• Title/Summary/Keyword: hidden unit number

Search Result 20, Processing Time 0.028 seconds

Recurrent Neural Network with Backpropagation Through Time Learning Algorithm for Arabic Phoneme Recognition

  • Ismail, Saliza;Ahmad, Abdul Manan
    • 제어로봇시스템학회:학술대회논문집
    • /
    • 2004.08a
    • /
    • pp.1033-1036
    • /
    • 2004
  • The study on speech recognition and understanding has been done for many years. In this paper, we propose a new type of recurrent neural network architecture for speech recognition, in which each output unit is connected to itself and is also fully connected to other output units and all hidden units [1]. Besides that, we also proposed the new architecture and the learning algorithm of recurrent neural network such as Backpropagation Through Time (BPTT, which well-suited. The aim of the study was to observe the difference of Arabic's alphabet like "alif" until "ya". The purpose of this research is to upgrade the people's knowledge and understanding on Arabic's alphabet or word by using Recurrent Neural Network (RNN) and Backpropagation Through Time (BPTT) learning algorithm. 4 speakers (a mixture of male and female) are trained in quiet environment. Neural network is well-known as a technique that has the ability to classified nonlinear problem. Today, lots of researches have been done in applying Neural Network towards the solution of speech recognition [2] such as Arabic. The Arabic language offers a number of challenges for speech recognition [3]. Even through positive results have been obtained from the continuous study, research on minimizing the error rate is still gaining lots attention. This research utilizes Recurrent Neural Network, one of Neural Network technique to observe the difference of alphabet "alif" until "ya".

  • PDF

Self-organized Learning in Complexity Growing of Radial Basis Function Networks

  • Arisariyawong, Somwang;Charoenseang, Siam
    • Proceedings of the IEEK Conference
    • /
    • 2002.07a
    • /
    • pp.30-33
    • /
    • 2002
  • To obtain good performance of radial basis function (RBF) neural networks, it needs very careful consideration in design. The selection of several parameters such as the number of centers and widths of the radial basis functions must be considered carefully since they critically affect the network's performance. We propose a learning algorithm for growing of complexity of RBF neural networks which is adapted automatically according to the complexity of tasks. The algorithm generates a new basis function based on the errors of network, the percentage of decreasing rate of errors and the nearest distance from input data to the center of hidden unit. The RBF's center is located at the point where the maximum of absolute interference error occurs in the input space. The width is calculated based on the standard deviation of distance between the center and inputs data. The steepest descent method is also applied for adjusting the weights, centers, and widths. To demonstrate the performance of the proposed algorithm, general problem of function estimation is evaluated. The results obtained from the simulation show that the proposed algorithm for RBF neural networks yields good performance in terms of convergence and accuracy compared with those obtained by conventional multilayer feedforward networks.

  • PDF

A Study on the Low Force Estimation of Skeletal Muscle by using ICA and Neuro-transmission Model (독립성분 분석과 신전달 모델을 이용한 근육의 미세한 힘의 추정에 관한 연구)

  • Yoo, Sae-Keun;Youm, Doo-Ho;Lee, Ho-Yong;Kim, Sung-Hwan
    • The Transactions of The Korean Institute of Electrical Engineers
    • /
    • v.56 no.3
    • /
    • pp.632-640
    • /
    • 2007
  • The low force estimation method of skeletal muscle was proposed by using ICA(independent component analysis) and neuro-transmission model. An EMG decomposition is the procedure by which the signal is classified into its constituent MUAP(motor unit action potential). The force index of electromyography was due to the generation of MUAP. To estimate low force, current analysis technique, such as RMS(root mean square) and MAV(mean absolute value), have not been shown to provide direct measures of the number and timing of motoneurons firing or their firing frequencies, but are used due to lack of other options. In this paper, the method based on ICA and chemical signal transmission mechanism from neuron to muscle was proposed. The force generation model consists of two linear, first-order low pass filters separated by a static non-linearity. The model takes a modulated IPI(inter pulse interval) as input and produces isometric force as output. Both the step and random train were applied to the neuro-transmission model. As a results, the ICA has shown remarkable enhancement by finding a hidden MAUP from the original superimposed EMG signal and estimating accurate IPI. And the proposed estimation technique shows good agreements with the low force measured comparing with RMS and MAV method to the input patterns.

A Study on Effective Digital Watermark Generation Method to Overcome Capacity Limit (저장 한계를 극복한 효율적인 디지털 워터마크 생성 방법 연구)

  • Kim Hee-Sun;Cho Dae-Jea
    • The Journal of the Korea Contents Association
    • /
    • v.5 no.6
    • /
    • pp.343-350
    • /
    • 2005
  • During the design of a successful digital watermarking systems, Pseudo-Noise(PN) sequences are widely used to modulate information bits into watermark signals. In this method, the number of bits that can be hidden within a small image by means of frequency domain watermarking is limited. In this paper, we show the possibility of introducing chaotic sequences into digital watermarking systems as potential substitutes to commonly used PN-sequences. And we propose a method that transforms the text to chaotic sequence. In our current implementation, we show how the sample text is expressed by an implied unit data(watermark) and the implied unit data is regenerated into the original left. Because we use this implied data as watermark for information hiding, we can insert much more watermark compared with previous method.

  • PDF

Elimination of Redundant Input Information and Parameters during Neural Network Training (신경망 학습 과정중 불필요한 입력 정보 및 파라미터들의 제거)

  • Won, Yong-Gwan;Park, Gwang-Gyu
    • The Transactions of the Korea Information Processing Society
    • /
    • v.3 no.3
    • /
    • pp.439-448
    • /
    • 1996
  • Extraction and selection of the informative features play a central role in pattern recognition. This paper describes a modified back-propagation algorithm that performs selection of the informative features and trains a neural network simultaneously. The algorithm is mainly composed of three repetitive steps : training, connection pruning, and input unit elimination. Afer initial training, the connections that have small magnitude are first pruned. Any unit that has a small number of connections to the hidden units is deleted,which is equivalent to excluding the feature corresponding to that unit.If the error increases,the network is retraned,again followed by connection pruning and input unit elimination.As a result,the algorithm selects the most im-portant features in the measurement space without a transformation to another space.Also,the selected features are the most-informative ones for the classification,because feature selection is tightly coupled with the classifi-cation performance.This algorithm helps avoid measurement of redundant or less informative features,which may be expensive.Furthermore,the final network does not include redundant parameters,i.e.,weights and biases,that may cause degradation of classification performance.In applications,the algorithm preserves the most informative features and significantly reduces the dimension of the feature vectors whiout performance degradation.

  • PDF

Long-Term Arrival Time Estimation Model Based on Service Time (버스의 정차시간을 고려한 장기 도착시간 예측 모델)

  • Park, Chul Young;Kim, Hong Geun;Shin, Chang Sun;Cho, Yong Yun;Park, Jang Woo
    • KIPS Transactions on Computer and Communication Systems
    • /
    • v.6 no.7
    • /
    • pp.297-306
    • /
    • 2017
  • Citizens want more accurate forecast information using Bus Information System. However, most bus information systems that use an average based short-term prediction algorithm include many errors because they do not consider the effects of the traffic flow, signal period, and halting time. In this paper, we try to improve the precision of forecast information by analyzing the influencing factors of the error, thereby making the convenience of the citizens. We analyzed the influence factors of the error using BIS data. It is shown in the analyzed data that the effects of the time characteristics and geographical conditions are mixed, and that effects on halting time and passes speed is different. Therefore, the halt time is constructed using Generalized Additive Model with explanatory variable such as hour, GPS coordinate and number of routes, and we used Hidden Markov Model to construct a pattern considering the influence of traffic flow on the unit section. As a result of the pattern construction, accurate real-time forecasting and long-term prediction of route travel time were possible. Finally, it is shown that this model is suitable for travel time prediction through statistical test between observed data and predicted data. As a result of this paper, we can provide more precise forecast information to the citizens, and we think that long-term forecasting can play an important role in decision making such as route scheduling.

Development of a model to predict Operating Speed (주행속도 예측을 위한 모형 개발 (2차로 지방부 도로 중심으로))

  • 이종필;김성호
    • Journal of Korean Society of Transportation
    • /
    • v.20 no.1
    • /
    • pp.131-139
    • /
    • 2002
  • This study introduces a developed artificial neural networks(ANN) model as a more efficient and reliable prediction model in operating speed Prediction with the 85th percentile horizontal curve of two-way rural highway in the aspect of evaluating highway design consistency. On the assumption that the speed is decided by highway geometry features, total 30 survey sites were selected. Data include currie radius, curve length, intersection angle, sight distance, lane width, and lane of those sites and were used as input layer data of the ANN. The optimized model structure was drawn by number of unit of hidden layer, learning coefficient, momentum coefficient, and change in learning frequency in multi-layer a ANN model. To verify learning Performance of ANN, 30 survey sites were selected while data in obtained from the 20 cites were used as learning data and those from the remaining 10 sites were used as predictive data. As a result of statistical verification, the model D of 4 types of ANN was evaluated as the most similar model to the actual operating speed value: R2 was 85% and %RMSE was 0.0204.

Development of a Freeway Travel Time Forecasting Model for Long Distance Section with Due Regard to Time-lag (시간처짐현상을 고려한 장거리구간 통행시간 예측 모형 개발)

  • 이의은;김정현
    • Journal of Korean Society of Transportation
    • /
    • v.20 no.4
    • /
    • pp.51-61
    • /
    • 2002
  • In this dissertation, We demonstrated the Travel Time forecasting model in the freeway of multi-section with regard of drives' attitude. Recently, the forecasted travel time that is furnished based on expected travel time data and advanced experiment isn't being able to reflect the time-lag phenomenon specially in case of long distance trip, so drivers don't believe any more forecasted travel time. And that's why the effects of ATIS(Advanced Traveler Information System) are reduced. Therefore, in this dissertation to forecast the travel time of the freeway of multi-section reflecting the time-lag phenomenon & the delay of tollgate, we used traffic volume data & TCS data that are collected by Korea Highway Cooperation. Also keep the data of mixed unusual to applicate real system. The applied model for forecasting is consisted of feed-forward structure which has three input units & two output units and the back-propagation is utilized as studying method. Furthermore, the optimal alternative was chosen through the twelve alternative ideas which is composed of the unit number of hidden-layer & repeating number which affect studying speed & forecasting capability. In order to compare the forecasting capability of developed ANN model. the algorithm which are currently used as an information source for freeway travel time. During the comparison with reference model, MSE, MARE, MAE & T-test were executed, as the result, the model which utilized the artificial neural network performed more superior forecasting capability among the comparison index. Moreover, the calculated through the particularity of data structure which was used in this experiment.

A Study on the Spatial Control Effect of Panjang in Donggwoldo (동궐도(東闕圖) 판장(板墻)의 공간통제 효과에 관한 연구)

  • HA Yujeong;KIM Choongsik
    • Korean Journal of Heritage: History & Science
    • /
    • v.55 no.4
    • /
    • pp.196-209
    • /
    • 2022
  • This study compared and analyzed the spatial division function and role of partitions by comparing the entire space and the spatial changes before and after the installation of partitions in <Donggwoldo>, which was manufactured in the late Joseon Dynasty. As a research method, a set standard was prepared to decompose the space of <Donggwoldo> into a unit space, and the standard was set according to the role and height of the space by classifying it into a main space, sub space, and transition space. Two convex maps were prepared according to before and after the installation of the Panjang, and the values of connectivity, control, and integration, which are spatial syntax variables, were calculated and analyzed. The results of the study are as follows. First, the partition in <Donggwoldo(東闕圖)> did not affect the overall spatial arrangement and control or connection of Donggwol, but the movement and access of space is limited to specific areas. Second, the partition was a facility intensively distributed in Naejeon(內殿) and Donggung(東宮) to be used actively in the way of space utilization. It shows that the unit space increased rapidly due to the installation of the partition. Since the partition was installed in the spaces that were open and under high control in the case of Naejeon(內殿), it helped to secure private spaces as closed ones under low control. On the other hand, for Donggung(東宮), the spaces were compartmented and divided with the partition to guide the movement path through narrow gates of the partition and increase the depth of the space. This helped to create spaces that are free and can be hidden as it increased the number of spaces coming through. Third, In addition to the functions of "eye blocking, space division, and movement path control" revealed in prior research, the partition has created a "space that is easy to control" within a specific area. The installation of the partition reduced the scale through the separation of spaces, but it occurred the expansion of the movement path and space. Also, the partition functioned to strengthen hiding and closure or increase openness as well through space division. This study is significant in that it revealed the value of the spatial control function of panjang through the analysis of spatial control and depth by analyzing the function of the partition with a mathematical model in addition to the analysis and study of the function and role of panjang. In addition, it is valuable in that it has prepared a framework for analysis tools that can be applied to traditional residential complexes similar to palaces by applying space syntax to <Donggungdo> to create convex spaces according to unit space division and connection types of palace architecture and landscape elements.

A Time Series Graph based Convolutional Neural Network Model for Effective Input Variable Pattern Learning : Application to the Prediction of Stock Market (효과적인 입력변수 패턴 학습을 위한 시계열 그래프 기반 합성곱 신경망 모형: 주식시장 예측에의 응용)

  • Lee, Mo-Se;Ahn, Hyunchul
    • Journal of Intelligence and Information Systems
    • /
    • v.24 no.1
    • /
    • pp.167-181
    • /
    • 2018
  • Over the past decade, deep learning has been in spotlight among various machine learning algorithms. In particular, CNN(Convolutional Neural Network), which is known as the effective solution for recognizing and classifying images or voices, has been popularly applied to classification and prediction problems. In this study, we investigate the way to apply CNN in business problem solving. Specifically, this study propose to apply CNN to stock market prediction, one of the most challenging tasks in the machine learning research. As mentioned, CNN has strength in interpreting images. Thus, the model proposed in this study adopts CNN as the binary classifier that predicts stock market direction (upward or downward) by using time series graphs as its inputs. That is, our proposal is to build a machine learning algorithm that mimics an experts called 'technical analysts' who examine the graph of past price movement, and predict future financial price movements. Our proposed model named 'CNN-FG(Convolutional Neural Network using Fluctuation Graph)' consists of five steps. In the first step, it divides the dataset into the intervals of 5 days. And then, it creates time series graphs for the divided dataset in step 2. The size of the image in which the graph is drawn is $40(pixels){\times}40(pixels)$, and the graph of each independent variable was drawn using different colors. In step 3, the model converts the images into the matrices. Each image is converted into the combination of three matrices in order to express the value of the color using R(red), G(green), and B(blue) scale. In the next step, it splits the dataset of the graph images into training and validation datasets. We used 80% of the total dataset as the training dataset, and the remaining 20% as the validation dataset. And then, CNN classifiers are trained using the images of training dataset in the final step. Regarding the parameters of CNN-FG, we adopted two convolution filters ($5{\times}5{\times}6$ and $5{\times}5{\times}9$) in the convolution layer. In the pooling layer, $2{\times}2$ max pooling filter was used. The numbers of the nodes in two hidden layers were set to, respectively, 900 and 32, and the number of the nodes in the output layer was set to 2(one is for the prediction of upward trend, and the other one is for downward trend). Activation functions for the convolution layer and the hidden layer were set to ReLU(Rectified Linear Unit), and one for the output layer set to Softmax function. To validate our model - CNN-FG, we applied it to the prediction of KOSPI200 for 2,026 days in eight years (from 2009 to 2016). To match the proportions of the two groups in the independent variable (i.e. tomorrow's stock market movement), we selected 1,950 samples by applying random sampling. Finally, we built the training dataset using 80% of the total dataset (1,560 samples), and the validation dataset using 20% (390 samples). The dependent variables of the experimental dataset included twelve technical indicators popularly been used in the previous studies. They include Stochastic %K, Stochastic %D, Momentum, ROC(rate of change), LW %R(Larry William's %R), A/D oscillator(accumulation/distribution oscillator), OSCP(price oscillator), CCI(commodity channel index), and so on. To confirm the superiority of CNN-FG, we compared its prediction accuracy with the ones of other classification models. Experimental results showed that CNN-FG outperforms LOGIT(logistic regression), ANN(artificial neural network), and SVM(support vector machine) with the statistical significance. These empirical results imply that converting time series business data into graphs and building CNN-based classification models using these graphs can be effective from the perspective of prediction accuracy. Thus, this paper sheds a light on how to apply deep learning techniques to the domain of business problem solving.