• Communications for Statistical Applications and Methods
• /
• v.20 no.6
• /
• pp.481-490
• /
• 2013

Classification is an important research area as data can be easily obtained even if the number of predictors becomes huge. The support vector machine(SVM) is widely used to classify a subject into a predetermined group because it gives sound theoretical background and better performance than other methods in many applications. The SVM can be viewed as a penalized method with the hinge loss function and penalty functions. Instead of $L_2$ penalty function Fan and Li (2001) proposed the smoothly clipped absolute deviation(SCAD) satisfying good statistical properties. Despite the ability of SVMs, they have drawbacks of non-robustness when there are outliers in the data. We develop a robust SVM method using a weight function with the SCAD penalty function based on the local quadratic approximation. We compare the performance of the proposed SVM with the SVM using the $L_1$ and $L_2$ penalty functions.

• Journal of Information Processing Systems
• /
• v.16 no.6
• /
• pp.1407-1423
• /
• 2020

A graph is a data structure consisting of nodes and edges between these nodes. Graph embedding is to generate a low dimensional vector for a given graph that best represents the characteristics of the graph. Recently, there have been studies on graph embedding, especially using deep learning techniques. However, until now, most deep learning-based graph embedding techniques have focused on unweighted graphs. Therefore, in this paper, we propose a graph embedding technique for weighted graphs based on long short-term memory (LSTM) autoencoders. Given weighted graphs, we traverse each graph to extract node-weight sequences from the graph. Each node-weight sequence represents a path in the graph consisting of nodes and the weights between these nodes. We then train an LSTM autoencoder on the extracted node-weight sequences and encode each nodeweight sequence into a fixed-length vector using the trained LSTM autoencoder. Finally, for each graph, we collect the encoding vectors obtained from the graph and combine them to generate the final embedding vector for the graph. These embedding vectors can be used to classify weighted graphs or to search for similar weighted graphs. The experiments on synthetic and real datasets show that the proposed method is effective in measuring the similarity between weighted graphs.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.4
• /
• pp.2038-2057
• /
• 2017

Gaussian learning is a new technology in the computational intelligence area. However, this technology weakens the learning ability of a particle swarm and achieves a lack of diversity. Thus, this paper proposes a vector Gaussian learning strategy and presents an effective approach, named particle swarm optimization based on vector Gaussian learning. The experiments show that the algorithm is more close to the optimal solution and the better search efficiency after we use vector Gaussian learning strategy. The strategy adopts vector Gaussian learning to generate the Gaussian solution of a swarm's optimal location, increases the learning ability of the swarm's optimal location, and maintains the diversity of the swarm. The method divides the states into normal and premature states by analyzing the state threshold of the swarm. If the swarm is in the premature category, the algorithm adopts an inertia weight strategy that decreases linearly in addition to vector Gaussian learning; otherwise, it uses a fixed inertia weight strategy. Experiments are conducted on eight well-known benchmark functions to verify the performance of the new approach. The results demonstrate promising performance of the new method in terms of convergence velocity and precision, with an improved ability to escape from a local optimum.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.9
• /
• pp.133-141
• /
• 1994

Conventional competitive learning algorithms compute the Euclidien distance to determine the winner neuron out of all predetermined output neurons. In such cases, there is a drawback that the performence of the learning algorithm depends on the initial reference(=weight) vectors. In this paper, we propose a new competitive learning algorithm that dynamically generates output neurons. The proposed method generates output neurons by dynamically changing the class thresholds for all output neurons. We compute the similarity between the input vector and the reference vector of each output neuron generated. If the two are similar, the reference vector is adjusted to make it still more like the input vector. Otherwise, the input vector is designated as the reference vector of a new outputneuron. Since the reference vectors of output neurons are dynamically assigned according to input pattern distribution, the proposed method gets around the phenomenon that learning is early determined due to redundant output neurons. Experiments using speech data have shown the proposed method to be superior to existint methods.

• ETRI Journal
• /
• v.23 no.2
• /
• pp.61-70
• /
• 2001

In this paper, we propose an efficient method for personal identification by analyzing iris patterns that have a high level of stability and distinctiveness. To improve the efficiency and accuracy of the proposed system, we present a new approach to making a feature vector compact and efficient by using wavelet transform, and two straightforward but efficient mechanisms for a competitive learning method such as a weight vector initialization and the winner selection. With all of these novel mechanisms, the experimental results showed that the proposed system could be used for personal identification in an efficient and effective manner.

• ETRI Journal
• /
• v.39 no.5
• /
• pp.621-631
• /
• 2017

Abnormal samples are usually difficult to obtain in production systems, resulting in imbalanced training sample sets. Namely, the number of positive samples is far less than the number of negative samples. Traditional Support Vector Machine (SVM)-based anomaly detection algorithms perform poorly for highly imbalanced datasets: the learned classification hyperplane skews toward the positive samples, resulting in a high false-negative rate. This article proposes a new imbalanced SVM (termed ImSVM)-based anomaly detection algorithm, which assigns a different weight for each positive support vector in the decision function. ImSVM adjusts the learned classification hyperplane to make the decision function achieve a maximum GMean measure value on the dataset. The above problem is converted into an unconstrained optimization problem to search the optimal weight vector. Experiments are carried out on both Cloud datasets and Knowledge Discovery and Data Mining datasets to evaluate ImSVM. Highly imbalanced training sample sets are constructed. The experimental results show that ImSVM outperforms over-sampling techniques and several existing imbalanced SVM-based techniques.

• Proceedings of the IEEK Conference
• /
• 1999.06a
• /
• pp.500-503
• /
• 1999

In this paper, we analyze weight distributions of neural networks. If we construct a vector containing all weights of a neural network, then training process can be viewed as finding a solution point in the weight space. In order to obtain insight into the training process of neural networks, we investigate the distribution of the solution points in the weight space Experiments provide some interesting results, showing that solution points tend to form clusters in the weight space and the information may be used to speed up the training process.

• Journal of the Korea Society of Computer and Information
• /
• v.21 no.1
• /
• pp.107-113
• /
• 2016

In this paper, by managing the biometric data is changed with the passage of time, a systematic and scientifically propose a framework to increase the bio-vector generation efficiency of the smart health care. Increasing the development of human life as a medicine and has emerged smart health care according to this. Organic and efficient health management becomes possible to generate a vector when the biological domain to the wireless communication infrastructure based on the measurement of the health status and to take action in accordance with the change of the physical condition. In this paper, we propose a framework to create a bio-vector that contains information about the current state of health of the person. In the proposed framework, Bio vectors may be generated by collecting the biometric data such as blood pressure, pulse, body weight. Biometric data is the raw data from the bio-vector. The scope of the primary data can be set to active. As the collecting biometric data from multiple items of the bio-recognition vectors may increase. The resulting bio-vector is used as a measure to determine the current health of the person. Bio-vector generating the proposed framework, it can aid in the efficiency and systemic health of healthcare for the individual.

• Journal of Korea Society of Industrial Information Systems
• /
• v.24 no.1
• /
• pp.1-12
• /
• 2019

This paper proposes a new motion vector smoothing and motion compensation method to increase the frame rate of videos. The proposed method reduces false motion vector smoothing by improving the weight with reflecting accuracy to overcome the limitation of the conventional motion vector smoothing using the adaptively weighted vector median. Also, to improve the interpolated image quality of the conventional OBMC(Overlapped Block Motion Compensation), a scheme that divides an original block into 4 sub-blocks and then generates the interpolated frame using the reestimated motion vector for each sub-block is proposed. The simulation results prove that the proposed method can provide much better objective and subjective image quality than the conventional method.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.1
• /
• pp.62-68
• /
• 2002

It is well known that the MDVR beamforming outperforms the conventional delay-sum beamformer in the sense of noise rejection and bearing resolution. However, the MDVR method requires long observation time to achieve high frequency resolution. The STMV method uses the steered covariance matrix of sensor data, so it has an ability to form an adaptive weight vector from a single time-series snapshot. But it uses the same weight vector across all frequencies. In this paper, we propose an SSMV method. The basic idea of the SSMV method is to decompose a full frequency band into several subbands to acquire a weight vector for each subband, individually. Also the wrap may be divided into several subarrays in order to reduce a computational load and the bandwidth of each subband. Simulations using real sea trial data show that the proposed SSMV method has good performance with short observation time.