• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.2
• /
• pp.81-86
• /
• 2022

Given that time series are used in various fields, such as finance, IoT, and manufacturing, data analytical methods for accurate time-series forecasting can serve to increase operational efficiency. Among time-series analysis methods, multi-horizon forecasting provides a better understanding of data because it can extract meaningful statistics and other characteristics of the entire time-series. Furthermore, time-series data with exogenous information can be accurately predicted by using multi-horizon forecasting methods. However, traditional deep learning-based models for time-series do not account for the heterogeneity of inputs. We proposed an improved time-series predicting method, called the temporal fusion transformer method, which combines multi-horizon forecasting with interpretable insights into temporal dynamics. Various real-world data such as stock prices, fine dust concentrates and electricity consumption were considered in experiments. Experimental results showed that our temporal fusion transformer method has better time-series forecasting performance than existing models.

• ETRI Journal
• /
• v.33 no.3
• /
• pp.310-319
• /
• 2011

In this paper, a new soft-fusion approach for multiple-receiver wireless communication systems is proposed. In the proposed approach, each individual receiver provides the central receiver with a confidence level rather than a binary decision. The confidence levels associated with the local receiver are modeled by means of soft-membership functions. The proposed approach can be applied to wireless digital communication systems, such as amplitude shift keying, frequency shift keying, phase shift keying, multi-carrier code division multiple access, and multiple inputs multiple outputs sensor networks. The performance of the proposed approach is evaluated and compared to the performance of the optimal diversity, majority voting, optimal partial decision, and selection diversity in case of binary noncoherent frequency shift keying on a Rayleigh faded additive white Gaussian noise channel. It is shown that the proposed approach achieves considerable performance improvement over optimal partial decision, majority voting, and selection diversity. It is also shown that the proposed approach achieves a performance comparable to the optimal diversity scheme.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.9
• /
• pp.2927-2941
• /
• 2022

The primary task of machine fault diagnosis is to judge whether the current state is normal or damaged, so it is a typical binary classification problem with mutual exclusion. Mutually exclusive events and out-of-domain detection have one thing in common: there are two types of data and no intersection. We proposed a fusion model method to improve the accuracy of machine fault diagnosis, which is based on the mutual exclusivity of events and the commonality of out-of-distribution detection, and finally generalized to all binary classification problems. It is reported that the performance of a convolutional neural network (CNN) will decrease as the recognition type increases, so the variational auto-encoder (VAE) is used as the primary model. Two VAE models are used to train the machine's normal and fault sound data. Two reconstruction probabilities will be obtained during the test. The smaller value is transformed into a correction value of another value according to the mutually exclusive characteristics. Finally, the classification result is obtained according to the fusion algorithm. Filtering normal data features from fault data features is proposed, which shields the interference and makes the fault features more prominent. We confirm that good performance improvements have been achieved in the machine fault detection data set, and the results are better than most mainstream models.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.11
• /
• pp.4224-4243
• /
• 2021

Cognitive radio (CR) is a feasible intelligent technology and can be used as an effective solution to spectrum scarcity and underutilization. As the key function of CR, cooperative spectrum sensing (CSS) is able to effectively prevent the harmful interference with primary users (PUs) and identify the available spectrum resources by exploiting the spatial diversity of multiple secondary users (SUs). However, the open nature of the cognitive radio networks (CRNs) framework makes CSS face many security threats, such as, the malicious user (MU) launches Byzantine attack to undermine CRNs. For this aim, we make an in-depth analysis of the motive and purpose from the MU's perspective in the interweave CR system, aiming to provide the future guideline for defense strategies. First, we formulate a dynamic Byzantine attack model by analyzing Byzantine behaviors in the process of CSS. On the basis of this, we further make an investigation on the condition of making the fusion center (FC) blind when the fusion rule is unknown for the MU. Moreover, the throughput and interference to the primary network are taken into consideration to evaluate the impact of Byzantine attack on the interweave CR system, and then analyze the optimal strategy of Byzantine attack when the fusion rule is known. Finally, theoretical proofs and simulation results verify the correctness and effectiveness of analyses about the impact of Byzantine attack strategy on the throughput and interference.

• Korean Journal of Artificial Intelligence
• /
• v.11 no.3
• /
• pp.17-22
• /
• 2023

In this paper, we propose a black ice detection system using Convolutional Neural Networks (CNNs). Black ice poses a serious threat to road safety, particularly during winter conditions. To overcome this problem, we introduce a CNN-based architecture for real-time black ice detection with an encoder-decoder network, specifically designed for real-time black ice detection using thermal images. To train the network, we establish a specialized experimental platform to capture thermal images of various black ice formations on diverse road surfaces, including cement and asphalt. This enables us to curate a comprehensive dataset of thermal road black ice images for a training and evaluation purpose. Additionally, in order to enhance the accuracy of black ice detection, we propose a multi-scale dilation convolution feature fusion (MsDC-FF) technique. This proposed technique dynamically adjusts the dilation ratios based on the input image's resolution, improving the network's ability to capture fine-grained details. Experimental results demonstrate the superior performance of our proposed network model compared to conventional image segmentation models. Our model achieved an mIoU of 95.93%, while LinkNet achieved an mIoU of 95.39%. Therefore, it is concluded that the proposed model in this paper could offer a promising solution for real-time black ice detection, thereby enhancing road safety during winter conditions.

• Journal of Powder Materials
• /
• v.30 no.3
• /
• pp.210-216
• /
• 2023

In this study, machine learning models are proposed to predict the Vickers hardness of AlSi10Mg alloys fabricated by laser powder bed fusion (LPBF). A total of 113 utilizable datasets were collected from the literature. The hyperparameters of the machine-learning models were adjusted to select an accurate predictive model. The random forest regression (RFR) model showed the best performance compared to support vector regression, artificial neural networks, and k-nearest neighbors. The variable importance and prediction mechanisms of the RFR were discussed by Shapley additive explanation (SHAP). Aging time had the greatest influence on the Vickers hardness, followed by solution time, solution temperature, layer thickness, scan speed, power, aging temperature, average particle size, and hatching distance. Detailed prediction mechanisms for RFR are analyzed using SHAP dependence plots.

• Proceedings of the KIEE Conference
• /
• 1998.07b
• /
• pp.552-554
• /
• 1998

In target tracking problems the fixed gain Kalman filter is primarily used to predict a target state vector. This filter, however, has a poor precision for maneuvering targets while it has a good performance for non-maneuvering targets. To overcome the problem this paper proposes the system which estimates the acceleration with neural networks using the input estimation technique. The ability to efficiently fuse information of different forms is one of the major capabilities of trained multi-layer neural networks. The primary motivation for employing neural networks in these applications comes from the efficiency with which more features can be utilized as inputs for estimating target maneuvers. The parallel processing capability of a properly trained neural network can permit fast processing of features to yield correct acceleration estimates. The features used as inputs can be extracted from the combinations of innovation data and heading changes, and for this we set the two dimensional model. The properly trained neural network system outputs the acceleration estimates and compensates for the primary Kalman filter. Finally the proposed system shows the optimum performance.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.50 no.9
• /
• pp.430-438
• /
• 2001

In this study, we introduce the adaptive Polynomial Neuro-Fuzzy Networks(PNFN) architecture generated from the fusion of fuzzy inference system and PNN algorithm. The PNFN dwells on the ideas of fuzzy rule-based computing and neural networks. Fuzzy inference system is applied in the 1st layer of PNFN and PNN algorithm is employed in the 2nd layer or higher. From these the multilayer structure of the PNFN is constructed. In order words, in the Fuzzy Inference System(FIS) used in the nodes of the 1st layer of PNFN, either the simplified or regression polynomial inference method is utilized. And as the premise part of the rules, both triangular and Gaussian like membership function are studied. In the 2nd layer or higher, PNN based on GMDH and regression polynomial is generated in a dynamic way, unlike in the case of the popular multilayer perceptron structure. That is, the PNN is an analytic technique for identifying nonlinear relationships between system's inputs and outputs and is a flexible network structure constructed through the successive generation of layers from nodes represented in partial descriptions of I/O relatio of data. The experiment part of the study involves representative time series such as Box-Jenkins gas furnace data used across various neurofuzzy systems and a comparative analysis is included as well.

• Journal of information and communication convergence engineering
• /
• v.15 no.1
• /
• pp.7-13
• /
• 2017

We consider a distributed estimation where many nodes remotely placed at known locations collect the measurements of the parameter of interest, quantize these measurements, and transmit the quantized data to a fusion node; this fusion node performs the parameter estimation. Noting that quantizers at nodes should operate in a non-regular framework where multiple codewords or quantization partitions can be mapped from a single measurement to improve the system performance, we propose a low-weight estimation algorithm that finds the most feasible combination of codewords. This combination is found by computing the weighted sum of the possible combinations whose weights are obtained by counting their occurrence in a learning process. Otherwise, tremendous complexity will be inevitable due to multiple codewords or partitions interpreted from non-regular quantized data. We conduct extensive experiments to demonstrate that the proposed algorithm provides a statistically significant performance gain with low complexity as compared to typical estimation techniques.

• Journal of information and communication convergence engineering
• /
• v.14 no.4
• /
• pp.215-221
• /
• 2016

We consider a design of a combined quantizer and estimator for distributed systems wherein each node quantizes its measurement without any communication among the nodes and transmits it to a fusion node for estimation. Noting that the quantization partitions minimizing the estimation error are not independently encoded at nodes, we focus on the parameter regions created by the partitions and propose a cluster-based quantization algorithm that iteratively finds a given number of clusters of parameter regions with each region being closer to the corresponding codeword than to the other codewords. We introduce a new metric to determine the distance between codewords and parameter regions. We also discuss that the fusion node can perform an efficient estimation by finding the intersection of the clusters sent from the nodes. We demonstrate through experiments that the proposed design achieves a significant performance gain with a low complexity as compared to the previous designs.