• 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.

• Communications for Statistical Applications and Methods
• /
• v.24 no.6
• /
• pp.651-662
• /
• 2017

We propose an alternative procedure to select penalty parameter in $L_1$ penalized robust regression. This procedure is based on marginalization of prior distribution over the penalty parameter. Thus, resulting objective function does not include the penalty parameter due to marginalizing it out. In addition, its estimating algorithm automatically chooses a penalty parameter using the previous estimate of regression coefficients. The proposed approach bypasses cross validation as well as saves computing time. Variable-wise penalization also performs best in prediction and variable selection perspectives. Numerical studies using simulation data demonstrate the performance of our proposals. The proposed methods are applied to Boston housing data. Through simulation study and real data application we demonstrate that our proposals are competitive to or much better than cross-validation in prediction, variable selection, and computing time perspectives.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.50 no.1
• /
• pp.1-5
• /
• 2001

In this paper, a new method for calculating the penalty factors of all generators including the slack bus is presented. A simple transposition of the angle reference, from the conventional slack bus to another bus where no generation exists, enables the derivation of the loss sensitivity of the slack bus. Penalty factors are obtained without any physical assumption through a simple substitution of the bus loss sensitivities. Penalty factors calculated by proposed method are not dependent on reference bus and can also be directly substituted into the general ELD equation for computing the optimal dispatch. Equations for loss sensitivities, Penalty factors and ELD are calculated simultaneously in normal power flow computation. A case study on a test system has proved the effectiveness of the proposed' angle reference transposition' method.

• Journal of IKEEE
• /
• v.18 no.3
• /
• pp.392-397
• /
• 2014

In this paper, we propose a memory operation system architecture for memory latency penalty reduction in SIMT architecture based stream processor. The proposed architecture applied non-blocking cache architecture to reduce cache miss penalty generated by blocking cache architecture. We verified that the proposed memory operation architecture improve the performance of the stream processor by comparing processing performances of various algorithms. We measured the performance improvement rate that was improved in accordance with the ratio of memory instruction in each algorithm. As a result, we confirmed that the performance of stream processor improves up to minimum 8.2% and maximum 46.5%.

• Journal of Biomedical Engineering Research
• /
• v.38 no.5
• /
• pp.219-226
• /
• 2017

Penalized-likelihood (PL) reconstruction methods for transmission tomography are known to provide improved image quality for reduced dose level by efficiently smoothing out noise while preserving edges. Unfortunately, however, most of the edge-preserving penalty functions used in conventional PL methods contain at least one free parameter which controls the shape of a non-quadratic penalty function to adjust the sensitivity of edge preservation. In this work, to avoid difficulties in finding a proper value of the free parameter involved in a non-quadratic penalty function, we propose a new adaptive method of space-variant smoothing with a simple quadratic penalty function. In this method, the smoothing parameter is adaptively selected for each pixel location at each iteration by using the image roughness measured by a pixel-wise standard deviation image calculated from the previous iteration. The experimental results demonstrate that our new method not only preserves edges, but also suppresses noise well in monotonic regions without requiring additional processes to select free parameters that may otherwise be included in a non-quadratic penalty function.

• Communications for Statistical Applications and Methods
• /
• v.15 no.4
• /
• pp.633-642
• /
• 2008

Variable selection algorithm for partial least square regression using penalty function is proposed. We use the fact that usual partial least square regression problem can be expressed as a maximization problem with appropriate constraints and we will add penalty function to this maximization problem. Then simulated annealing algorithm can be used in searching for optimal solutions of above maximization problem with penalty functions added. The HARD penalty function would be suggested as the best in several aspects. Illustrations with real and simulated examples are provided.

• Journal of Korean Society of Transportation
• /
• v.27 no.1
• /
• pp.157-167
• /
• 2009

This study deals with the turn penalty functions in the urban transportation demand forecasting. The objectives are to develop the penalty functions of left-turn traffic in the case of signalized intersections, and to analyze the applicability of the functions to the traffic assignment models. This is based on the background that the existing models can not effectively account for the delays of left-turn traffic which is bigger than that of through traffic. In pursuing the above, this study gives particular attention to developing the penalty functions based on the degrees of saturation by simulation results of Transyt-7F, and analyzing the applicability of the functions by the case study of Cheongju. The major findings are the followings. First, two penalty functions developed according to the degrees of saturation, are evaluated to be all statistically significant. Second, the results that the above functions apply to the Cheongju network, are analyzed to be converging, though the iteration numbers increase. Third, the link volumes forecasted by turn penalty functions are evaluated to be better fitted to the observed data than those by the existing models. Finally, the differences of traffic volumes assigned by two functions, which are exponential and divided forms, are analyzed to be very small.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.9 no.3
• /
• pp.12-19
• /
• 2010

In road network, especially in urban area, inefficiency of travel time is caused by signal control and turn maneuver at intersection and this inefficiency has substantial effects on travel time. When searching for the shortest path, this inefficiency which is caused by turn maneuver must be considered. Therefore, travel time, vehicle volume and delay for each link were calculated by using simulation package, PARAMICS V5.2 for adaptation of turn penalty at 16 intersections of Gangnam-gu. Turn penalty was calculated respectively for each intersection. Within the same intersection, turn penalty differs by each approaching road and turn direction so the delay was calculated for each approaching road and turn direction. Shortest path dealing with 16 intersections searched by Dijkstra algorithm using travel time as cost, considering random turn penalty, and algorithm considering calculated turn penalty was compared and analyzed. The result shows that by considering turn penalty searching the shortest path can decrease the travel time can be decreased. Also, searching the shortest path which considers turn penalty can represent reality appropriately and the shortest path considering turn penalty can be utilized as an alternative.

• Korean Journal of Cognitive Science
• /
• v.28 no.1
• /
• pp.65-90
• /
• 2017

Rewards or penalties become informative only when contingent on an immediately preceding response. Our goal was to determine if the brain responds differently to motivational events depending on whether they provide feedback with the contingencies effective for learning. Event-related fMRI data were obtained from 22 volunteers performing a visuomotor categorical task. In learning-condition trials, participants learned by trial and error to make left or right responses to letter cues (16 consonants). Monetary rewards (+500) or penalties (-500) were given as feedback (learning feedback). In random-condition trials, cues (4 vowels) appeared right or left of the display center, and participants were instructed to respond with the appropriate hand. However, rewards or penalties (random feedback) were given randomly (50/50%) regardless of the correctness of response. Feedback-associated BOLD responses were analyzed with ANOVA [trial type (learning vs. random) x feedback type (reward vs. penalty)] using SPM8 (voxel-wise FWE p < .001). The right caudate nucleus and right cerebellum showed activation, whereas the left parahippocampus and other regions as the default mode network showed deactivation, both greater for learning trials than random trials. Activations associated with reward feedback did not differ between the two trial types for any brain region. For penalty, both learning-penalty and random-penalty enhanced activity in the left insular cortex, but not the right. The left insula, however, as well as the left dorsolateral prefrontal cortex and dorsomedial prefrontal cortex/dorsal anterior cingulate cortex, showed much greater responses for learning-penalty than for random-penalty. These findings suggest that learning-penalty plays a critical role in learning, unlike rewards or random-penalty, probably not only due to its evoking of aversive emotional responses, but also because of error-detection processing, either of which might lead to changes in planning or strategy.

• Geomechanics and Engineering
• /
• v.15 no.4
• /
• pp.947-955
• /
• 2018

In this study, we propose a three-dimensional simplified slope stability analysis using a hybrid-type penalty method (HPM). In this method, a solid element obtained by the HPM is applied to a column that divides the slope into a lattice. Therefore, it can obtain a safety factor in the same way as simplified methods on the slip surface. Furthermore, it can obtain results (displacement and strain) that cannot be obtained by conventional limit equilibrium methods such as the Hovland method. The continuity condition of displacement between adjacent columns and between elements for each depth is considered to incorporate a penalty function and the relative displacement. For a slip surface between the bottom surface and the boundary condition to express the slip of slope, we introduce a penalty function based on the Mohr-Coulomb failure criterion. To compute the state of the slip surface, an r-min method is used in the load incremental method. Using the result of the simple three-dimensional slope stability analysis, we obtain a safety factor that is the same as the conventional method. Furthermore, the movement of the slope was calculated quantitatively and qualitatively because the displacement and strain of each element are obtained.