• The KIPS Transactions:PartD
• /
• v.8D no.2
• /
• pp.175-180
• /
• 2001

This paper presents the optimal fraction of validation set to obtain a prediction accuracy of software failure count or failure time in the future by a neuro-fuzzy system. Given a fixed amount of training data, the most popular effective approach to avoiding underfitting and overfitting is early stopping, and hence getting optimal generalization. But there is unresolved practical issues : How many data do you assign to the training and validation set\ulcorner Rules of thumb abound, the solution is acquired by trial-and-error and we spend long time in this method. For the sake of optimal fraction of validation set, the variant specific fraction for the validation set be provided. It shows that minimal fraction of the validation data set is sufficient to achieve good next-step prediction. This result can be considered as a practical guideline in a prediction of software reliability by neuro-fuzzy system.

• Journal of the Korean Statistical Society
• /
• v.24 no.2
• /
• pp.293-301
• /
• 1995

Sums of independent random variables $S_n = X_1 + \cdots + X_n$ are considered, where the $X_n$ are chosen according to a stationary process of distributions. For $c > 0$, let $t_c$ be the smallest positive integer n such that $$\mid$S_n$\mid$ > cn^{\frac{1}{2}}$. In this set up we are concerned with finiteness of expectation of $t_c$ and we have some results of sign-invariant process as applications.

• Proceedings of the KSR Conference
• /
• 2009.05a
• /
• pp.270-277
• /
• 2009

This paper describes the method of how to improve the probability of the tracing stopping ratio which is the most important factor in Automatic Train Operation (ATO). Aspects of improving the performance of automatic driving, the followings are investigated and studied : the interface between the signal system and the vehicle, the need to improve braking system, test method of blending brakes, how to minimize the delay of commands for real-time control. In this study, we applied this method to prove the effectiveness in DAEJEON LINE1

• ETRI Journal
• /
• v.43 no.5
• /
• pp.775-786
• /
• 2021

To translate in real time, a simultaneous translation system should determine when to stop reading source tokens and generate target tokens corresponding to a partial source sentence read up to that point. However, conventional attention-based neural machine translation (NMT) models cannot produce translations with adequate latency in online scenarios because they wait until a source sentence is completed to compute alignment between the source and target tokens. To address this issue, we propose a reinforced learning (RL)-based attention mechanism, the reinforced attention mechanism, which allows a neural translation model to jointly train the stopping criterion and a partial translation model. The proposed attention mechanism comprises two modules, one to ensure translation quality and the other to address latency. Different from previous RL-based simultaneous translation systems, which learn the stopping criterion from a fixed NMT model, the modules can be trained jointly with a novel reward function. In our experiments, the proposed model has better translation quality and comparable latency compared to previous models.

• Bulletin of the Korean Mathematical Society
• /
• v.56 no.5
• /
• pp.1211-1217
• /
• 2019

Let $0<{\alpha}<{\infty}$ be fixed, and let $X=(X_t)_{t{\geq}0}$ be a Bessel process with dimension $0<{\theta}{\leq}1$ starting at $x{\geq}0$. In this paper, it is proved that there are positive constants A and D depending only on ${\theta}$ and ${\alpha}$ such that $$E_x\({\exp}[{\alpha}\;\max_{0{\leq}t{\leq}{\tau}}\;X_t]\){\leq}AE_x\({\exp}[D_{\tau}]\)$$ for any stopping time ${\tau}$ of X. This inequality is also shown to be sharp.

• Management Science and Financial Engineering
• /
• v.22 no.1
• /
• pp.13-20
• /
• 2016

We investigate the optimal consumption and investment problem when a working debtor has an option to file for bankruptcy. By applying the duality approach, the closed-form solutions are obtained for the case of CRRA utility function. The optimal bankruptcy time is determined by the first hitting time when the financial wealth hits the wealth threshold derived from the optimal stopping time problem. Moreover, the numerical results show that the investment increases as the wealth approaches the threshold and the value gain from the bankruptcy option is vanished as wealth increases.

• Proceedings of the Safety Management and Science Conference
• /
• 2008.11a
• /
• pp.35-39
• /
• 2008

This paper considers the sequencing of products in mixed model assembly lines. The sequence which minimizes overall utility work in car assembly lines reduce the cycle time, the number of utility workers, and the risk of conveyor stopping. The sequencing problem is solved using Tabu Search. Tabu Search is a heuristic method which can provide a near optimal solution in real time. Various examples are presented and experimental results are reported to demonstrate the efficiency of the technique.

• Journal of applied mathematics & informatics
• /
• v.11 no.1_2
• /
• pp.373-378
• /
• 2003

consider the problem of sequentially hypotheses about a parameter $\theta$ in the presence of the nuisance parameter $\rho$. and we investigate further to computing the error probabilities and expected sample sizes in the frequentist properties of the adaptive S.P.R.T. for $\theta$.

• Journal of the Korean Statistical Society
• /
• v.19 no.2
• /
• pp.160-170
• /
• 1990

A new approximation method is proposed for the ARL of CUSUM chart which is based on the log probability ratio statistic. This method uses the condition of before-stopping time to derive the expectation of excess over boundaries. The proposed method is compared to some other approximation methods in normal and exponential cases.

• Journal of Korean Society of Transportation
• /
• v.22 no.4 s.75
• /
• pp.147-158
• /
• 2004

A class of SWP(Stochastic Wane Propagation) models microscopically mimics individual vehicles' stochastic behavior and traffic jam propagation with simplified car-following models based on CA(Cellular Automata) theory and macroscopically captures dynamic traffic flow relationships based on statistical physics. SWP model, a program-oriented model using both discrete time-space and integer data structure, can simulate a huge road network with high-speed computing time. However, the model has shortcomings to both the capturing of low speed within a jam microscopically and that of the density and back propagation speed of traffic congestion macroscopically because of the generation of spontaneous jam through unrealistic collision avoidance. In this paper, two additional rules are integrated into the NaSch model. The one is SMR(Stopping Maneuver Rule) to mimic vehicles' stopping process more realistically in the tail of traffic jams. the other is LAR(Low Acceleration Rule) for the explanation of low speed characteristics within traffic jams. Therefore, the CA car-following model with the two rules prevents the lockup condition within a heavily traffic density capturing both the stopping maneuver behavior in the tail of traffic jam and the low acceleration behavior within jam microscopically, and generates more various macroscopic traffic flow mechanism than NaSch model's with the explanation of propagation speed and density of traffic jam.