This article introduces the global-local image perceptual score (GLIPS), an image metric designed to assess the photorealistic image quality of AI-generated images with a high degree of alignment to human visual perception. Traditional metrics such as Fréchet inception distance (FID) and kernel inception distance scores do not align closely with human evaluations. The proposed metric incorporates advanced transformer-based attention mechanisms to assess local similarity and maximum mean discrepancy to evaluate global distributional similarity. To evaluate the performance of GLIPS, we conducted a human study on photorealistic image quality. Comprehensive tests across various generative models demonstrate that GLIPS consistently outperforms existing metrics like FID, structural similarity index measure, and multiscale structural similarity index measure in terms of correlation with human scores. In addition, we introduce the interpolative binning scale, a refined scaling method that enhances the interpretability of metric scores by aligning them more closely with human evaluative standards. The proposed metric and scaling approach not only provide more reliable assessments of AI-generated images but also suggest pathways for future enhancements in image generation technologies.

Federated learning (FL), a distributed learning strategy, improves security and privacy by eliminating the need for clients to share their local data; however, FL struggles with non-IID (non independent and identically distributed) data. Clustered FL aims to remedy this by grouping similar clients and training a model per group; nevertheless, it faces difficulties in determining clusters without sharing local data and conducting model evaluation. Clustered FL evaluation on unseen clients typically applies all models, selecting the best-performer for each client - approach known as best-fit cluster evaluation. This paper challenges such evaluation process arguing that it violates a fundamental machine learning principle: test dataset labels should be used only for performance calculation, not for model selection. We show that best-fit cluster evaluation results in significant accuracy overestimates. Moreover, we present an evaluation approach that maintains the separation between model selection and evaluation by reserving a portion of the target client data for model selection, while the remaining data is used for accuracy estimation. Experiments on four datasets, encompassing various IID and non-IID scenarios, demonstrate that the best-fit cluster evaluation produces overestimates that are statistically different from our evaluation.

Smart grids enable real-time monitoring and optimization of energy generation, transmission, and consumption. Ensuring the integrity of data collected by smart meters is crucial for grid operation and data-driven decision-making, but missing data due to communication failures or device malfunctions can compromise data reliability. Recent studies have proposed deep learning techniques to impute missing points; however, these models require large training data, are computationally expensive, and struggle to scale to a large consumer base. Moreover, the interpretability of the imputation is also a concern, hindering trust and acceptance. To address these issues, this paper proposes Temporally Chained Equations (TCE), an interpretable, computationally lightweight missing data imputation approach for smart meters with low data requirements. TCE forms chained equations across the temporal axis using lead and lag features and imputes missing points in a manner coherent with neighboring and seasonally correlated points. Local normalization reduces the impact of outliers, and an iterative process refines the estimates until convergence. Experiments on a real-world dataset show that TCE outperforms related techniques, particularly for random missing points or short sequences of continuous missing points.

The increasing impact of human-induced climate change and unplanned urban constructions has increased flooding incidents in recent years. Accurate identification of flooded areas is crucial for effective disaster management and urban planning. While few works have utilized convolutional neural networks and transformer-based semantic segmentation techniques for identifying flooded areas from aerial footage, recent developments in graph neural networks have created improvement opportunities. This paper proposes an innovative approach, the Graph Attention Convolutional U-NET (GAC-UNET) model, based on graph neural networks for automated identification of flooded areas. The model incorporates a graph attention mechanism and Chebyshev layers into the U-Net architecture. Furthermore, this paper explores the applicability of transfer learning and model reprogramming to enhance the accuracy of flood area segmentation models. Empirical results demonstrate that the proposed GAC-UNET model, outperforms other approaches with 91% mAP, 94% dice score, and 89% IoU, providing valuable insights for informed decision-making and better planning of future infrastructures in flood-prone areas.

Civil structures worldwide are confronted with a growing threat of structural deterioration, aggravated by various factors such as climate change, population growth, and increased traffic. The latent nature of these issues often leads to undetected vulnerabilities until a catastrophic failure occurs, resulting in substantial losses. To address this challenge, there is a critical need for improved structural monitoring, condition assessment, and maintenance practices. Traditional inspection methods, relying on visual estimation and heavy equipment for inaccessible areas, present formidable obstacles to inspectors. These methods impede the safe and fast examination of structural damage, complicating tracking of structural deterioration, and hindering efficient condition assessment. Recognizing these challenges, this paper proposes a remote collaborative framework to enhance the efficiency of structural inspections by leveraging the capabilities of Augmented Reality (AR), QR code, and 5G network. The proposed framework centers on real-time remote collaboration among on-site and off-site inspectors, aiming to elevate safety, accessibility, and overall inspection efficacy. The integration of real-time data sharing and collaboration facilitates immediate decision-making, enabling inspectors to proactively address structural vulnerabilities and prevent potential failures. This study concludes that the proposed framework effectively facilitates real-time structural condition assessment for on-site AR users. Simultaneously, off-site web users can instantly track the progression of data over time through the utilization of 5G technology. The proposed advanced AR framework effectively demonstrates real-time structural condition assessment through a lab-scale experimental beam and a full-scale bridge.

Few-shot image classification (FSIC) is a computer vision task from the few-shot learning (FSL) category in which the model learns to classify images using only a few training samples. It has been demonstrated that even neural networks trained on large scale datasets are vulnerable to adversarial samples. This vulnerability is magnified in FSIC due to the low volume of training data. This paper proposes the use of hypersphere embedding and supervised contrastive learning to improve the adversarial robustness of representation learning-based FSIC. Contrastive learning contributes through its ability to bring together similar samples while pushing away dissimilar ones. On the other hand, hypersphere embedding has been successful in the representation learning tasks by restricting the embeddings to a hypersphere manifold. The proposed approach was evaluated on both 5-shot and 1-shot learning using two standard FSL networks and the standard Mini-ImageNet benchmark dataset. The evaluation shows that supervised contrastive training provides inherent adversarial robustness to the FSIC model while hypersphere embedding with cosine distance metrics improves the accuracy of the FSIC model and, when used in conjunction with an adversarial defense mechanism, boosts the adversarial performance.

In recent years, electric vehicles (EVs) have been widely adopted because of their environmental benefits. However, the increasing volume of EVs poses capacity issues for grid operators as simultaneously charging many EVs may result in grid instabilities. Scheduling EV charging for grid load balancing has a potential to prevent load peaks caused by simultaneous EV charging and contribute to balance of supply and demand. This paper proposes a user-preference-based scheduling approach to minimize costs for the user while balancing grid loads. The EV owners benefit by charging when the electricity cost is lower, but still within the user-defined preferred charging periods. On the other hand, the approach reduces the pressure on the grid by balancing the grid load. Two methods, the greedy algorithm and nonlinear programming, are considered along with users' charging preferences and durations. For scheduling small numbers of charging activities, the nonlinear programming model achieves better load balancing than the greedy algorithm; however, for scheduling medium to large numbers of charging activities, the greedy algorithm has a clear advantage in terms of time complexity.

Rapid growth in numbers of connected devices including sensors, mobile, wearable, and other Internet of Things (IoT) devices, is creating an explosion of data that are moving across the network. To carry out machine learning (ML), IoT data are typically transferred to the cloud or another centralized system for storage and processing; however, this causes latencies and increases network traffic. Edge computing has the potential to remedy those issues by moving computation closer to the network edge and data sources. On the other hand, edge computing is limited in terms of computational power, and thus, is not well-suited for ML tasks. Consequently, this article aims to combine edge and cloud computing for IoT data analytics by taking advantage of edge nodes to reduce data transfer. In order to process data close to the source, sensors are grouped according to locations, and feature learning is performed on the close by edge node. For comparison reasons, similarity-based processing is also considered. Feature learning is carried out with deep learning - the encoder part of the trained autoencoder is placed on the edge and the decoder part is placed on the cloud. The evaluation was performed on the task of human activity recognition from sensor data. The results show that when sliding windows are used in the preparation step, data can be reduced on the edge up to 80% without significant loss in accuracy.

Automated service classification plays a crucial role in service discovery, selection, and composition. Machine learning has been widely used for service classification in recent years. However, the performance of conventional machine learning methods highly depends on the quality of manual feature engineering. In this paper, we present a novel deep neural network to automatically abstract low-level representation of both service name and service description to high-level merged features without feature engineering and the length limitation, and then predict service classification on 50 service categories. To demonstrate the effectiveness of our approach, we conduct a comprehensive experimental study by comparing 10 machine learning methods on 10,000 real-world web services. The result shows that the proposed deep neural network can achieve higher accuracy in classification and more robust than other machine learning methods.

With the development of smart electricity metering technologies, huge amounts of consumption data can be retrieved on a daily and hourly basis. Energy consumption forecasting facilitates electricity demand management and utilities load planning. Most studies have been focussed on commercial customers or residential building-level energy consumption, or have used behavioral and occupancy sensor data to characterize an individual household's electrical consumption. This study has analyzed energy consumption at single household level using smart meter data to improve residential energy services and gain insights into planning demand response programs. Electricity consumption for anonymous individual households has been predicted using a Support Vector Regression (SVR) modelling with both daily and hourly data granularity. The electricity usage data set for 2014 to 2016 was obtained from a Canadian utility company. Exploratory data analysis (EDA) was used for data visualization and feature selection. The analysis presented here demonstrates that forecasting residential energy consumption for individual households is feasible, but the accuracy is highly dependable on household behaviour variability.

Recommender systems have dramatically changed the way we consume content. Internet applications rely on these systems to help users navigate among the ever-increasing number of choices available. However, most current systems ignore the fact that user preferences can change according to context, resulting in recommendations that do not fit user interests. This research addresses these issues by proposing the $$({ CF})^2$$ architecture, which uses local learning techniques to embed contextual awareness into collaborative filtering models. The proposed architecture is demonstrated on two large-scale case studies involving over 130 million and over 7 million unique samples, respectively. Results show that contextual models trained with a small fraction of the data provided similar accuracy to collaborative filtering models trained with the complete dataset. Moreover, the impact of taking into account context in real-world datasets has been demonstrated by higher accuracy of context-based models in comparison to random selection models.

Buildings are responsible for a significant amount of total global energy consumption and as a result account for a substantial portion of overall carbon emissions. Moreover, buildings have a great potential for helping to meet energy efficiency targets. Hence, energy saving goals that target buildings can have a significant contribution in reducing environmental impact. Today's smart buildings achieve energy efficiency by monitoring energy usage with the aim of detecting and diagnosing abnormal energy consumption behaviour. This research proposes a generic collective contextual anomaly detection (CCAD) framework that uses sliding window approach and integrates historic sensor data along with generated and contextual features to train an autoencoder to recognize normal consumption patterns. Subsequently, by determining a threshold that optimizes sensitivity and specificity, the framework identifies abnormal consumption behaviour. The research compares two models trained with different features using real-world data provided by Powersmiths, located in Brampton, Ontario, Canada.

Paper presented May 2015. Also published in Proceedings of the 5th International Conference on Cloud Computing and Services Science, pages 186-193; ISBN 978-989-758-104-5.

Predicting energy demand peak is a key factor for reducing energy demand and electricity bills for commercial customers. Features influencing energy demand are many and complex, such as occupant behaviours and temperature. Feature selection can decrease prediction model complexity without sacrificing performance. In this paper, features were selected based on their multiple linear regression correlation coefficients. This paper discusses the capabilities of M5 model trees in energy demand prediction for commercial buildings. M5 model trees are similar to regression trees, however they are more suitable for continuous prediction problems. The M5 model tree prediction was developed based on a selected feature set including sensor energy demand readings, day of the week, season, humidity, and weather conditions (sunny, rain, etc.). The performance of the M5 model tree was evaluated by comparing it to the support vector regression (SVR) and artificial neural networks (ANN) models. The M5 model tree outperformed the SVR and ANN models with a mean absolute error (MAE) of 8.94 compared to 10.02 and 12.04 for the SVR and ANN models respectively.

The demand for knowledge extraction has been increasing. With the growing amount of data being generated by global data sources (e.g., social media and mobile apps) and the popularization of context-specific data (e.g., the Internet of Things), companies and researchers need to connect all these data and extract valuable information. Machine learning has been gaining much attention in data mining, leveraging the birth of new solutions. This paper proposes an architecture to create a flexible and scalable machine learning as a service. An open source solution was implemented and presented. As a case study, a forecast of electricity demand was generated using real-world sensor and weather data by running different algorithms at the same time.

In the Big Data community, MapReduce has been seen as one of the key enabling approaches for meeting continuously increasing demands on computing resources imposed by massive data sets. The reason for this is the high scalability of the MapReduce paradigm which allows for massively parallel and distributed execution over a large number of computing nodes. This paper identifies MapReduce issues and challenges in handling Big Data with the objective of providing an overview of the field, facilitating better planning and management of Big Data projects, and identifying opportunities for future research in this field. The identified challenges are grouped into four main categories corresponding to Big Data tasks types: data storage (relational databases and NoSQL stores), Big Data analytics (machine learning and interactive analytics), online processing, and security and privacy. Moreover, current efforts aimed at improving and extending MapReduce to address identified challenges are presented. Consequently, by identifying issues and challenges MapReduce faces when handling Big Data, this study encourages future Big Data research.