Search

The goal of neural architecture search (NAS) is to either downsize the neural architecture and model of a deep neural network (DNN), adjust a neural architecture to improve its end result, or even speed up the whole training process. Such improvements make it possible to generate or install the model of a DNN on a small device, such as a device of internet of things or wireless sensor network. Because most NAS algorithms are time-consuming, finding out a way to reduce their computation costs has now become a critical research issue. The training-free method (also called the zero-shot learning) provides an alternative way to estimate how good a neural architecture is more efficiently during the process of NAS by using a lightweight score function instead of a general training process to avoid incurring heavy costs. This paper starts with a brief discussion of DNN and NAS, followed by a brief review of both model-dependent and model-independent training-free score functions. A brief introduction to the search algorithms and benchmarks that were widely used in a training-free NAS will also be given in this paper. The changes, potential, open issues, and future trends of this research topic are then addressed in the end of this paper.

Abstract Fog computing aims to mitigate data communication delay by deploying fog nodes to provide servers in the proximity of users and offload resource-hungry tasks that would otherwise be sent to distant cloud servers. In this paper, we propose an effective fog device deployment algorithm based on a new metaheuristic algorithm–search economics–to solve the optimization problem for the deployment of fog computing systems. The term “effective” in this paper refers to that the developed algorithm can achieve better performance in terms of metrics such as lower latency and less resource usage. Compared with conventional metaheuristic algorithms, the proposed algorithm is unique in that it first divides the solution space into a set of regions to increase search diversity of the search and then allocates different computational resources to each region according to its potential. To verify the effectiveness of the proposed algorithm, we compare it with several classical fog computing deployment algorithms. The simulation results indicate that the proposed algorithm provides lower network latency and higher quality of service than the other deployment algorithms evaluated in this study.

Abstract Urine test paper is a standard, noninvasive detection method for direct bilirubin, but this method can only achieve qualitative analysis and cannot achieve quantitative analysis. This study used Mini-LEDs as the light source, and direct bilirubin was oxidized to biliverdin by an enzymatic method with ferric chloride (FeCl3) for labeling. Images were captured with a smartphone and evaluated for red (R), green (G), and blue (B) colors to analyze the linear relationship between the spectral change of the test paper image and the direct bilirubin concentration. This method achieved noninvasive detection of bilirubin. The experimental results demonstrated that Mini-LEDs can be used as the light source to analyze the grayscale value of the image RGB. For the direct bilirubin concentration range of 0.1–2 mg/dL, the green channel had the highest coefficient of determination coefficient (R 2) of 0.9313 and a limit of detection of 0.56 mg/dL. With this method, direct bilirubin concentrations higher than 1.86 mg/dL can be quantitatively analyzed with the advantage of rapid and noninvasive detection.

The unlimited possibilities of machine learning have been shown in several successful reports and applications. However, how to make sure that the searched results of a machine learning system are not tampered by anyone and how to prevent the other users in the same network environment from easily getting our private data are two critical research issues when we immerse into powerful machine learning-based systems or applications. This situation is just like other modern information systems that confront security and privacy issues. The development of blockchain provides us an alternative way to address these two issues. That is why some recent studies have attempted to develop machine learning systems with blockchain technologies or to apply machine learning methods to blockchain systems. To show what the combination of blockchain and machine learning is capable of doing, in this paper, we proposed a parallel framework to find out suitable hyperparameters of deep learning in a blockchain environment by using a metaheuristic algorithm. The proposed framework also takes into account the issue of communication cost, by limiting the number of information exchanges between miners and blockchain.

It is an irrefutable fact that the Internet of Things (IoT) will eventually change our daily lives because its applications and relevant technologies have been or will be penetrating our daily lives. Also, the IoT is aimed to connect all the things (e.g., devices and systems) together via the Internet, thus making it easy to collect the data of users or environments and to find out useful information from the gathered data by using data mining technologies. As a consequence, how intelligent systems are developed for the IoT has become a critical research topic today. This means that artificial intelligence (AI) technologies (e.g., supervised learning, unsupervised learning, and semi-supervised learning) were used in the development of intelligent systems for analyzing the data captured from IoT devices or making decisions for IoT systems. It can be easily seen that AI can make an IoT system more intelligent and thus more accurate. For example, various sensors can be used for a smart home system to pinpoint the location and analyze the behavior of a human; however, with AI technologies, a more accurate prediction can be provided on the two pieces of information of a human. One of the most important uses for AI technologies is to make IoT systems more intelligent in order to provide a more convenient environment for users; thus, how to use existing AI technologies or develop new AI technologies to construct a better IoT system has attracted the attention of researchers from different disciplines in recent years. That is why, besides using existing supervised, unsupervised, semi-supervised learning algorithms, data mining algorithms, and machine learning algorithms, several recent studies have also attempted to develop new intelligent methods for the devices or systems for the IoT. All these approaches for making an IoT system more intelligent can also be found in the articles of this Special Section.

Metaheuristic algorithms have been widely used as an effective and efficient way for solving various complex optimization problems; there is, however, plenty of room for improvement. In this research area, the two most important issues that greatly influence the final results of single-solution-based metaheuristic algorithms are in that: (1) some of them are extremely sensitive to the initial solutions, and (2) some of them are easy to fall into a local optimum at early iterations. For these reasons, an effective framework, called multiple-search multi-start for single-solution-based metaheuristic algorithm (MSMS-S), is presented in this paper to mitigate the impact of these issues. MSMS-S ensures that a search procedure will be given different search directions based on the so-called re-start mechanism. To evaluate the performance of the proposed framework, we compare it with several well-known single-solution-based metaheuristic algorithms for clustering and codebook generation problems. Simulation results show that the proposed framework is capable of improving the performance of single-solution-based metaheuristic algorithms.

This report outlines a proposed method of packaging wide-angle (WA) mini-light-emitting diode (mini-LED) devices without optical lenses to create a highly efficient, ultrathin, flexible planar backlight for portable quantum dot light-emitting diode (QLED) displays. Since the luminous intensity curve for mini-LEDs generally recommends a beam angle of 120°, numerous LEDs are necessary to achieve a uniform surface light source for a QLED backlight. The light-guide layer and diffusion layer were packaged together on a chip surface to create WA mini-LEDs with a viewing angle of 180°. These chips were then combined with a quantum dot (QD) film and an optical film to create a high-efficiency, ultrathin, flexible planar light source with excellent color purity that can be used as a QLED display backlight. A 6 in (14.4 cm) light source was used as an experimental sample. When 1.44 W was supplied to the sample, the 3200-piece WA mini-LED with a flexible planar QLED display had a beam angle of 180° on the luminous intensity curve, a planar backlight thickness of 0.98 mm, a luminance of 10,322 nits, and a luminance uniformity of 92%.

The underlying concept of the Internet of Things (IoT) is simply to connect all devices and systems together via the Internet so that more suitable services can be provided to users. Many infrastructures, systems, and devices of the IoT have matured while some are still being developed. This is why several recent studies have claimed that IoT will dramatically change our lives. Today, we can find research topics driven by IoT technologies and can imagine that the era of smart homes and cities will be coming in the foreseeable future. The development of the IoT has reached a crossroad. One of the current research trends is to make this kind of system smarter, by using intelligent technologies to provide a much more convenient environment for humans. Among the intelligent technologies, how to handle the massive amount of data generated by the systems and devices of the IoT has been widely considered. Many other technologies, such as data mining, big data analytics, statistical and other analysis technologies, have also been used for analyzing data generated from the IoT. In addition to the analysis technologies, intelligent system technologies also provide many possibilities for the IoT because they can be used to enhance not only the performance of a system and its devices, but they can also be aware of events that have occurred.

Three different electrothermally-actuated MEMS micromirrors with Cr/Au-Si bimorph actuators are proposed. The devices are fabricated with the SOIMUMPs process developed by MEMSCAP, Inc. (Durham, NC, USA). A silicon-on-insulator MEMS process has been employed for the fabrication of these micromirrors. Electrothermal actuation has achieved a large angular movement in the micromirrors. Application of an external electric current 0.04 A to the bending-type, restricted-torsion-type, and free-torsion-type mirrors achieved rotation angles of 1.69°, 3.28°, and 3.64°, respectively.

This paper presents a method using multiobjective particle swarm optimization (PSO) approach to improve the consistency matrix in analytic hierarchy process (AHP), called PSOMOF. The purpose of this method is to optimize two objectives which conflict each other, while improving the consistency matrix. They are minimizing consistent ratio (CR) and deviation matrix. This study focuses on fuzzy preference matrix as one model comparison matrix in AHP. Some inconsistent matrices are repaired successfully to be consistent by this method. This proposed method offers some alternative consistent matrices as solutions.

Finding ways to distribute workloads to each processor core and efficiently reduce power consumption is of vital importance, especially for real-time systems. In this paper, a novel scheduling algorithm is proposed for real-time multicore systems to balance the computation loads and save power. The developed algorithm simultaneously considers multiple criteria, a novel factor, and task deadline, and is called power and deadline-aware multicore scheduling (PDAMS). Experiment results show that the proposed algorithm can greatly reduce energy consumption by up to 54.2% and the deadline times missed, as compared to the other scheduling algorithms outlined in this paper.

This paper presents a simple but efficient algorithm for reducing the computation time of genetic algorithm (GA) and its variants. The proposed algorithm is motivated by the observation that genes common to all the individuals of a GA have a high probability of surviving the evolution and ending up being part of the final solution; as such, they can be saved away to eliminate the redundant computations at the later generations of a GA. To evaluate the performance of the proposed algorithm, we use it not only to solve the traveling salesman problem but also to provide an extensive analysis on the impact it may have on the quality of the end result. Our experimental results indicate that the proposed algorithm can significantly reduce the computation time of GA and GA-based algorithms while limiting the degradation of the quality of the end result to a very small percentage compared to traditional GA.