• Journal of the Society of Naval Architects of Korea
• /
• v.59 no.2
• /
• pp.72-79
• /
• 2022

Deep-learning-based Super-Resolution (SR) methods were evaluated to reconstruct pressure fields with a high resolution from low-resolution images taken from a coarse grid simulation. In addition to a canonical SRCNN(super-resolution convolutional neural network) model, two modified models from SRCNN, adding an activation function (ReLU or Sigmoid function) to the output layer, were considered in the present study. High resolution images obtained by three models were more vivid and reliable qualitatively, compared with a conventional super-resolution method of bicubic interpolation. A quantitative comparison of statistical similarity showed that SRCNN model with Sigmoid function achieved best performance with less dependency on original resolution of input images.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.1
• /
• pp.123-132
• /
• 1999

In this paper, a quadratic sigmoid neural equalizer(QSNE) is proposed to improve the performance of conventional neural equalizer in terms of bit error probability by using a quadratic sigmoid function as the activation function of neural networks. Conventional neural equalizers which have been used to compensate for nonlinear distortions adopt the sigmoid function. In the case of sigmoid neural equalizer, each neuron has one linear decision boundary. So many neurons are required when the neural equalizer has to separate complicated structure. But in case of the proposed QSNF and quadratic sigmoid neural decision feedback equalizer(QSNDFE), each neuron separates decision region with two parallel lines. Therefore, QSNE and QSNDFE have better performance and simpler structure than the conventional neural equalizers in terms of bit error probability. When the proposed QSNDFE is applied to communication systems and digital magnetic recording systems, it is an improvement of approximately 1.5dB~8.3dB in signal to moise ratio(SNR) over the conventional decision feedback equalizer(DEF) and neural decision feedback equalizer(NDFE). As intersymbol interference(ISI) and nonlinear distortions become severer, QSNDFE shows astounding SNR shows astounding SNR performance gain over the conventional equalizers in the same bit error probability.

• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.39 no.8
• /
• pp.856-867
• /
• 1990

In this paper, the convergence and the learning accuracy of the back-propagation (BP) method in neural network are investigated by 1) analyzing the reason for decelerating the convergence of BP method and examining the rapid deceleration of the convergence when the learning is executed on the part of sigmoid activation function with the very small first derivative and 2) proposing the modified logistic activation function by defining, the convergence factor based on the analysis. Learning on the output patterns of binary as well as analog forms are tested by the proposed method. In binary output patter, the test results show that the convergence is accelerated and the learning accuracy is improved, and the weights and thresholds are converged so that the stability of neural network can be enhanced. In analog output patter, the results show that with extensive initial transient phenomena the learning error is decreased according to the convergence factor, subsequently the learning accuracy is enhanced.

• International Journal of Internet, Broadcasting and Communication
• /
• v.14 no.3
• /
• pp.143-148
• /
• 2022

In this paper, we employ deep neural network (DNN) to discern Android user interaction data distribution from artificial data distribution. We utilize real Android user interaction trace dataset collected from [1] to evaluate our DNN design. In particular, we use sequential model with 4 dense hidden layers and 1 dense output layer in TensorFlow and Keras. We also deploy sigmoid activation function for a dense output layer with 1 neuron and ReLU activation function for each dense hidden layer with 32 neurons. Our evaluation shows that our DNN design fulfills high test accuracy of at least 0.9955 and low test loss of at most 0.0116 in all cases of artificial data distributions.

• The Journal of the Convergence on Culture Technology
• /
• v.9 no.4
• /
• pp.563-568
• /
• 2023

ECG(electrocardiogram) is used to measure the rate and regularity of heartbeats, as well as the size and position of the chambers, the presence of any damage to the heart, and the cause of all heart diseases can be found. Because the ECG signal obtained using the ECG-KIT includes noise in the ECG signal, noise must be removed from the ECG signal to apply to the deep learning. In this paper, Noise included in the ECG signal was removed by using a lowpass filter of the Digital FIR Hamming window function. When the performance evaluation of the three activation functions, sigmoid(), ReLU(), and tanh() functions, which was confirmed that the activation function with the smallest error was the tanh() function, the elapsed time was longer when the batch size was small than large. Also, it was confirmed that result of the performance evaluation for the GRU model was superior to that of the LSTM model.

• Journal of Animal Science and Technology
• /
• v.65 no.3
• /
• pp.627-637
• /
• 2023

As the population and income levels rise, meat consumption steadily increases annually. However, the number of farms and farmers producing meat decrease during the same period, reducing meat sufficiency. Information and Communications Technology (ICT) has begun to be applied to reduce labor and production costs of livestock farms and improve productivity. This technology can be used for rapid pregnancy diagnosis of sows; the location and size of the gestation sacs of sows are directly related to the productivity of the farm. In this study, a system proposes to determine the number of gestation sacs of sows from ultrasound images. The system used the YOLOv7-E6E model, changing the activation function from sigmoid-weighted linear unit (SiLU) to a multi-activation function (SiLU + Mish). Also, the upsampling method was modified from nearest to bicubic to improve performance. The model trained with the original model using the original data achieved mean average precision of 86.3%. When the proposed multi-activation function, upsampling, and AutoAugment were applied, the performance improved by 0.3%, 0.9%, and 0.9%, respectively. When all three proposed methods were simultaneously applied, a significant performance improvement of 3.5% to 89.8% was achieved.

• ETRI Journal
• /
• v.46 no.5
• /
• pp.890-903
• /
• 2024

Anomaly detection in video surveillance is crucial but challenging due to the rarity of irregular events and ambiguity of defining anomalies. We propose a method called AONet that utilizes a spatiotemporal module to extract spatiotemporal features efficiently, as well as a residual autoencoder equipped with an attention network for effective future frame prediction in video anomaly detection. AONet utilizes a novel activation function called OptAF that combines the strengths of the ReLU, leaky ReLU, and sigmoid functions. Furthermore, the proposed method employs a combination of robust loss functions to address various aspects of prediction errors and enhance training effectiveness. The performance of the proposed method is evaluated on three widely used benchmark datasets. The results indicate that the proposed method outperforms existing state-of-the-art methods and demonstrates comparable performance, achieving area under the curve values of 97.0%, 86.9%, and 73.8% on the UCSD Ped2, CUHK Avenue, and ShanghaiTech Campus datasets, respectively. Additionally, the high speed of the proposed method enables its application to real-time tasks.

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.47.1-47
• /
• 2001

The drilling process has a great importance for the production technology due to its widerspread use in the manufacturing industry. In order to enhance a maximum production rate and prevent the drill from the damage, it is important to monitor and control the drilling system. Thrust force and cutting torque are the main output variables in the design of drilling control systems. In this paper, an alternative estimation method of thrust force by using flexible neural networks is proposed. Flexible neural network uses the sigmoid activation function with adjustable parameter in order to enhance the approximation accuracy ...

• Proceedings of the IEEK Conference
• /
• 1999.11a
• /
• pp.368-371
• /
• 1999

In this paper, we propose a neural network processor architecture with on-chip learning and with reconfigurability according to the data dependencies of the algorithm applied. For the neural network model applied, the proposed architecture can be configured into either SIMD or SRA(Systolic Ring Array) without my changing of on-chip configuration so as to obtain a high throughput. However, changing of system configuration can be controlled by user program. To process activation function, which needs amount of cycles to get its value, we design it by using PWL(Piece-Wise Linear) function approximation method. This unit has only single latency and the processing ability of non-linear function such as sigmoid gaussian function etc. And we verified the processing mechanism with EBP(Error Back-Propagation) model.

• The Journal of the Convergence on Culture Technology
• /
• v.9 no.4
• /
• pp.611-616
• /
• 2023

ECG(electrocardiogram) is a test used to measure the rate and regularity of heartbeats, as well as the size and position of the chambers, the presence of any damage to the heart, and the cause of all heart diseases can be found. Because the ECG signal obtained using the ECG-KIT includes noise in the ECG signal, noise must be removed from the ECG signal to apply to the deep learning. In this paper, the noise of the ECG signal was removed using the digital IIR Butterworth low-pass filter. When the performance evaluation of the three activation functions, sigmoid(), ReLU(), and tanh() functions, was compared using the deep learning model of LSTM, it was confirmed that the activation function with the smallest error was the tanh() function. Also, When the performance evaluation and elapsed time were compared for LSTM and GRU models, it was confirmed that the GRU model was superior to the LSTM model.