• International journal of advanced smart convergence
• /
• v.13 no.2
• /
• pp.80-87
• /
• 2024

This paper focuses on improving accuracy in constrained computing settings by employing the ReLU (Rectified Linear Unit) activation function. The research conducted involves modifying parameters of the ReLU function and comparing performance in terms of accuracy and computational time. This paper specifically focuses on optimizing ReLU in the context of a Multilayer Perceptron (MLP) by determining the ideal values for features such as the dimensions of the linear layers and the learning rate (Ir). In order to optimize performance, the paper experiments with adjusting parameters like the size dimensions of linear layers and Ir values to induce the best performance outcomes. The experimental results show that using ReLU alone yielded the highest accuracy of 96.7% when the dimension sizes were 30 - 10 and the Ir value was 1. When combining ReLU with the Adam optimizer, the optimal model configuration had dimension sizes of 60 - 40 - 10, and an Ir value of 0.001, which resulted in the highest accuracy of 97.07%.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.10
• /
• pp.3989-4006
• /
• 2020

The intelligent agriculture monitoring is based on the perception and analysis of environmental data, which enables the monitoring of the production environment and the control of environmental regulation equipment. As the scale of the application continues to expand, a large amount of data will be generated from the perception layer and uploaded to the cloud service, which will bring challenges of insufficient bandwidth and processing capacity. A fog-based offline and real-time hybrid data analysis architecture was proposed in this paper, which combines offline and real-time analysis to enable real-time data processing on resource-constrained IoT devices. Furthermore, we propose a data process-ing algorithm based on the incremental principal component analysis, which can achieve data dimensionality reduction and update of principal components. We also introduce the concept of Squared Prediction Error (SPE) value and realize the abnormal detection of data through the combination of SPE value and data fusion algorithm. To ensure the accuracy and effectiveness of the algorithm, we design a regular-SPE hybrid model update strategy, which enables the principal component to be updated on demand when data anomalies are found. In addition, this strategy can significantly reduce resource consumption growth due to the data analysis architectures. Practical datasets-based simulations have confirmed that the proposed algorithm can perform data fusion and exception processing in real-time on resource-constrained devices; Our model update strategy can reduce the overall system resource consumption while ensuring the accuracy of the algorithm.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.11 no.1
• /
• pp.101-106
• /
• 2010

Due to significant advances in Internet technology, there have been many e-learning courses offered by universities and academic institutes nowadays through the Internet. And these courses have benefited many students who might be constrained by distance and time. Nevertheless, most web-based courses are lecturing courses that cannot fulfill the needs for engineering technology education. In this paper, we propose the design and implementation of web-based programmable logic controller(PLC) laboratory to support learning and training for industrial automation. The proposed web-based PLC laboratory system consists of virtual labs and remote labs. This web-based PLC laboratory can be accessed by registered students to practice PLC experiments at their own home, enhancing the quality of education without much increasing in the overall cost. With the help of web cameras, the students can even have experience the live PLC experiments through the Internet.

• Journal of Information Processing Systems
• /
• v.16 no.1
• /
• pp.6-29
• /
• 2020

Biometrics identification using multiple modalities has attracted the attention of many researchers as it produces more robust and trustworthy results than single modality biometrics. In this paper, we present a novel multimodal recognition system that trains a deep learning network to automatically learn features after extracting multiple biometric modalities from a single data source, i.e., facial video clips. Utilizing different modalities, i.e., left ear, left profile face, frontal face, right profile face, and right ear, present in the facial video clips, we train supervised denoising auto-encoders to automatically extract robust and non-redundant features. The automatically learned features are then used to train modality specific sparse classifiers to perform the multimodal recognition. Moreover, the proposed technique has proven robust when some of the above modalities were missing during the testing. The proposed system has three main components that are responsible for detection, which consists of modality specific detectors to automatically detect images of different modalities present in facial video clips; feature selection, which uses supervised denoising sparse auto-encoders network to capture discriminative representations that are robust to the illumination and pose variations; and classification, which consists of a set of modality specific sparse representation classifiers for unimodal recognition, followed by score level fusion of the recognition results of the available modalities. Experiments conducted on the constrained facial video dataset (WVU) and the unconstrained facial video dataset (HONDA/UCSD), resulted in a 99.17% and 97.14% Rank-1 recognition rates, respectively. The multimodal recognition accuracy demonstrates the superiority and robustness of the proposed approach irrespective of the illumination, non-planar movement, and pose variations present in the video clips even in the situation of missing modalities.

• Journal of IKEEE
• /
• v.26 no.1
• /
• pp.62-66
• /
• 2022

Recently, there have been many research efforts based on data-based deep learning technologies to deal with the interference problem between heterogeneous wireless communication devices in unlicensed frequency bands. However, existing approaches are commonly based on the use of complex neural network models, which require high computational power, limiting their efficiency in resource-constrained network interfaces and Internet of Things (IoT) devices. In this study, we address the problem of classifying heterogeneous wireless technologies including Wi-Fi and ZigBee in unlicensed spectrum bands. We focus on a data-driven approach that employs a supervised-learning method that uses received signal strength indicator (RSSI) data to train Deep Convolutional Neural Networks (CNNs). We propose a simple measurement methodology for collecting RSSI training data which preserves temporal and spectral properties of the target signal. Real experimental results using an open-source 2.4 GHz wireless development platform Ubertooth show that the proposed sampling method maintains the same accuracy with only a 10% level of sampling data for the same neural network architecture.

• 한국사회정책
• /
• v.25 no.2
• /
• pp.95-124
• /
• 2018

This study aims to compare and analyze the work-based learning programs of the U.K. in order to find out the successful strategies of the Korea's youth internship which has not got employment effects. Furthermore, to seek a fundamental improvement plan of the Korea's work-based learning programs. The implications through comparative analysis are as follows. For internship performances, the curriculum should be standardized on the basis of industry-specific job criteria, which can be reflected by industrial demands and be correspond to needs at the workplace. Also, it is imperative that the curriculum is linked to national qualifications to guarantee the job ability. Additionally, an infrastructure for effectively monitoring, managing, and evaluating programs should be established. However, these strategies are still constrained in improving the job ability of vulnerable targets and fostering the future-leading manpower in the new industry. In order to reform the work-based learning program, it needs to be structured with a variety of programs suitable for different levels of participants by introducing traineeship, apprenticeship and further education.

• Journal of the Korean Society for Precision Engineering
• /
• v.21 no.4
• /
• pp.120-131
• /
• 2004

In this paper, we propose the method of a motion planning generation in which the movement of the 3-link leg subsystem is constrained to a slider-link and a singular posture can be easily avoided. The proposed method is the jumping control moving in vertical direction which mimics a cat's behavior. That is, it is jumping toward wall and kicking it to get a higher-place. Considering the movement from the point of constraint mechanical system, the robotic system which realizes the motion changes its configuration according to the position and it has several phases such as; ⅰ) an one-leg phase, ⅱ) in an air-phase. In other words, the system is under nonholonomic constraint due to the reservation of its momentum. Especially, in an air-phase, we will use a control method using state transformation and linearization in order to control the landing posture. Also, an iterative learning control algorithm is applied in order to improve the robustness of the control. The simulation results for jumping control will illustrate the effectiveness of the proposed control method.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.8
• /
• pp.3270-3294
• /
• 2020

Some things come easily to humans, one of them is the ability to navigate around. This capability of navigation suffers significantly in case of partial or complete blindness, restricting life activity. Advances in the technological landscape have given way to new solutions aiding navigation for the visually impaired. In this paper, we analyze the existing works and identify the challenges of path selection, context awareness, obstacle detection/identification and integration of visual and nonvisual information associated with real-time assisted mobility. In the process, we explore machine learning approaches for robotic path planning, multi constrained optimal path computation and sensor based wearable assistive devices for the visually impaired. It is observed that the solution to problem is complex and computationally intensive and significant effort is required towards the development of richer and comfortable paths for safe and smooth navigation of visually impaired people. We cannot overlook to explore more effective strategies of acquiring surrounding information towards autonomous mobility.

• ETRI Journal
• /
• v.41 no.4
• /
• pp.415-425
• /
• 2019

Generative adversarial networks (GAN) have been successfully used in many image restoration tasks, including image denoising, super-resolution, and compression artifact reduction. By fully exploiting its characteristics, state-of-the-art image restoration techniques can be used to generate images with photorealistic details. However, there are many applications that require faithful rather than visually appealing image reconstruction, such as medical imaging, surveillance, and video coding. We found that previous GAN-training methods that used a loss function in the form of a weighted sum of fidelity and adversarial loss fails to reduce fidelity loss. This results in non-negligible degradation of the objective image quality, including peak signal-to-noise ratio. Our approach is to alternate between fidelity and adversarial loss in a way that the minimization of adversarial loss does not deteriorate the fidelity. Experimental results on compression-artifact reduction and super-resolution tasks show that the proposed method can perform faithful and photorealistic image restoration.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2009.10a
• /
• pp.227-230
• /
• 2009

This paper presents an architecture of various devices convergence for ubiquitous network. Integration of a variety of devices be can connect to every kind of device should not be constrained. We construct PLC to UPnP protocol architecture and UPnP Bridge Module for interconnecting Non-IP devices with heterogeneous network interfaces to UPnP devices on UPnP networks.