Various real-world problems can be attributed to constrained multiobjective optimization problems (CMOPs). Although there are various solution methods, it is still very challenging to automatically select efficient solving strategies for CMOPs. Given this, a process knowledge-guided constrained multiobjective autonomous evolutionary optimization method is proposed. First, the effects of different solving strategies on population states are evaluated in the early evolutionary stage. Then, the mapping model of population states and solving strategies is established. Finally, the model recommends subsequent solving strategies based on the current population state. This method can be embedded into existing evolutionary algorithms, which can improve their performances to different degrees. The proposed method is applied to 41 benchmarks and 30 dispatch optimization problems of the integrated coal mine energy system. Experimental results verify the effectiveness and superiority of the proposed method in solving CMOPs.

Non-intrusive load monitoring (NILM), which usually utilizes machine learning methods and is effective in disaggregating smart meter readings from the household-level into appliance-level consumption, can help to analyze electricity consumption behaviours of users and enable practical smart energy and smart grid applications. However, smart meters are privately owned and distributed, which make real-world applications of NILM challenging. To this end, this paper develops a distributed and privacy-preserving federated deep learning framework for NILM (FederatedNILM), which combines federated learning with a state-of-the-art deep learning architecture to conduct NILM for the classification of typical states of household appliances. Through extensive comparative experiments, the effectiveness of the proposed FederatedNILM framework is demonstrated.

Prior feature transformation based approaches to Unsupervised Domain Adaptation (UDA) employ the deep features extracted by pre-trained deep models without fine-tuning them on the specific source or target domain data for a particular domain adaptation task. In contrast, end-to-end learning based approaches optimise the pre-trained backbones and the customised adaptation modules simultaneously to learn domain-invariant features for UDA. In this work, we explore the potential of combining fine-tuned features and feature transformation based UDA methods for improved domain adaptation performance. Specifically, we integrate the prevalent progressive pseudo-labelling techniques into the fine-tuning framework to extract fine-tuned features which are subsequently used in a state-of-the-art feature transformation based domain adaptation method SPL (Selective Pseudo-Labeling). Thorough experiments with multiple deep models including ResNet-50/101 and DeiT-small/base are conducted to demonstrate the combination of fine-tuned features and SPL can achieve state-of-the-art performance on several benchmark datasets.

Abstract Online federated learning (OFL) and online transfer learning (OTL) are two collaborative paradigms for overcoming modern machine learning challenges such as data silos, streaming data, and data security. This survey explores OFL and OTL throughout their major evolutionary routes to enhance understanding of online federated and transfer learning. Practical aspects of popular datasets and cutting-edge applications for online federated and transfer learning are also highlighted in this work. Furthermore, this survey provides insight into potential future research areas and aims to serve as a resource for professionals developing online federated and transfer learning frameworks.

We address the Unsupervised Domain Adaptation (UDA) problem in image classification from a new perspective. In contrast to most existing works which either align the data distributions or learn domain-invariant features, we directly learn a unified classifier for both the source and target domains in the high-dimensional homogeneous feature space without explicit domain alignment. To this end, we employ the effective Selective Pseudo-Labelling (SPL) technique to take advantage of the unlabelled samples in the target domain. Surprisingly, data distribution discrepancy across the source and target domains can be well handled by a computationally simple classifier (e.g., a shallow Multi-Layer Perceptron) trained in the original feature space. Besides, we propose a novel generative model norm-AE to generate synthetic features for the target domain as a data augmentation strategy to enhance the classifier training. Experimental results on several benchmark datasets demonstrate the pseudo-labelling strategy itself can lead to comparable performance to many state-of-the-art methods whilst the use of norm-AE for feature augmentation can further improve the performance in most cases. As a result, our proposed methods (i.e. naive-SPL and norm-AE-SPL) can achieve comparable performance with state-of-the-art methods with the average accuracy of 93.4% and 90.4% on Office-Caltech and ImageCLEF-DA datasets, and achieve competitive performance on Digits, Office31 and Office-Home datasets with the average accuracy of 97.2%, 87.6% and 68.6% respectively.

In design for additive manufacturing, an essential task is to determine the optimal build orientation of a part according to one or multiple factors. Heuristic search is used by the most part orientation methods to select the optimal orientation from a large solution space. Search algorithms occasionally converge towards the local optimum and waste considerable time on trial and error. This issue could be addressed if there was an intelligent agent that knew the optimal search path for a given 3D model. A straightforward method to construct such an agent is reinforcement learning (RL). By adopting this idea, the time-consuming online searches in existing part orientation methods will be moved to the offline learning stage, potentially improving part orientation performance. This is a challenging problem because the goal is to build an agent capable of rotating arbitrary 3D models, whereas RL agents frequently struggle to generalize in new scenarios. Therefore, this paper suggests a generalizable reinforcement learning (GRL) framework to train the agent, and a GRL benchmark to support the training, testing, and comparison of part orientation approaches. Experimental results de- monstrate that the proposed method on average outperforms others in terms of effectiveness and efficiency. It is proved to have the potential to solve the local minima problems raised in the existing approaches, to swiftly discover the global (sub-)optimal solution (i.e., on average 2.62× to 229.00× faster than the random search algorithm), and to generalize beyond the environment in which it was trained.

Due to the growing needs in decision-making under uncertainty, existing studies introduced consistency and consensus-driven algorithms for group decision-making (GDM) problems with hesitant fuzzy linguistic preference relations (HFLPRs). A decision support system (DSS) that can host these GDM algorithms to provide decision-support services or tools for practical use is urgently needed. However, the state-of-the-art architectures cannot organize these algorithms and related data to run within one framework. This is mainly due to the running environments for these GDM algorithms are different since these algorithms were not originally designed to be compatible. Given the multilingual consistency and consensus-based decision support algorithms, how to design and implement a cloud-based DSS in a multilingual environment is still an open question. To fill this gap, this paper provides a web-based and cloud-based DSS with a novel architecture that utilizes the advantages of microservices. The proposed system implements a multilingual support framework to dynamically upload, manage and run multilingual consistency and consensus-based decision support algorithms. An algorithm recommendation module is developed to help users choose suitable decision support algorithms. Tokenization is applied to deal with regulatory issues of knowledge protection, data privacy, and security while storing, analyzing, and transforming data into different algorithms for effective decision-making. An expert feedback study verified that our web and cloud-based DSS is a right artifact to fulfill the objective claimed in this paper.

The ageing of the population and the increasing number of patients with dementia in modern society undoubtedly put tremendous pressure on the medical system. Providing telehealth care for potential patients and patients with dementia can reduce the burden on both the health system and care-givers. This paper describes a telehealth framework for dementia early detection and dementia care. Specifically, we propose an improved deep neural network model for Non-Intrusive Load Monitoring (NILM), which disaggregates the household's overall energy usage into those of individual appliances based on the sequence-to-point model and transfer learning. The daily behaviour regularities of patients are then inferred by combining principal component analysis and$K$-means clustering based on the disaggregated appliance-level consumptions. Experiments show that the proposed model can significantly improve training efficiency and maintain load disaggregation accuracy, and the inferred behaviour regularities have great potential to be used as useful inputs and prior knowledge to the dementia condition detection platform for early detection and real-time monitoring of patient's conditions.

With the installation of smart meters, high resolution building-level energy consumption data become increasingly accessible, which not only provides more accurate data for energy forecasting at the aggregated level but also enables data-driven energy forecasting for individual buildings. On the one hand, individual buildings exhibit high randomness, making the forecasting problem at the building-level more challenging. On the other hand, buildings usually have their own characteristics, therefore such valuable information needs to be considered in the forecast models at the aggregation level. In this paper we investigate how unique characteristics of buildings could affect the performance of forecasting models and aim to identify defining patterns of buildings. The usefulness of the proposed approach is demonstrated using data from three real-world buildings.

Considering that it is difficult for experts to provide precise preference values for the site selection of electric vehicle charging station in risky environment, this paper develops an approach for linear ordinal ranking aggregation to validly improve the efficiency and accuracy of electric vehicle charging station site selection. At first, the inverse value function of prospect theory is applied to reduce the impact of risk. Then, through combining with the concept of information energy, the experts' weights can be derived. Besides, a consistency constraint is added to the individual ranking-based alternatives' weights deriving model, which can guarantee the consistency degree at an acceptable level. Additionally, a consensus and standard deviation-based model is established to aggregate the alternatives' weights. Finally, a numerical case about the electric vehicle charging station site selection is presented to show the usage of the approach, meanwhile, comparative analysis and sensitivity analysis are also conducted which show the robustness and practicability of the approach.

Computing estimated time of arrival (ETA) is one of the most important services for online ride-hailing platforms like DiDi and Uber. With billions of service queries per day on such platforms, a fast inference ETA module ensures the efficiency of the overall decision system to guarantee satisfied user experience, as well as saving significant operating cost. In this paper, we develop a novel ETA learning system named as CompactETA, which provides an accurate online travel time inference within 100 microseconds. In the proposed method, we encode high order spatial and temporal dependency into sophisticated representations by applying graph attention network on a spatiotemporal weighted road network graph. We further encode the sequential information of the travel route by positional encoding to avoid the recurrent network structure. The properly learnt representations enable us to apply a very simple multi-layer perceptron model for online real-time inference. Evaluation of both offline experiments and online A/B testing verifies that CompactETA reduces the inference latency by more than 100 times compared to a state-of-the-art system, while maintains competing prediction accuracy.

This paper proposes a hybrid approach to optimal day-ahead pricing for demand response management. At the customer-side, a comprehensive energy management system, which includes most commonly used appliances and an effective waiting time cost model is proposed to manage the energy usages in households (lower level problem). At the retailer-side, the best retail prices are determined to maximize the retailer's profit (upper level problem). The interactions between the electricity retailer and its customers can be cast as a bilevel optimization problem. To overcome the infeasibility of conventional Karush-Kuhn-Tucker (KKT) approach for this particular type of bilevel problem, a hybrid pricing optimization approach, which adopts the multi-population genetic algorithms for the upper level problem and distributed individual optimization algorithms for the lower level problem, is proposed. Numerical results show the applicability and effectiveness of the proposed approach and its benefit to the retailer and its customers by improving the retailer's profit and reducing the customers' bills.

In this paper, we propose a distributed optimization algorithm for the demand response management with a comprehensive customer demand modeling framework in the smart grid. Different from the existing literature, the considered demand modeling framework considers not only the energy management modeling but also the appliance-level usage pattern learning models, both for time-shiftable loads. More specific, a bill minimization based demand optimization model is firstly proposed for the customers choosing to use a home energy management software. Secondly, an appliance level probability behaviour model via calculating the probability distribution of different electricity consumption patterns in response to the dynamic prices is proposed for the customers choosing to manage their energy usages by themselves. Based on the optimization and learning results, we further propose a multi-population genetic algorithm based pricing optimization model for demand response management with the aim to maximize the retailer's profit and maximize customers' benefits. Numerical results indicate the applicability and effectiveness of the proposed models and its benefits.

Output consensus problems for multi-agent systems under disturbances are investigated. Distributed dynamic consensus protocols, which estimate the state and the disturbances of each agent and distribute the estimations to other neighbor agents, are applied, and output consensus problems under disturbances are converted into output regulation problems. On this basis, a necessary and sufficient condition for output consensus in the form of a simultaneous stabilization condition and the solvability of regulator equations which have the same dimensions as the dynamic model of a single agent is shown, and an analytical expression of the output consensus function is obtained. Further, a sufficient condition for output consensualizability is given and a procedure for control protocol design is summarized. Finally, theoretical results are demonstrated by a numerical simulation example.

MOTIVATION: miRNAs play crucial roles in human diseases and newly discovered could be targeted by small molecule (SM) drug compounds. Thus, the identification of small molecule drug compounds (SM) that target dysregulated miRNAs in cancers will provide new insight into cancer biology and accelerate drug discovery for cancer therapy. RESULTS: In this study, we aimed to develop a novel computational method to comprehensively identify associations between SMs and miRNAs. To this end, exploiting multiple molecular interaction databases, we first established an integrated SM-miRNA association network based on 690 561 SM to SM interactions, 291 600 miRNA to miRNA associations, as well as 664 known SM to miRNA targeting pairs. Then, by performing Random Walk with Restart algorithm on the integrated network, we prioritized the miRNAs associated to each of the SMs. By validating our results utilizing an independent dataset we obtained an area under the ROC curve greater than 0.7. Furthermore, comparisons indicated our integrated approach significantly improved the identification performance of those simple modeled methods. This computational framework as well as the prioritized SM-miRNA targeting relationships will promote the further developments of targeted cancer therapies. CONTACT: [email protected], [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

This paper proposes a real-time pricing scheme for demand response management in the context of smart grids. The electricity retailer determines the retail price first and announces the price information to the customers through the smart meter systems. According to the announced price, the customers automatically manage the energy use of appliances in the households by the proposed energy management system with the aim to maximize their own benefits. We model the interactions between the electricity retailer and its customers as a 1-leader, N-follower Stackelberg game. By taking advantage of the two-way communication infrastructure, the sequential equilibrium can be obtained through backward induction. At the followers' side, given the electricity price information, we develop efficient algorithms to maximize customers' satisfaction. At the leader's side, we develop a genetic algorithms based real-time pricing scheme by considering the expected customers' reactions to maximize retailer's profit. Experimental results indicate that the proposed scheme can not only benefit the retailers but also the customers.

In this paper, we propose two learning models to study the electricity consumption patterns of customers responding to real-time prices. We firstly divide home appliances into non-shift able appliances, shift able appliances and curtail able appliances according to their load types. Since non-shift able appliances have fixed operation routines, the consumption patterns of these appliances are obvious, thus need no learning. A learning model based on mean price ranking has been designed with the aim to study the electricity consumption patterns of using shift able appliances. For curtail able appliances, we propose a learning model based on multiple linear regression to learn the consumption pattern of customers. The simulation results in this paper show that the proposed learning models are feasible and efficient. Most importantly, this paper provides a new perspective for further research in learning and analysing the behaviour of electricity customers in the context of the smart grid.

MOTIVATION: Alzheimer's disease (AD) is a severe neurodegenerative disease of the central nervous system that may be caused by perturbation of regulatory pathways rather than the dysfunction of a single gene. However, the pathology of AD has yet to be fully elucidated. RESULTS: In this study, we systematically analyzed AD-related mRNA and miRNA expression profiles as well as curated transcription factor (TF) and miRNA regulation to identify active TF and miRNA regulatory pathways in AD. By mapping differentially expressed genes and miRNAs to the curated TF and miRNA regulatory network as active seed nodes, we obtained a potential active subnetwork in AD. Next, by using the breadth-first-search technique, potential active regulatory pathways, which are the regulatory cascade of TFs, miRNAs and their target genes, were identified. Finally, based on the known AD-related genes and miRNAs, the hypergeometric test was used to identify active pathways in AD. As a result, nine pathways were found to be significantly activated in AD. A comprehensive literature review revealed that eight out of nine genes and miRNAs in these active pathways were associated with AD. In addition, we inferred that the pathway hsa-miR-146a→STAT1→MYC, which is the source of all nine significantly active pathways, may play an important role in AD progression, which should be further validated by biological experiments. Thus, this study provides an effective approach to finding active TF and miRNA regulatory pathways in AD and can be easily applied to other complex diseases.

UNLABELLED: The inappropriate expression of microRNAs (miRNAs) is closely related with disease diagnosis, prognosis and therapy response. Recently, many studies have demonstrated that bioactive small molecules (or drugs) can regulate miRNA expression, which indicates that targeting miRNAs with small molecules is a new therapy for human diseases. In this study, we established the SM2miR database, which recorded 2925 relationships between 151 small molecules and 747 miRNAs in 17 species after manual curation from nearly 2000 articles. Each entry contains the detailed information about small molecules, miRNAs and evidences of their relationships, such as species, miRBase Accession number, DrugBank Accession number, PubChem Compound Identifier (CID), expression pattern of miRNA, experimental method, tissues or conditions for detection. SM2miR database has a user-friendly interface to retrieve by miRNA or small molecule. In addition, we offered a submission page. Thus, SM2miR provides a fairly comprehensive repository about the influences of small molecules on miRNA expression, which will promote the development of miRNA therapeutics. AVAILABILITY: SM2miR is freely available at http://bioinfo.hrbmu.edu.cn/SM2miR/.