Search

Federated learning (FL) has been facilitating privacy-preserving deep learning in many walks of life such as medical image classification, network intrusion detection, and so forth. Whereas it necessitates a central parameter server for model aggregation, which brings about delayed model communication and vulnerability to adversarial attacks. A fully decentralized architecture like Swarm Learning allows peer-to-peer communication among distributed nodes, without the central server. One of the most challenging issues in decentralized deep learning is that data owned by each node are usually non-independent and identically distributed (non-IID), causing time-consuming convergence of model training. To this end, we propose a decentralized learning model called Homogeneous Learning (HL) for tackling non-IID data with a self-attention mechanism. In HL, training performs on each round’s selected node, and the trained model of a node is sent to the next selected node at the end of each round. Notably, for the selection, the self-attention mechanism leverages reinforcement learning to observe a node’s inner state and its surrounding environment’s state, and find out which node should be selected to optimize the training. We evaluate our method with various scenarios for two different image classification tasks. The result suggests that HL can achieve a better performance compared with standalone learning and greatly reduce both the total training rounds by 50.8% and the communication cost by 74.6% for decentralized learning with non-IID data.

For small and medium-sized enterprises and large enterprises alike, crowdsourcing innovation has become an important element of a product’s whole life cycle. It is the open call process of soliciting consumers to harvest and evaluate ideas or other intellectual assets. The previous proposed taxonomic framework for charactering this process is mainly for general crowdsourcing process and summarized by empirical study. The purpose of this paper is to propose a conceptual model for crowdsourcing innovation from a cybernetic and knowledge management perspective by normative research. The authors performed a normative study and deduced five systemic characteristics from the general laws of control system that guarantee ongoing efficiency for the innovation process. The normative research results provide two key contributions. Firstly, general control laws deduce five indispensable characteristics, and they reveal the intrinsic mechanism of crowdsourcing innovation: the knowledge flow controls, which are also the connotation of open innovation. Secondly, the authors have analyzed a five-characteristics system model and tested the model in several classical cases to show the design tricks of cases. This study provides a new conceptual framework that integrates the theory of open innovation and cybernetics to provide a new view of crowdsourcing innovation process design. In practice, this framework guides managers through the design criteria needed to implement a success crowdsourcing process.

Distributed intelligence is a well-known approach for optimizing interactions among numerous smart devices that interconnect and operate together as Internet of Things (IoT) systems. A modern form of human-machine collective intelligence emerges when humans interact with IoT systems in sociotechnical environments such as smart homes. Fifth-generation (5G) communication networks are designed for high-speed reliable wireless connectivity and expected to boost IoT and (distributed) collective intelligence by revolutionizing human–device–human interactions. In this paper, we contribute a comprehensive review of state-of-the-art sociotechnical environments that exhibit collective intelligence, supported by 5G-enabled IoT. We discuss the latest developments in 5G and their implications for collective intelligence. Further, we explain the key challenges for using 5G to support collective intelligence, e.g., data processing, security, and radio resource management. Finally, we describe four practical applications of collective intelligence to sociotechnical environments—road traffic control, unmanned aerial vehicles, electrical load demand response, and augmented democracy.

A vehicular fog network is an emerging paradigm adopted to facilitate delay-sensitive and innovative applications. Since vehicular environments are inherently dynamic, it becomes a challenge to effectively utilize all available resources. Often a centralized resource distribution model is adopted for effective resource utilization but this comes with significant overhead. Thus, a distributed model seems more suited for highly dynamic vehicular environments; however, a distributed model without collective intelligence can end up in uneven workload distribution. In this paper, we propose a distributed non-cooperative task offloading framework for efficient resource utilization. Here, vehicles with heterogeneous task requirements interact with one another without directly influencing the actions of the neighboring vehicles. That is, the framework allows vehicles to communicate with neighbors to gather contextual information to revisit their offloading decisions. The shared information includes resource type, task residence time, system cost, and offloading inference. The effectiveness of the framework is evaluated against baseline schemes in terms of service delay, transmission delay, system cost, delivery rate, and system efficiency. The results demonstrate significantly reduced task residence times by 50%, highest throughput at 83 Mbits/s, which in turn contributes to an improved system utilization up to 70% across the resource-sharing network.

Mobile crowdsensing (MCS) is a well-established paradigm that leverages mobile devices’ ubiquitous nature and processing capabilities for large-scale data collection to monitor phenomena of common interest. Crowd-powered data collection is significantly faster and more cost-effective than traditional methods. However, it poses challenges in assessing the accuracy and extracting information from large volumes of user-generated data. SmartRoadSense (SRS) is an MCS technology that utilises sensors embedded in mobile phones to monitor the quality of road surfaces by computing a crowdsensed road roughness index (referred to as $P_{PE}$ ). The present work performs statistical modelling of $P_{PE}$ to analyse its distribution across the road network and elucidate how it can be efficiently analysed and interpreted. Joint statistical analysis of open datasets is then carried out to investigate the effect of both internal and external road features on $P_{PE}$ . Several road properties affecting $P_{PE}$ as predicted are identified, providing evidence that SRS can be effectively applied to assess road quality conditions. Finally, the effect of road category and the speed limit on the mean and standard deviation of $P_{PE}$ is evaluated, incorporating previous results on the relationship between vehicle speed and $P_{PE}$ . These results enable more effective and confident use of the SRS platform and its data to help inform road construction and renovation decisions, especially where a lack of resources limits the use of conventional approaches. The work also exemplifies how crowdsensing technologies can benefit from open data integration and highlights the importance of making coherent, comprehensive, and well-structured open datasets available to the public.

The advantage of human-machine collective intelligence for decision support systems is the ability to make better decisions due to the mitigation of human biases in the generation of potential solutions and their evaluation. So far, the potential of human-machine collective intelligence was used only in few decision support systems, however, teamwork between humans and machines has not been achieved. This is partly due to the lack of interoperability in these systems. In earlier works, the authors proposed the apparatus of multi-aspect ontologies implying the integration of multiple domain ontologies to provide interoperability between humans and machines and coordinate interrelated processes going on in the systems of the considered type. Such ontologies have proved efficient for systems that require intensive information and knowledge exchange between loosely-related dynamic autonomous domains (e.g., enterprise knowledge management, product lifecycle management, or human-machine collective intelligence systems). However, existing ontology development methodologies fail to recommend a process that would support cross-domain knowledge integration during the multi-aspect ontology development. Moreover, the structure of the multi-aspect ontology imposes some restrictions on the integration approach. The paper proposes such a methodology for the multi-aspect ontology development that incorporates the aspect integration approach at multiple levels. The methodology is applied to develop a multi-aspect ontology for decision support based on human-machine collective intelligence. An example from the “e-tourism” domain demonstrates the applicability of the proposed methodology as well as the usage of the multi-aspect ontology for a human-machine environment aimed at solving real-world problems. The proposed methodology can facilitate the development of ontologies for complex knowledge-based systems that operate with knowledge from multiple loosely-connected domains.

Decision making is one of the most frequently performed processes nowadays. Our future or other people’s future may depend on the decision we make, therefore we try to take it as good as possible. The rapid development of technology facilitates it for us, as we are able to store, process and analyse more and more data. However, the power of human knowledge extended by experience and intuition still overperforms computer systems in solving such problems. The power of the combined knowledge of a group of people is known as a collective intelligence. In this paper, an analysis of one of the properties describing the consensus is presented, id est the influence between its members. The work is based on graph theory and a multi-agent system approach to formalise the problem. The new measures describing the collective and the accuracy of the final prediction were proposed. Experimental evaluations, based on the developed environment, followed by the statistical analysis of the obtained results allowed to observed four relationships between the considered measures describing the collective and the accuracy of collective prediction. The independence of the collective members plays a crucial role in the decision-making, therefore it should be carefully considered when selecting people for the collective, especially the large ones.

Collective intelligence is manifested when multiple agents coherently work in observation, interaction, decision-making and action. In this paper, we define and quantify the intelligence level of heterogeneous agents group organized in a flat structure without explicit leadership by the improved Anytime Universal Intelligence Test (AUIT), an extension of the existing evaluation of homogeneous agents group. The relationship of intelligence level with agents' composition, group size, spatial complexity and testing time is analyzed. The intelligence level of heterogeneous agents groups is compared with that for the homogeneous groups as well, so as to demonstrate the effects of heterogeneity on collective intelligence. Our work will contribute to understand the essence of collective intelligence more deeply and reveal the effect of various key factors on group intelligence level.

Nowadays, to solve a problem, people/systems typically use knowledge from different sources. A binary vector is a useful structure to represent knowledge states, and determining the consensus for a binary vector collective is helpful in many areas. However, determining a consensus that satisfies postulate 2-Optimality is an NP-hard problem; therefore, many heuristic algorithms have been proposed. The basic heuristic algorithm is the fastest in the literature, and most widely used to solve this problem. The computational complexity of the basic heuristic algorithm is O(m2n). In this study, we propose a quick algorithm (called QADC) to determine the 2-Optimality consensus. The QADC algorithm is developed based on a new approach for calculating the distances from a candidate consensus to the collective members. The computational complexity of the QADC algorithm has been reduced to O(mn), and the consensus quality of QADC algorithm and the basic heuristic algorithm is the same.

Collective intelligence projects based on citizen participation are gaining momentum in today's society. Citizen science applies crowdsourcing techniques to produce reliable data, quickly and easily. These projects allow getting new knowledge and help professional scientists to come to real conclusions. This paper proposes that the use of reputation systems improves the results obtained in citizen science projects. To prove this hypothesis, a reputation system is applied to a real experiment and the results are analyzed. The goal of the experiment is to calculate the real-time solar activity, known as the Wolf number, using the infrastructure and user community of the GLORIA project (a set of professional robotic telescopes running since 2013). The sample size of the study are 196 end-users and 2,108 executions of the experiment. The key findings presented in the paper are: 1) the online experiment with volunteers correctly reproduces the traditional method of the year 1848 performed by astronomers or advanced amateurs, 2) the model is contrasted and validated with the values published by the official organization, and 3) the reputation system reduces the error in calculations by more than half, discarding the contributions of the users with lowest karma.

ACOR is a well-established ant colony optimization algorithm for continuous-domain optimization. We present an approach for the dynamic adaptation of the ACOR algorithm's controlling parameters, focusing on the search width parameter, based on using several pre-specified parameter configurations, which we call personalities. Before an ant starts to generate a candidate solution, it stochastically adopts a personality based on the relative past success of different personalities. The success of a personality is measured, in turn, by the survival rate of the previous solutions generated by the ants adopting that personality. The premise of our approach is that some personalities will be more appropriate than others for different phases of the search. In addition, our adaptive approach can accommodate solution recombination, the use of which within ACOR was recently explored in the previous work. It allows the frequency of applying recombination and the type of recombination operator to be dynamically adapted, by having one or more recombination personalities among the competing personalities. We evaluate these proposals experimentally on two applications: 1) training feedforward neural networks for classification using 65 benchmark datasets from the University of California Irvine repository and 2) optimizing several popular synthetic benchmark continuous-domain functions. Our experimental results indicate that our proposals perform better than the standard ACOR on both applications, to a statistically significant extent.

Collective intelligence (CI) refers to the intelligence that emerges at the macro-level of a collection and transcends that of the individuals. CI is a continuously popular research topic that is studied by researchers in different areas, such as sociology, economics, biology, and artificial intelligence. In this survey, we summarize the works of CI in various fields. First, according to the existence of interactions between individuals and the feedback mechanism in the aggregation process, we establish CI taxonomy that includes three paradigms: isolation, collaboration and feedback. We then conduct statistical literature analysis to explain the differences among three paradigms and their development in recent years. Second, we elaborate the types of CI under each paradigm and discuss the generation mechanism or theoretical basis of the different types of CI. Third, we describe certain CI-related applications in 2019, which can be appropriately categorized by our proposed taxonomy. Finally, we summarize the future research directions of CI under each paradigm. We hope that this survey helps researchers understand the current conditions of CI and clears the directions of future research.