different sequence parameters for the visual place recognition task is yet to be explored. In this work, we explore the parametrization of sequence encoding with VSA in the SeqNet variant of sequence-based visual place recognition and introduce a new hierarchical VPR method, which utilizes the proposed parametrization. We show that with our parametrization the VSA realization of sequence-based visual place recognition achieves on par results to conventional algorithms, while featuring the capability of being implemented on novel neuromorphic hardware for efficient execution.

Abstract The advancements of the Internet of Things (IoT) and voice-based multimedia applications have resulted in the generation of big data consisting of patterns, trends and associations capturing and representing many features of human behaviour. The latent representations of many aspects and the basis of human behaviour is naturally embedded within the expression of emotions found in human speech. This signifies the importance of mining audio data collected from human conversations for extracting human emotion. Ability to capture and represent human emotions will be an important feature in next-generation artificial intelligence, with the expectation of closer interaction with humans. Although the textual representations of human conversations have shown promising results for the extraction of emotions, the acoustic feature-based emotion detection from audio still lags behind in terms of accuracy. This paper proposes a novel approach for feature extraction consisting of Bag-of-Audio-Words (BoAW) based feature embeddings for conversational audio data. A Recurrent Neural Network (RNN) based state-of-the-art emotion detection model is proposed that captures the conversation-context and individual party states when making real-time categorical emotion predictions. The performance of the proposed approach and the model is evaluated using two benchmark datasets along with an empirical evaluation on real-time prediction capability. The proposed approach reported 60.87% weighted accuracy and 60.97% unweighted accuracy for six basic emotions for IEMOCAP dataset, significantly outperforming current state-of-the-art models.

The emergence of low latency and high bandwidth 5G networks, alongside localized computation and data storage of edge computing are enabling real-time applications in industrial settings, such as smart grid, smart cities, and smart factories. The resolution, frequency and variety of data streams generated by such applications are not effectively processed and analysed by contemporary machine learning algorithms. This challenge is further complicated by the unlabelled and non-deterministic nature of the data streams. Hardware accelerated machine learning has been proposed to address some of these challenges but limited work has been published on unsupervised learning from unlabelled data. In this paper, we extend the hardware accelerated Self Organizing Map (SOM) algorithm by optimizing the learning rule for computational efficiency, followed by a comparative empirical evaluation with two other variants, tri-state SOM and integer SOM. We have used two datasets representative of real-time industrial applications in 5G networks and smart grids, for this evaluation.

Smart cities endeavor to deliver safe and sustainable infrastructure services that enable individuals, organizations, and communities alike to be productive, healthy, informed, and actively involved in rapid urbanization. The widespread installation of closed-circuit television cameras and continuously generated video streams are a strategic data source that can contribute toward safety and sustainability through efficient surveillance of smart city assets and resources. Recent advances in deep learning methods are able to detect and localize salient objects in a video stream. However, a number of practical issues remain unaddressed, such as suboptimality, latency, predictive accuracy, and most importantly the contextualization of all detected salient objects for informed decisions that aligns with ethical surveillance. In this article, we propose a Generative Latent Space (GenLS) approach that overcomes these challenges, specifically in road surveillance. We demonstrate an adaptation of this approach for a prominent use-case in road surveillance, License Plate Detection. GenLS was evaluated for accuracy, robustness, computational cost, and cogency, using a state-of-the-art benchmark dataset on road traffic. Results from these experiments and the corresponding ablation study validate GenLS and confirm its suitability for real-time smart city road surveillance.

Work-related stress is an uncompromising burden with compounding effects on individuals, communities, organizations, and the economy. Highly automated digital environments, such as smart factories and cyber-physical ecosystems, are significantly impacted by these negative effects due to the inherent constraints on human engagement and social interaction. Verbal communication between human operators is a critical resource in such environments, as it can be effectively utilized to address this challenge. A mix of statistical and artificial intelligence (AI) techniques have been reported in recent literature for the detection of stress-related indicators from audio recordings. However, none of these studies has focused on the challenges of detecting work-related stress in cyber-physical environments, such as smart factories, where verbal communication is not only constrained but also impaired by background noise and other disturbances common to factory settings. In this paper, we address these challenges by proposing a novel deep learning approach, based on the workings of the Convolutional Neural Network (CNN) and Growing Self-Organizing Map (GSOM) algorithms. In addition to this novel AI approach, sampling strategy and speaker diarization techniques are utilised for noise reduction. Pitch and speech augmentation techniques address the data imbalance issue common in most real-world datasets. The accuracy and effectiveness of the proposed approach is demonstrated using a benchmark dataset (DAIC-WOZ). We report F1 scores of 82% and 64% for normal and distressed classes respectively, which outperform the state-of-the-art models. We conclude with a discussion on the empirical evaluation of the proposed approach in cyber-physical environments and directions for future work.

Humans have continuously endeavored to enhance their sensory perception and awareness of the physical surroundings for the betterment of themselves as individuals as well as communities. Advancing this notion to the present day, the Internet of Things (IoT) provides a unique opportunity to attempt the same in an increasingly digital landscape. The prevalence and pervasiveness of IoT data streams gives us this ability to represent a situation with clarity and nuance, leading to a refined awareness of the environment. However, there are several challenges in managing the scale, velocity, and magnitude of IoT data streams, as well as the cohesive representation of these varied sources in a single frame of reference. In this article, we present a new algorithm, the deep growing self-organizing map (deep GSOM) algorithm that addresses these challenges. Deep GSOM incrementally generates a latent representation of situational awareness from high entropy to low entropy IoT data streams. It utilizes an implementation of the fuzzy integral to define a metric that can be moved across spatial and temporal situations to profile the density of congestion. We have also expanded deep GSOM into an IoT platform that can collate, aggregate, and process an entire network of IoT data streams. We demonstrate the workings of deep GSOM on the real-life scenario of profiling vehicular and pedestrian movement using IoT data streams of two highly urbanized cities. The results of these experiments confirm the validity and effectiveness of the proposed approach for generating situational awareness from multiple IoT data streams.

The rapid growth in autonomous industrial environments has increased the need for intelligent video surveillance. As a predominant element of video surveillance, recognition of complex human movements is important in a wide range of surveillance applications. However, the current state-of-the-art video surveillance techniques use supervised deep learning pipelines for human activity recognition (HAR). A key shortcoming of such techniques is the inability to learn from unlabeled video streams. To operate effectively in natural environments, video surveillance techniques have to be able to handle huge volumes of unlabeled video data, monitor and generate alerts and insights derived from multiple characteristics such as spatial structure, motion flow, color distribution, etc. Furthermore, most conventional learning systems lack memory persistence capability which can reduce the influence of outdated information in memory-guided decision-making resulting in limiting plasticity and overfitting based on specific past events. In this article, we propose a new adaptation of the Growing Self-Organizing Map (GSOM) to address these shortcomings by 1) adopting two proven concepts of traditional deep learning, hierarchical, and multistream learning, applied into GSOM self-structuring architecture to accommodate learning from unlabeled video data and their diverse characteristics, 2) address overfitting and the influence of outdated information on neural architecture by implementing a transience property in the algorithm. We demonstrate the proposed model using three benchmark video datasets and the results confirm its validity and usability for HAR.

Rapid developments in urbanization and autonomous industrial environments have augmented and expedited the need for intelligent real-time video surveillance. Recent developments in artificial intelligence for anomaly detection in video surveillance only address some of the challenges, largely overlooking the evolving nature of anomalous behaviors over time. Tightly coupled dependence on a known normality training dataset and sparse evaluation based on reconstruction error are further limitations. In this article, we propose the incremental spatiotemporal learner (ISTL) to address challenges and limitations of anomaly detection and localization for real-time video surveillance. ISTL is an unsupervised deep-learning approach that utilizes active learning with fuzzy aggregation, to continuously update and distinguish between new anomalies and normality that evolve over time. ISTL is demonstrated and evaluated on accuracy, robustness, computational overhead as well as contextual indicators, using three benchmark datasets. Results of these experiments validate our contribution and confirm its suitability for real-time video surveillance.

The Self-Organizing Map algorithm has been proven and demonstrated to be a useful paradigm for unsupervised machine learning of two-dimensional projections of multidimensional data. The tri-state Self-Organizing Maps have been proposed as an accelerated resource-efficient alternative to the Self-Organizing Maps for implementation on field-programmable gate array (FPGA) hardware. This paper presents a generalization of the tri-state Self-Organizing Maps. The proposed generalization, which we call integer Self-Organizing Maps, requires only integer operations for weight updates. The presented experiments demonstrated that the integer Self-Organizing Maps achieve better accuracy in a classification task when compared to the original tri-state Self-Organizing Maps.

Industrial applications are increasingly adopting conversational agents (chatbots) for tasks ranging from primitive conversation interfaces to intelligent human assistants. In this emerging field of research, there has been a strong drive towards modelling human-like characteristics and behaviors in chatbots. However, only a limited number of research endeavors have focused on a chatbot for automatic characterization of end-user emotions. In order to address this limitation, we propose an artificial intelligence based cognitive model for emotion awareness in chatbots using Markov chains, word embedding, and Natural Language Processing. The proposed model is able to extract emotions from conversations, detect emotion transitions over time, predict real-time emotions and intelligently profile human participants based on their distinct emotional characteristics. We conducted experiments using a real-world end-user dataset to demonstrate the functionality of the proposed model. Results from experiments confirm the plausibility of this model for emotion awareness in industrial conversational agents.

Recognition of complex human activities is a prominent area of research in intelligent video surveillance. The current state-of-the-art techniques are largely based on supervised deep learning algorithms. The inability to learn from unlabeled video streams is a key shortcoming in supervised techniques in most current applications where large volumes of unlabeled video data are utilized. Furthermore, the dominant focus on persistence in traditional machine learning algorithms has induced two limitations; the influence of outdated information in memory- guided decision making, and overfitting of acquired knowledge on specific past events, weakening the plasticity of the learning system. To address the above requirements, we propose a new adaptation of the Growing Self Organizing Map (GSOM), formed in a hierarchical two-stream learning pipeline to accommodate unlabeled video data for human activity recognition, which facilitates plasticity by implementing a transience property, without losing the stability of the learning system. The proposed model is evaluated using two benchmark video datasets, confirming its validity and usability for human activity recognition.

The fusion of sensor data is an essential requirement for active perception in autonomous robots in order to accurately perceive an environment, make appropriate adjustments that improve the understanding of the environment and generate actionable insights. Despite a large body of literature in this domain, only a handful of research focuses on unsupervised machine learning, which is increasingly important in unlabeled, unknown environments. This paper proposes a scalable self-organizing neural architecture for environmental perception based on multimodal fusion using unsupervised machine learning. Inspired by the biological counterpart, the neural architecture consists of topographic maps across each modality and incorporates a novel scalable self-organization process that handles high volume, high-velocity sensor data. The co-occurrence relationships across modalities are captured in cross-modal connections and the reverse Hebbian projection is used for multimodal representation. This proposed unsupervised machine learning-based scalable fusion technique was implemented on Apache Hadoop and Apache Spark. It was validated using an extensive multimodal human activity sensor dataset to demonstrate efficient representation and multimodal fusion for active perception.

The emergence of artificial intelligence (AI), empowered by robust computing infrastructure and abundance of data, maintains potential for radical transformation of human society, essentially a third phase in evolution. Numerous research endeavor, policy development, and thought-leadership are presently in progress aimed at discovering data-driven intelligent decision-making solutions for smart cities, smart grids, smart homes, and informed citizens as well as addressing potential risks posed by AI workplace automation. Joining this broad effort, this Special Section contributes six research articles that consolidate recent developments in AI for industrial informatics.

The technological landscape of intelligent transport systems (ITS) has been radically transformed by the emergence of the big data streams generated by the Internet of Things (IoT), smart sensors, surveillance feeds, social media, as well as growing infrastructure needs. It is timely and pertinent that ITS harness the potential of an artificial intelligence (AI) to develop the big data-driven smart traffic management solutions for effective decision-making. The existing AI techniques that function in isolation exhibit clear limitations in developing a comprehensive platform due to the dynamicity of big data streams, high-frequency unlabeled data generation from the heterogeneous data sources, and volatility of traffic conditions. In this paper, we propose an expansive smart traffic management platform (STMP) based on the unsupervised online incremental machine learning, deep learning, and deep reinforcement learning to address these limitations. The STMP integrates the heterogeneous big data streams, such as the IoT, smart sensors, and social media, to detect concept drifts, distinguish between the recurrent and non-recurrent traffic events, and impact propagation, traffic flow forecasting, commuter sentiment analysis, and optimized traffic control decisions. The platform is successfully demonstrated on 190 million records of smart sensor network traffic data generated by 545,851 commuters and corresponding social media data on the arterial road network of Victoria, Australia.

Multimodal sensory fusion is a fundamental requirement for autonomous robots to form an unambiguous and meaningful representation of their surroundings. In this paper, we propose a multisensory self-organizing neural architecture for multimodal fusion using unsupervised machine learning. Inspired by biological evidence from the organization of the human sensory system, the proposed architecture consists of self-organizing neural layers for learning individual modalities. We have incorporated scalable computing for self-organization, so the processing can be scaled to support large datasets and short computation times. The lateral associative connections capture the co-occurrence relationships across individual modalities for cross-modal fusion in obtaining a multimodal representation. Experiments are conducted on an audio-visual dataset consisting of utterances to evaluate the quality of multimodal fusion over individual unimodal representations. Multimodal representation achieves significant improvements over the unimodal representations. These results indicate the proposed architecture is capable of forming effective multimodal representations in short computation times from congruent multisensory stimuli.

Driver behavior recognition is an essential determinant for effective coordination and communication between autonomous and human-driven vehicles. Effective coordination will ensure traffic flow optimization, collision avoidance and hazard detection. Humans exhibit a diverse array of driving behaviors and respond differently to task demands and capabilities of each driving situation. These behaviors can be captured from driving data generated by the vehicle, such as acceleration, brake position, steering wheel position, and transformed into driver behavior characterizations using unsupervised incremental machine learning. In this paper, we extend the Driver Demands and Capabilities Model for intelligent change detection. This model is implemented as an unsupervised incremental machine learning algorithm for behavior change detection, and differentiation between abrupt behavior change and repeating behavior change. Experiments were conducted using the first openly available dataset of annotated DAVIS driving recordings accompanying driving data. Results demonstrate and confirm the capabilities of the proposed model and algorithm for detecting driver behavior change, distinguishing between abrupt and repeat behavior changes. Detected changes can be communicated to all vehicles within the immediate vicinity for effective coordination and improved situational understanding.

The traditional approach to health care is being revolutionized by the rapid adoption of patient‐centered healthcare models. The successful transformation of patients from passive recipients to active participants is largely attributed to increased access to healthcare information. Online support groups present a platform to seek and exchange information in an inclusive environment. As the volume of text on online support groups continues to grow exponentially, it is imperative to improve the quality of retrieved information in terms of relevance, reliability, and usefulness. We present a text‐mining approach that generates a knowledge extraction layer to address this void in personalized information retrieval from online support groups. The knowledge extraction layer encapsulates an ensemble of text‐mining techniques with a domain ontology to interpose an investigable and extensible structure on hitherto unstructured text. This structure is not limited to personalized information retrieval for patients, as it also imparts aggregates for crowdsourcing analytics by healthcare researchers. The proposed approach was successfully trialed on an active online support group consisting of 800,000 posts by 72,066 participants. Demonstrations for both patient and researcher use cases accentuate the value of the proposed approach to unlock a broad spectrum of personalized and aggregate knowledge concealed within crowdsourced content.

The proliferation of Internets of Things (IoT) technologies in both industrial and non-industrial settings has led to the accumulation of Big Data sets. Analysis of these high-volume, high-velocity datasets require advanced processing techniques that incorporate parallel and distributed computations. In this paper, we present a novel distributed self-adaptive neural-network algorithm, the Distributed Growing Self-Organizing Map (DGSOM) algorithm to address the growing need for unsupervised machine learning of Big Data sets on distributed computing environments. The algorithm was tested on a Big Data set of sensor recordings of human activity collected from wearable devices, 2.8 million records. Results indicate that the distributed algorithm significantly reduces execution time compared to its serial counterpart. Moreover, the self-adaptive nature and controlled growth of the algorithm demonstrates data-driven structure adaptation and multi-granular pattern analysis. Overall, the proposed algorithm addresses the need for pattern discovery and visualization from Big Data sets generated by IoT devices which are increasingly commonplace in industrial scenarios.

IoT systems deployed in industrial and smart factory settings generate large volumes of data at high velocity. Context awareness is mandatory for knowledge discovery and actionable insights from such high-velocity, high-volume IoT data streams. Changes to the context of a data stream are represented in the underlying data distribution. Research in concept drift aims to detect and adapt to such changes in a data distribution. Concept drift detection can be extended to suit ad hoc Big Data streams generated by IoT systems, by introducing the cognitive principles of learning. This paper proposes an unsupervised incremental learning algorithm for detection and adaption of concept drift based on the cognitive principles of learning. It executes in automated time windows, detects concept drift using movement in space and determines the type of concept drift using movement in time. The algorithm was applied to a Big Data set representing an IoT system for urban vehicular movement and traffic. Results confirm that the proposed algorithm generates context-awareness by detection and adaptation to concept drift in high-volume, high-velocity IoT systems.

The world is witnessing a remarkable increase in the usage of video surveillance systems. Besides fulfilling an imperative security and safety purpose, it also contributes towards operations monitoring, hazard detection and facility management in industry/smart factory settings. Most existing surveillance techniques use hand-crafted features analyzed using standard machine learning pipelines for action recognition and event detection. A key shortcoming of such techniques is the inability to learn from unlabeled video streams. The entire video stream is unlabeled when the requirement is to detect irregular, unforeseen and abnormal behaviors, anomalies. Recent developments in intelligent high-level video analysis have been successful in identifying individual elements in a video frame. However, the detection of anomalies in an entire video feed requires incremental and unsupervised machine learning. This paper presents a novel approach that incorporates high-level video analysis outcomes with incremental knowledge acquisition and self-learning for autonomous video surveillance. The proposed approach is capable of detecting changes that occur over time and separating irregularities from re-occurrences, without the prerequisite of a labeled dataset. We demonstrate the proposed approach using a benchmark video dataset and the results confirm its validity and usability for autonomous video surveillance.

The global popularity of microblogs has led to an increasing accumulation of large volumes of text data on microblogging platforms such as Twitter. These corpora are untapped resources to understand social expressions on diverse subjects. Microblog analysis aims to unlock the value of such expressions by discovering insights and events of significance hidden among swathes of text. Besides velocity; diversity of content, brevity, absence of structure and time‐sensitivity are key challenges in microblog analysis. In this paper, we propose an unsupervised incremental machine learning and event detection technique to address these challenges. The proposed technique separates a microblog discussion into topics to address the key problem of diversity. It maintains a record of the evolution of each topic over time. Brevity, time‐sensitivity and unstructured nature are addressed by these individual topic pathways which contribute to generate a temporal, topic‐driven structure of a microblog discussion. The proposed event detection method continuously monitors these topic pathways using multiple domain‐independent event indicators for events of significance. The autonomous nature of topic separation, topic pathway generation, new topic identification and event detection, appropriates the proposed technique for extensive applications in microblog analysis. We demonstrate these capabilities on tweets containing #microsoft and tweets containing #obama.

As the world advances into the information age, the proliferating demand for energy entails an increasing need for effective smart energy management systems. The resulting smart grids and smart energy management solutions are beginning to generate Big Data; high-variety data at cumulative volumes and velocity. Simultaneously, data analytics techniques and methodologies are being introduced to comprehend this changing nature of data. A number of conventional data mining techniques have successfully transitioned into the Big Data landscape. However, integration of multi-source information in this landscape remains hitherto unaddressed. In this paper, we present a novel data fusion technique that incrementally integrates information from multiple sources. Based on an incremental, unsupervised learning algorithm and possibilistic fusion, the technique overcomes limitations to continuous learning and integrating information of mixed granularity. The technique is also extensible into the Big Data landscape. The collective effect of these features postulate smart energy management as a fitting application domain for the proposed technique. We demonstrate practical applicability of the technique using a household energy and utility consumption dataset. Results and analytics outcomes from these experiments confirm its effectiveness as a data fusion technique and its extensibility into further high-volume applications in smart energy management.

Growing volumes of text and increasing expectations on the complexity of analysis entail advanced approaches to text mining. Unsupervised text clustering is an efficient approach to determine structural groupings in a text corpus without the impact of external bias. The information content of such structural groupings needs to be enhanced by integrating semantics into the cluster outcomes. This integration can eventuate at different stages of the clustering process where semantic relatedness measures can be used for the integration. In this paper we propose a novel method of semantics integration to cluster separation in a dynamic self-organizing map algorithm. We demonstrate the effectiveness of the proposed method and the value of semantics integration for cluster separation with empirical results from two benchmark text datasets.

This paper presents a novel data mining framework for the exploration and extraction of actionable knowledge from data generated by electricity meters. Although a rich source of information for energy consumption analysis, electricity meters produce a voluminous, fast-paced, transient stream of data that conventional approaches are unable to address entirely. In order to overcome these issues, it is important for a data mining framework to incorporate functionality for interim summarization and incremental analysis using intelligent techniques. The proposed Incremental Summarization and Pattern Characterization (ISPC) framework demonstrates this capability. Stream data is structured in a data warehouse based on key dimensions enabling rapid interim summarization. Independently, the IPCL algorithm incrementally characterizes patterns in stream data and correlates these across time. Eventually, characterized patterns are consolidated with interim summarization to facilitate an overall analysis and prediction of energy consumption trends. Results of experiments conducted using the actual data from electricity meters confirm applicability of the ISPC framework.

Incremental learning is a core necessity in developments towards intelligent machines. Artificial learning as implemented in contemporary neural network algorithms does not fully encompass an incremental, autonomous learning capacity. In this paper we present a self learning algorithm capable of incrementally acquiring knowledge across learning periods. A dynamic unsupervised learning algorithm, the GSOM algorithm, forms the basis of the presented incrementally knowledge acquiring self learning (IKASL) algorithm, to which we have introduced a layer of aggregation for continuous learning, knowledge acquisition and retention. We also present a novel application of the IKASL algorithm for continuous learning of hidden patterns from semantics of text.