• Journal of IKEEE
• /
• v.22 no.4
• /
• pp.1115-1122
• /
• 2018

In this paper, we analyze the performance degradation of previous 1-Bit convolutional neural network method and introduce ways to mitigate it. Previous work applies 32-Bit operation to first and last layers. But our method applies 32-Bit operation to second layer too. We also show that nonlinear activation function can be removed after binarizing inputs and weights. In order to verify the method proposed in this paper, we experiment the object detection neural network for korean license plate detection. Our method results in 96.1% accuracy, but the existing method results in 74% accuracy.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.26 no.1
• /
• pp.197-208
• /
• 2016

In this paper, we propose a method based on a convolutional neural network which is one of the deep neural network. So, we convert a malware code to malware image and train the convolutional neural network. In experiment with classify 9-families, the proposed method records a 96.2%, 98.7% of top-1, 2 error rate. And our model can classify 27 families with 82.9%, 89% of top-1,2 error rate.

• Journal of Biomedical Engineering Research
• /
• v.12 no.3
• /
• pp.191-202
• /
• 1991

This paper describes ECG data compression algorithm using neural network. As a learning method, we use back error propagation algorithm. ECG data compression is performed using learning ability of neural network. CSE database, which is sampled 12bit digitized at 500samp1e/sec, is selected as a input signal. In order to reduce unit number of input layer, we modify sampling ratio 250samples/sec in QRS complex, 125samples/sec in P & T wave respectively. hs a input pattern of neural network, from 35 points backward to 45 points forward sample Points of R peak are used.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.3 no.1
• /
• pp.7-12
• /
• 2003

In recent years, many attempts have been made to predict the behavior of bonds, currencies, stock or other economic markets. Most previous experiments used the neural network models for the stock market forecasting. The KOSPI 200 (Korea Composite Stock Price Index 200) is modeled by using different neural networks and fuzzy logic. In this paper, the neural network, the dynamic polynomial neural network (DPNN) and the fuzzy logic employed for the prediction of the KOSPI 200. The prediction results are compared by the root mean squared error (RMSE) and scatter plot, respectively. The results show that the performance of the fuzzy system is little bit worse than that of the DPNN but better than that of the neural network. We can develop the desired fuzzy system by optimization methods.

• Journal of Electrical Engineering and information Science
• /
• v.1 no.1
• /
• pp.108-117
• /
• 1996

In this paper, we propose a new type of recurrent neural network architecture in which each output unit is connected with itself and fully-connected with other output units and all hidden units. The proposed recurrent network differs from Jordan's and Elman's recurrent networks in view of functions and architectures because it was originally extended from the multilayer feedforward neural network for improving the discrimination and generalization power. We also prove the convergence property of learning algorithm of the proposed recurrent neural network and analyze the performance of the proposed recurrent neural network by performing recognition experiments with the totally unconstrained handwritten numeral database of Concordia University of Canada. Experimental results confirmed that the proposed recurrent neural network improves the discrimination and generalization power in recognizing spatial patterns.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.22 no.1
• /
• pp.1-8
• /
• 2018

Recently, the convolutional neural network shows high performance in many computer vision tasks. However, convolutional neural networks require enormous amount of operation, so it is difficult to adopt them in the embedded environments. To solve this problem, many studies are performed on the ASIC or FPGA implementation, where an efficient representation method is required. The fixed-point representation is adequate for the ASIC or FPGA implementation but causes a performance degradation. This paper proposes a separate optimization of representations for the convolutional layers and the batch normalization layers. With the proposed method, the required bit width for the convolutional layers is reduced from 16 bits to 10 bits for the ResNet-50 neural network. Since the computation amount of the convolutional layers occupies the most of the entire computation, the bit width reduction in the convolutional layers enables the efficient implementation of the convolutional neural networks.

• Management & Information Systems Review
• /
• v.3
• /
• pp.201-214
• /
• 1999

Adaptive correlator receiver with neural network based on complex multilayer perceptron is persented for suppressing interference of narrow-band of direct spread spectrum communication systems. Recursive least square algorithm with backpropagation error is used for fast convergence and better performance in adaptive correlator scheme. According to signal noise and transmission power, computer simulation results show that bit error ratio of adaptive correlator using neural network improved that of adative transversal filter of direct sequence spread spectrum considering of jamming and narrow-band interference. Bit error ratio of adaptive correlator with neural network is reduced about 10-1 than that of adaptive transversal filter where interference versus signal ratio is 5dB.

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1314-1320
• /
• 2019

In this paper, we propose a new method for optimizing bit operations and applying them to DNN(Deep Neural Network) in software environment. As a method for this, we propose a packing function for bitwise optimization and a masking matrix multiplication operation for application to DNN. The packing function converts 32-bit real value to 2-bit quantization value through threshold comparison operation. When this sequence is over, four 32-bit real values are changed to one 8-bit value. The masking matrix multiplication operation consists of a special operation for multiplying the packed weight value with the normal input value. And each operation was then processed in parallel using a GPU accelerator. As a result of this experiment, memory saved about 16 times than 32-bit DNN Model. Nevertheless, the accuracy was within 1%, similar to the 32-bit model.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.2 no.1
• /
• pp.109-116
• /
• 1998

In this paper, a back propagation neural network teaming algorithm based on the complex multilyer perceptron is represented for suppressing narrowband interference of the received signals in DS-SS mobile communication system. We proposed neural network adaptive correlator(NNAC) which has fast convergence rate and good performance with combining back propagation neural network and the receiver of DS-SS. We analyzed and proved that NNAC has lower bit error probability than that of traditional RAKE receiver through results of computer simulation in the presence of the tone and narrow-band interference and the co-channel interference.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.1 no.3
• /
• pp.77-82
• /
• 2008

In this paper, in the middle of progressing popular music chord, a method of inserting melody is addressed, which utilized by analyzing chord progress pattern. Firstly, a method for transforming melody into bit pattern which is to be used for neural network input is described. In order to insert the melody, composition pattern is learned from back propagation neural network, and based on these data new melody is to be generated. Experimental results verified the possibility of neural network based computer composition.