Intelligent Collectives: Theory,Applications, and Research Challenges

ABSTRACT

Nowadays, along with the rapid development of information technologies, very often solving a common problem is entrusted to many autonomous units (people, systems). With such an approach, one can tap into the so-called collective intelligence (CI)—emerging from the collaboration and competition of many individuals. In this paper, we present recent research on CI related to the effectiveness of using the wisdom of crowds to perform a wide range of problems. For this aim, we first introduce a general framework of CI involving key characteristics of intelligent collectives. Next, we focus on the problem of how diversity and collective cardinality influence collective performance. Then, its applications, which are widely used such as prediction markets and Delphi method, will be presented. Furthermore, some research challenges on the capacity of combining CI with other research fields such as machine learning and social networks are also discussed.     

INCONSISTENCY OF KNOWLEDGE AND COLLECTIVE INTELLIGENCE

Is the intelligence of a collective larger than the intelligence of its members? How does one determine the knowledge of a collective on the basis of the knowledge of its members? In this paper, we try to answer these questions. Many examples show that the knowledge of a collective is not a usual union of the knowledge of its members. If we assume that the members of a collective possess their knowledge states about some common real world, and these states reflect to some degree the proper (real) state of the knowledge about the real world, then a question arises: How does one determine the knowledge of the collective, and what is its quality? In this paper, we use Consensus Theory to solve this problem and to show that, in many cases, the collective knowledge state is more proper than the knowledge states of the collective intelligence.     

Manufacturing production plan optimization in three-stage supply chains under Bass model market effects

The Bass model is a very successful parametric approach to forecast the diffusion process of new products. In recent years, applications of the Bass model have been extended to other operational research fields such as managing customer demands, controlling inventory levels, optimizing advertisement strategies, and so forth. This study attempts to establish an application for optimizing manufacturers’ production plans in a three-stage supply chain under the Bass model’s effects on the market. The supply chain structure considered in this research is similar to other common supply chains comprised of three stages, namely retailer, distributor and manufacturer. The retailer stage has to handle customer demands following the Bass diffusion process. Market parameters and essential information are assumed to be available and ready for access. Each stage is expected to determine its inventory policy rationally. That is, each stage will attempt to maximize its own profits. These decisions will back-propagate their effects to upper stages. This study adopts a dynamic programming approach to determine the inventory policies of each stage so as to optimize manufacturers’ production plans.     

A mathematical model to predict surface roughness and pattern thickness in roll-to-roll gravure printed electronics

The roll-to-roll printing processes have recently been applied for the manufacturing of printed electronics due to their advantages, such as their high-throughput capabilities and low associated costs. In a roll-to-roll printing machine, a web or substrate is moved from an unwinding roll to a winding roll. During printing, the operating tension is important for correct substrate handling to prevent substrate defects, such as wrinkles, scratches and breaks. Accordingly, the operating conditions of the moving web can affect the quality of the printed pattern. In this study, a mathematical model has been developed to predict the thickness and surface roughness of printed patterns. Because the dynamics of roll-to-roll printing systems are complicated and non-linear, a statistical model is preferred. A full factorial method has been used with four independent variables: operating tension, print speed, ink viscosity and theoretical transfer volume. This model accurately predicted the surface roughness and thickness of the printed pattern.     

A software-based dynamic-warp scheduling approach for load-balancing the Viola–Jones face detection algorithm on GPUs

Face detection is a key component in applications such as security surveillance and human–computer interaction systems, and real-time recognition is essential in many scenarios. The Viola–Jones algorithm is an attractive means of meeting the real time requirement, and has been widely implemented on custom hardware, FPGAs and GPUs. We demonstrate a GPU implementation that achieves competitive performance, but with low development costs. Our solution treats the irregularity inherent to the algorithm using a novel dynamic warp scheduling approach that eliminates thread divergence. This new scheme also employs a thread pool mechanism, which significantly alleviates the cost of creating, switching, and terminating threads. Compared to static thread scheduling, our dynamic warp scheduling approach reduces the execution time by a factor of 3. To maximize detection throughput, we also run on multiple GPUs, realizing 95.6 FPS on 5 Fermi GPUs.     

An optimized hybrid remote display protocol using GPU-assisted M-JPEG encoding and novel high-motion detection algorithm

In this paper, we design a novel hybrid remote display for mobile thin-client system. The remote frame buffer (RFB) protocol and motion JPEG (M-JPEG) protocol are assigned to handle remote display tasks in the slow-motion region and high-motion region, respectively. Graphic processing units (GPU) are utilized to do a part of a real-time JPEG compression task. A novel quality of experience (QoE)-based high-motion detection algorithm is also proposed to reduce the network bandwidth consumption and the server-side computing resource consumption. The continuity of screen delivery remains whenever the JPEG compression is applied to different screen regions. The proposed hybrid remote display approach has many good features which have been justified by comprehensive simulation studies.     

Enhancing the attacking efficiency of the node capture attack in WSN: a matrix approach

In the node capture attack, the adversary intelligently captures nodes and extracts the cryptographic keys from their memories to destroy the security, reliability and confidentiality of the wireless sensor networks. However, it suffers from low attacking efficiency and high resource expenditure. In this paper, we approach this attack from an adversarial view and develop a matrix-based method to model the process of the node capture attack. We establish a matrix to indicate the compromising relationship between the nodes and the paths. We propose a Matrix-based node capture attack Algorithm (MA in short), which can maximize the destructiveness while consuming the minimum resource expenditure. We conduct several experiments to show the performance of MA. Experimental results manifest that MA can reduce the attacking round, shorten the execution time, enhance the attacking efficiency and conserve the energy cost.     

Deflation-based power iteration clustering

Spectral clustering (SC) is currently one of the most popular clustering techniques because of its advantages over conventional approaches such as K-means and hierarchical clustering. However, SC requires the use of computing eigenvectors, making it time consuming. To overcome this limitation, Lin and Cohen proposed the power iteration clustering (PIC) technique (Lin and Cohen in Proceedings of the 27th International Conference on Machine Learning, pp. 655–662, 2010), which is a simple and fast version of SC. Instead of finding the eigenvectors, PIC finds only one pseudo-eigenvector, which is a linear combination of the eigenvectors in linear time. However, in certain critical situations, using only one pseudo-eigenvector is not enough for clustering because of the inter-class collision problem. In this paper, we propose a novel method based on the deflation technique to compute multiple orthogonal pseudo-eigenvectors (orthogonality is used to avoid redundancy). Our method is more accurate than PIC but has the same computational complexity. Experiments on synthetic and real datasets demonstrate the improvement of our approach.     

Uni-REPM: validated and improved

Software products are usually developed for either a specific customer (bespoke) or a broader market (market-driven). Due to their characteristic, bespoke and market-driven development face different challenges, especially concerning requirements engineering. Many challenges are caused by an inadequate requirements engineering process, and hence there is a need for process improvement frameworks based on empirical research and industry needs. In a previous article we introduced Uni-REPM, a lightweight requirements engineering process assessment framework based on a review of empirically motivated practices in market-driven and bespoke requirements engineering literature. In this article, we validate this framework in academia as well as industry, in order to prepare Uni-REPM for widespread industry use. We conduct two validations; a static validation based on interviews with seven academic experts and a dynamic validation where Uni-REPM is applied in four industrial organisations. Uni-REPM is refined according to the feedback obtained in the validations. The study shows that Uni-REPM is a quick, simple, and cost-effective solution to assess the maturity level of the requirements engineering process of projects. Moreover, the assessment method using checklists is highly usable and applicable in various international development environments.