• Journal of Broadcast Engineering
• /
• v.26 no.7
• /
• pp.855-861
• /
• 2021

Neural Architecture Search (NAS) is cutting-edge technology in the machine learning community. NAS Without Training (NASWOT) recently has been proposed to tackle the high demand of computational resources in NAS by leveraging some indicators to predict the performance of architectures before training. The advantage of these indicators is that they do not require any training. Thus, NASWOT reduces the searching time and computational cost significantly. However, NASWOT only considers high-performing networks which does not guarantee a fast inference speed on hardware devices. In this paper, we propose a multi objectives reward function, which considers the network's latency and the predicted performance, and incorporate it into the Reinforcement Learning approach to search for the best networks with low latency. Unlike other methods, which use FLOPs to measure the latency that does not reflect the actual latency, we obtain the network's latency from the hardware NAS bench. We conduct extensive experiments on NAS-Bench-201 using CIFAR-10, CIFAR-100, and ImageNet-16-120 datasets, and show that the proposed method is capable of generating the best network under latency constrained without training subnetworks.

• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.58 no.4
• /
• pp.511-515
• /
• 2009

For detecting of the faulty insulator, Correlation Propagation Neural Networks(CPNN) has been proposed. Faulty insulator is reduced the rate of insulation extremely, and taken the results dirty and injured. It is necessary to detect the faulty insulator and exchange the new one. And thus, we have designed the CPNN to be detected that insulators by the real time computation method through the inter-node diffusion. In the network, a node corresponds to a state in the quantized input space. Each node is composed of a processing unit and fixed weights from its neighbor nodes as well as its input terminal. Information propagates among neighbor nodes laterally and inter-node interpolation is achieved. 1-D CPNN hardware has been implemented with general purpose. Experiments with static and dynamic signals have been done upon the CPNN hardware. Through the results of simulation experiments, we define the ability of real-time detecting the faulty insulators.

• Smart Media Journal
• /
• v.11 no.2
• /
• pp.70-76
• /
• 2022

IoT application development using a cloud server causes problems such as data transmission and reception delay, network traffic, and cost for real-time processing support in network connected hardware. To solve this problem, edge cloud-based platforms can use neuromorphic hardware to enable fast data transfer. In this paper, we propose a model optimization method for supporting spiking neural networks on FPGA hardware. We focused on auto-adjusting network model parameters optimized for neuromorphic hardware. The proposed method performs optimization to show higher performance based on user requirements for accuracy. As a result of performance analysis, it satisfies all requirements of accuracy and showed higher performance in terms of expected execution time, unlike the naive method supported by the existing open source framework.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.55 no.11
• /
• pp.476-482
• /
• 2006

Based on a field-oriented model of induction motor, an adaptive backstepping control approach using neural networks is proposed in this paper for the speed control of induction motors with uncertainties at a minimum of information. Neural networks are used to approximate most of uncertainties which are derived from unknown motor parameters, load torque disturbances and unknown nonlinearities and an adaptive backstepping controller is used to derive adaptive law of neural networks and control input directly. The controller is implemented by the hardware using DSP and the effectiveness of the proposed approach is verified by carrying out the experiment.

• Journal of Sensor Science and Technology
• /
• v.31 no.1
• /
• pp.16-23
• /
• 2022

Cardiac disease is the most common cause of death worldwide. Therefore, detection and classification of electrocardiogram (ECG) signals are crucial to extend life expectancy. In this study, we aimed to implement an artificial intelligence signal recognition system in field programmable gate array (FPGA), which can recognize patterns of bio-signals such as ECG in edge devices that require batteries. Despite the increment in classification accuracy, deep learning models require exorbitant computational resources and power, which makes the mapping of deep neural networks slow and implementation on wearable devices challenging. To overcome these limitations, spiking neural networks (SNNs) have been applied. SNNs are biologically inspired, event-driven neural networks that compute and transfer information using discrete spikes, which require fewer operations and less complex hardware resources. Thus, they are more energy-efficient compared to other artificial neural networks algorithms.

• Electronics and Telecommunications Trends
• /
• v.35 no.3
• /
• pp.76-84
• /
• 2020

Neuromorphic hardware refers to brain-inspired computers or components that model an artificial neural network comprising densely connected parallel neurons and synapses. The major element in the widespread deployment of neural networks in embedded devices are efficient architecture for neuromorphic hardware with regard to performance, power consumption, and chip area. Spiking neural networks (SiNNs) are brain-inspired in which the communication among neurons is modeled in the form of spikes. Owing to brainlike operating modes, SNNs can be power efficient. However, issues still exist with research and actual application of SNNs. In this issue, we focus on the technology development cases and market trends of two typical tracks, which are listed above, from the point of view of artificial intelligence neuromorphic circuits and subsequently describe their future development prospects.

• Proceedings of the KIEE Conference
• /
• 1998.11b
• /
• pp.651-653
• /
• 1998

This thesis propose the prefilter type control strategies using modified chaotic neural networks #or the trajectory control of robotic manipulator. Since the structure of chaotic neural networks and neurons, chaotic neural networks can show the robust characteristics for controlling highly nonlinear dynamics like robotic manipulators. For its application, the trajectory controller of the three-axis PUMA robot is designed by CNN. The CNN controller acts as the compensator of the PD controller. Simulation results show that learning error decrease drastically via on- line learning and the performance is excellent. The CNN controller have much better controllability and shorter calculation time compared to the RNN controller. Another advantage of the proposed controller could be attached to conventional robot controller without hardware changes.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.50 no.4
• /
• pp.184-191
• /
• 2001

This paper proposes a prefilter type velocity compensating control system using modified chaotic neural networks for the trajectory control of robotic manipulator. Since the structure of modified chaotic neural networks(MCNN) and neurons have highly nonlinear dynamic characteristics, MCNN can show the robust characteristics for controlling highly nonlinear dynamics like robotic manipulators. For its application, the trajectory controller of the three-axis robot manipulator is designed by MCNN. The MCNN controller acts as the compensator of the PD controller. Simulation results show that learning error decrease drastically via on-line learning and the performance is excellent. The MCNN controller showed much better control performance and shorter calculation time compared to the RNN controller, Another advantage of the proposed controller could by attached to conventional robot controller without hardware changes.

• International Journal of Computer Science & Network Security
• /
• v.22 no.3
• /
• pp.61-66
• /
• 2022

Object classification is one of the main fields in neural networks and has attracted the interest of many researchers. Although there have been vast advancements in this area, still there are many challenges that are faced even in the current era due to its inefficiency in handling large data, linguistic and dimensional complexities. Powerful hardware and software approaches in Neural Networks such as Deep Neural Networks present efficient mechanisms and contribute a lot to the field of object recognition as well as to handle time series classification. Due to the high rate of accuracy in terms of prediction rate, a neural network is often preferred in applications that require identification, segmentation, and detection based on features. Neural networks self-learning ability has revolutionized computing power and has its application in numerous fields such as powering unmanned self-driving vehicles, speech recognition, etc. In this paper, the experiment is conducted to implement a neural approach to identify numbers in different formats without human intervention. Measures are taken to improve the efficiency of the machines to classify and identify numbers. Experimental results show the importance of having training sets to achieve better recognition accuracy.

• Smart Media Journal
• /
• v.12 no.4
• /
• pp.19-26
• /
• 2023

IoT edge services utilizing neuromorphic hardware architectures are suitable for autonomous IoT applications as they perform intelligent processing on the device itself. However, spiking neural networks applied to neuromorphic hardware are difficult for IoT developers to comprehend due to their complex structures and various hyper-parameters. In this paper, we propose a method for generating spiking neural network (SNN) models that satisfy user performance requirements while considering the constraints of neuromorphic hardware. Our proposed method utilizes previously trained models from pre-processed data to find optimal SNN model parameters from profiling data. Comparing our method to a naive search method, both methods satisfy user requirements, but our proposed method shows better performance in terms of runtime. Additionally, even if the constraints of new hardware are not clearly known, the proposed method can provide high scalability by utilizing the profiled data of the hardware.