• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.9
• /
• pp.1215-1230
• /
• 1995

The purpose of this paper is to build a computer vision system that endows an autonomous mobile robot the ability of automatic measuring of the analog and digital meters installed in nuclear power plant(NPP). This computer vision system takes a significant part in the organization of automatic surveillance and measurement system having the instruments and gadzets in NPP under automatic control situation. In the meter image captured by the camera, the meter area is sorted out using mainly the thresholding and the region labeling and the meter value recognition process follows. The positions and the angles of the needles in analog meter images are detected using the projection based method. In the case of digital meters, digits and points are extracted and finally recognized through the neural network classifier. To use available database containing relevant information about meters and to build fully automatic meter recognition system, the segmentation and recognition of the function-name in the meter printed around the meter area should be achieved for enhancing identification reliability. For thus, the function- name of the meter needs to be identified and furthermore the scale distributions and values are also required to be analyzed for building the more sophisticated system and making the meter recognition fully automatic.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.9
• /
• pp.49-61
• /
• 1996

Back-propagation(BP) algorithm needs a lot of time to train the artificial neural network (ANN) to get high accuracy level in classification tasks. So there have been extensive researches to process back-propagation algorithm on parallel processors. This paper prsents a linear systolic array which calculates forward-backward propagation of BP algorithm at the same time using effective space-time transformation and PE structure. First, we analyze data flow of forwared and backward propagations and then, represent the BP algorithm into data dapendency graph (DG) which shows parallelism inherent in the BP algorithm. Next, apply space-time transformation on the DG of ANN is turn with orthogonal direction projection. By doing so, we can get a snakelike systolic array. Also we calculate the interval of input for parallel processing, calculate the indices to make the right datas be used at the right PE when forward and bvackward propagations are processed in the same PE. And then verify the correctness of output when forward and backward propagations are executed at the same time. By doing so, the proposed system maximizes parallelism of BP algorithm, minimizes th enumber of PEs. And it reduces the execution time by 2 times through making idle PEs participate in forward-backward propagation at the same time.

• Journal of the Institute of Electronics Engineers of Korea TE
• /
• v.39 no.2
• /
• pp.17-28
• /
• 2002

Recently, the study on non-linear equalization, self-recovering equalization using the neural Network structure or Fuzzy logic, is lively in progress. In this thesis, if the value of error difference is large, coefficient adaptation rate is bigger, and if being small, it is smaller. We proposed the new FSG(Fuzzy Stochastic Gradient)/CMA algorithm combining TS(Tagaki-Sugeno) fuzzy model having fast convergence rate and low mean square error(MSE) and CMA(Constant Modulus Algorithm) which is prone to ISI and insensitive to phase alteration. As a simulation result of the designed channel adaptive equalizer using the proposed FSG/CMA algorithm, it is shown that SNR is improved about 3.5dB comparing to the conventional algorithm. 

• IEMEK Journal of Embedded Systems and Applications
• /
• v.15 no.4
• /
• pp.197-204
• /
• 2020

The ImageNet dataset is a large scale dataset and contains various natural scene images. In this paper, we propose a convolutional neural network (CNN)-based weighted ensemble technique for the ImageNet classification task. First, in order to fuse several models, our technique uses weights for each model, unlike the existing average-based ensemble technique. Then we propose an algorithm that automatically finds the coefficients used in later ensemble process. Our algorithm sequentially selects the model with the best performance of the validation set, and then obtains a weight that improves performance when combined with existing selected models. We applied the proposed algorithm to a total of 13 heterogeneous models, and as a result, 5 models were selected. These selected models were combined with weights, and we achieved 3.297% Top-5 error rate on the ImageNet test dataset.

• Journal of IKEEE
• /
• v.21 no.3
• /
• pp.309-319
• /
• 2017

The research investigates the inhibition of fatty acid biosynthesis of the human Acetyl-CoA Carboxylase enzyme by the aromatic-structure inhibitors (also known as ligands) containing variables of substituents, contributing an important role in the treatment of fatty-acid metabolic syndrome expressed by the group of cardiovascular risk factors increasing the incidence of coronary heart disease and type-2 diabetes. The effective interoperability between ligand and enzyme is characterized by a 50% concentration of enzyme inhibitor ($IC_{50}$) which was determined by experiment, and the factor of geometry structure of the ligands which are modeled by quantum mechanical methods using HyperChem 8.0.10 and Gaussian 09W softwares, combining with the calculation of quantum chemical and chemico-physical structural parameters using HyperChem 8.0.10 and Padel Descriptor 2.21 softwares. The result data are processed with the combination of classical statistical methods and modern bioinformatics methods using the statistical softwares of Department of Pharmaceutical Technology - Jadavpur University - India and R v3.3.1 software in order to accomplish a model of the quantitative structure - activity relationship between aromatic-structure ligands inhibiting fatty acid biosynthesis of the human Acetyl-CoA Carboxylase.

• International Journal of Advanced Culture Technology
• /
• v.5 no.1
• /
• pp.83-89
• /
• 2017

In this paper, we propose an extended method of one-hot Long Short-Term Memory (LSTM) and evaluate the performance on spam filtering task. Most of traditional methods proposed for spam filtering task use word occurrences to represent spam or non-spam messages and all syntactic and semantic information are ignored. Major issue appears when both spam and non-spam messages share many common words and noise words. Therefore, it becomes challenging to the system to filter correct labels between spam and non-spam. Unlike previous studies on information filtering task, instead of using only word occurrence and word context as in probabilistic models, we apply a neural network-based approach to train the system filter for a better performance. In addition to one-hot representation, using term weight with attention mechanism allows classifier to focus on potential words which most likely appear in spam and non-spam collection. As a result, we obtained some improvement over the performances of the previous methods. We find out using region embedding and pooling features on the top of LSTM along with attention mechanism allows system to explore a better document representation for filtering task in general.

• The Journal of Korean Association of Computer Education
• /
• v.9 no.2
• /
• pp.89-99
• /
• 2006

This paper proposes a computational model which can simulate peculiar language phenomena observed in human lexical decision task. The model is designed to mimic major language phenomena such as frequency effect, lexical status effect, word similarity, and semantic priming effect. The experimental results show that the propose model replicated the major language phenomena and performed similar performance with that of human in LDT.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.43 no.2
• /
• pp.33-38
• /
• 2020

Anomaly detection of Machine Learning such as PCA anomaly detection and CNN image classification has been focused on cross-sectional data. In this paper, two approaches has been suggested to apply ML techniques for identifying the failure time of big time series data. PCA anomaly detection to identify time rows as normal or abnormal was suggested by converting subjects identification problem to time domain. CNN image classification was suggested to identify the failure time by re-structuring of time series data, which computed the correlation matrix of one minute data and converted to tiff image format. Also, LASSO, one of feature selection methods, was applied to select the most affecting variables which could identify the failure status. For the empirical study, time series data was collected in seconds from a power generator of 214 components for 25 minutes including 20 minutes before the failure time. The failure time was predicted and detected 9 minutes 17 seconds before the failure time by PCA anomaly detection, but was not detected by the combination of LASSO and PCA because the target variable was binary variable which was assigned on the base of the failure time. CNN image classification with the train data of 10 normal status image and 5 failure status images detected just one minute before.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.2
• /
• pp.50-59
• /
• 1994

This paper presents a speech recognition method utilizing third-order cumulants as a feature vector and a neural network for recognition. The use of higher-order cumulants provides desirable uncoupling between the gaussian noise and speech, which enables us to estimate the coefficients of AR model without bias. Unlike the conventional method using second-order statistics, the proposed one exhibits low bias even in SNR as low as 0 dB at the expense of higher variance. It is confirmed through computer simulation that recognition rate of korean single-vowels with the cumulant-based method is much higher than the results with the conventional method even in low SNR.

• Journal of the Korean Chemical Society
• /
• v.60 no.4
• /
• pp.225-234
• /
• 2016

The activity of 34 sulfonamide derivatives has been estimated by means of multiple linear regression (MLR), artificial neural network (ANN), simulated annealing (SA) and genetic algorithm (GA) techniques. These models were also utilized to select the most efficient subsets of descriptors in a cross-validation procedure for non-linear -log (IC50) prediction. The results obtained using GA-ANN were compared with MLR-MLR, MLR-ANN, SA-ANN and GA-ANN approaches. A high predictive ability was observed for the MLR-MLR, MLR-ANN, SA-ANN and MLR-GA models, with root mean sum square errors (RMSE) of 0.3958, 0.1006, 0.0359, 0.0326 and 0.0282 in gas phase and 0.2871, 0.0475, 0.0268, 0.0376 and 0.0097 in solvent, respectively (N=34). The results obtained using the GA-ANN method indicated that the activity of derivatives of sulfonamides depends on different parameters including DP03, BID, AAC, RDF035v, JGI9, TIE, R7e+, BELM6 descriptors in gas phase and Mor 32u, ESpm03d, RDF070v, ATS8m, MATS2e and R4p, L1u and R3m in solvent. In conclusion, the comparison of the quality of the ANN with different MLR models showed that ANN has a better predictive ability.