• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.5 no.1
• /
• pp.52-58
• /
• 2005

We have developed a Group Method of Data Handling (GMDH) type algorithm for designing multi-layered neural networks. The algorithm is general enough that it will accept any number of inputs and any sized training set. Each neuron of the resulting network is a function of two of the inputs to the layer. The equation for each of the neurons is a quadratic polynomial. Several forms of the equation are tested for each neuron to make sure that only the best equation of two inputs is kept. All possible combinations of two inputs to each layer are also tested. By carefully testing each resulting neuron, we have developed an algorithm to keep only the best neurons at each level. The algorithm's goal is to create as accurate a network as possible while minimizing the size of the network. Software was developed to train and simulate networks using our algorithm. Several applications were modeled using our software, and the result was that our algorithm succeeded in developing small, accurate, multi-layer networks.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.12
• /
• pp.6038-6053
• /
• 2017

In this study, we propose an automatic melody extraction algorithm using deep learning. In this algorithm, feature images, generated using the energy of frequency band, are extracted from polyphonic audio files and a deep learning technique, a convolutional neural network (CNN), is applied on the feature images. In the training data, a short frame of polyphonic music is labeled as a musical note and a classifier based on CNN is learned in order to determine a pitch value of a short frame of audio signal. We want to build a novel structure of melody extraction, thus the proposed algorithm has a simple structure and instead of using various signal processing techniques for melody extraction, we use only a CNN to find a melody from a polyphonic audio. Despite of simple structure, the promising results are obtained in the experiments. Compared with state-of-the-art algorithms, the proposed algorithm did not give the best result, but comparable results were obtained and we believe they could be improved with the appropriate training data. In this paper, melody extraction and the proposed algorithm are introduced first, and the proposed algorithm is then further explained in detail. Finally, we present our experiment and the comparison of results follows.

• Convergence Security Journal
• /
• v.18 no.4
• /
• pp.81-87
• /
• 2018

Military training is being carried out to win the combats through the quick and accurate response exercise for changeable engagement situations on battle fields. However, practically, it is really difficult to do actual fight training. Even though ROK Army is doing effort for practical training in KCTC(Korea army advanced Combat Training Center) supplying such as MILES equipments but a single platoon is able to use KCTC facilities or MILES equipments only 10 days a year. In order to find solution on this problem many researches suggesting AR or VR technology are still on the way. Nevertheless these are not fully covered the training done in the real field. In this regard, this paper proposes how the AR technology algorithm to apply on small units during field training exercise.

• The KIPS Transactions:PartB
• /
• v.14B no.4
• /
• pp.255-262
• /
• 2007

This paper proposes an enhanced skin color-based detection method to find a region of human face in color images. The proposed detection method combines three color spaces, RGB, $YC_bC_r$, YIQ and builds color union histograms of luminance and chrominance components respectively. Combined color union histograms are then fed in to the back-propagation neural network for training and Levenberg-Marquadt algorithm is applied to the iteration process of training. Proposed method with Levenberg-Marquadt algorithm applied to training process of neural network contributes to solve a local minimum problem of back-propagation neural network, one of common methods of training for face detection, and lead to make lower a detection error rate. Further, proposed color-based detection method using combined color union histograms which give emphasis to chrominance components divided from luminance components inputs more confident values at the neural network and shows higher detection accuracy in comparison to the histogram of single color space. The experiments show that these approaches perform a good capability for face region detection, and these are robust to illumination conditions.

• Journal of KSNVE
• /
• v.4 no.1
• /
• pp.59-69
• /
• 1994

A new LMS algorithm titled constrained LMS' is proposed for problems with constrained structure. The conventional LMS algorithm can not be used because it destroys the constrained structures of the weights or parameters. Proposed method uses error-back propagation, which is popular in training neural networks, for error minimization. The illustrative examplesare shown to demonstrate the applicability of the proposed algorithm.

• ETRI Journal
• /
• v.32 no.3
• /
• pp.486-489
• /
• 2010

This letter presents a modified Pegasos algorithm for graph-based dependency parsing. We show that the modified Pegasos algorithm outperforms the margin infused relaxed and passive aggressive algorithms in three languages.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.9
• /
• pp.125-139
• /
• 2014

In this paper, I propose a response modeling with a Semi-Supervised Support Vector Regression (SS-SVR) algorithm. In order to increase the accuracy and profit of response modeling, unlabeled data in the customer dataset are used with the labeled data during training. The proposed SS-SVR algorithm is designed to be a batch learning to reduce the training complexity. The label distributions of unlabeled data are estimated in order to consider the uncertainty of labeling. Then, multiple training data are generated from the unlabeled data and their estimated label distributions with oversampling to construct the training dataset with the labeled data. Finally, a data selection algorithm, Expected Margin based Pattern Selection (EMPS), is employed to reduce the training complexity. The experimental results conducted on a real-world marketing dataset showed that the proposed response modeling method trained efficiently, and improved the accuracy and the expected profit.

• Speech Sciences
• /
• v.5 no.1
• /
• pp.7-21
• /
• 1999

This paper proposes an improved training procedure in speech recognition based on the continuous density of the Hidden Markov Model (CDHMM). Of the three parameters (initial state distribution probability, state transition probability, output probability density function (p.d.f.) of state) governing the CDHMM model, we focus on the third parameter and propose an efficient algorithm that determines the p.d.f. of each state. It is known that the resulting CDHMM model converges to a local maximum point of parameter estimation via the iterative Expectation Maximization procedure. Specifically, we propose two independent algorithms that can be embedded in the segmental K -means training procedure by replacing relevant key steps; the adaptation of the number of mixture Gaussian p.d.f. and the initialization using the CDHMM parameters previously estimated. The proposed adaptation algorithm searches for the optimal number of mixture Gaussian humps to ensure that the p.d.f. is consistently re-estimated, enabling the model to converge toward the global maximum point. By applying an appropriate threshold value, which measures the amount of collective changes of weighted variances, the optimized number of mixture Gaussian branch is determined. The initialization algorithm essentially exploits the CDHMM parameters previously estimated and uses them as the basis for the current initial segmentation subroutine. It captures the trend of previous training history whereas the uniform segmentation decimates it. The recognition performance of the proposed adaptation procedures along with the suggested initialization is verified to be always better than that of existing training procedure using fixed number of mixture Gaussian p.d.f.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.18 no.6
• /
• pp.259-268
• /
• 2018

The purpose of this study is to propose a training method customized to the level of cognitive abilities to increase users' interest and engagement while using cognitive function training contents. The level of cognitive ability of the users was based on the clustering based on the users' information and Mini-Mental Statue Examination-Korea Child test score using the K-means algorithm applied collaborative filtering. The results were applied to the integrated cognitive function training system, and the contents order and difficulty level of the cognitive function training area were recommended to the user's cognitive ability level. Particularly, the contents difficulty control was designed to give a high immersion feeling by applying the 'flow theory' method that users can repeatedly feel tension and comfort. In conclusion, the user-customized cognitive function training method proposed in this paper can be expected to be more effective and rehabilitative results than existing therapists' subjective setting of contents order and difficulty level.

• Journal of the Korea Society of Computer and Information
• /
• v.14 no.3
• /
• pp.139-147
• /
• 2009

The memory based reasoning just stores in the memory in the form of the training pattern of the representative pattern. And it classifies through the distance calculation with the test pattern. Because it uses the techniques which stores the training pattern whole in the memory or in which it replaces training patterns with the representative pattern. Due to this, the memory in which it is a lot for the other machine learning techniques is required. And as the moreover stored training pattern increases, the time required for a classification is very much required. In this paper, We propose the EAS(Evaluation And Selection) algorithm in order to minimize memory usage and to improve classification performance. After partitioning the training space, this evaluates each partitioned space as MDL and PM method. The partitioned space in which the evaluation result is most excellent makes into the representative pattern. Remainder partitioned spaces again partitions and repeat the evaluation. We verify the performance of Proposed algorithm using benchmark data sets from UCI Machine Learning Repository.