Ming Ha Lee

Ming Ha Lee

Auto-generated profile, based on publicly available information

Ming Ha Lee

Faculty of Engineering, Computing and Science, Swinburne University of Technology Sarawak Campus, 93350 Kuching, Sarawak, Malaysia

Research Focus & Metrics

Academic Career Length

18Years (2009-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 December 2023 in IEEE Access
Verified
Journal Article
Open access
OJOlanrewaju Victor Johnson
Olanrewaju Victor Johnson
XCXinYing Chew
XinYing Chew
KKKHAI WAH KHAW
KHAI WAH KHAW
MLMing Ha Lee
Ming Ha Lee
There are continued efforts to build on the performance of deep learning (DL) models in various fields of application. Developing new DL models continues to open unprecedented opportunities in diverse application areas despite the enormous resources required. Generally, the learning mechanism of DL models depends on the term "cost function" (CF) or "loss function" (LF), and DL models require varied hyperparameter settings and, precisely, parameters that can help the model to continually minimize the cost function until faster convergence, with better generalization over the data in the loss landscape, is assumed. The learning rate (LR) update seeks to find the optimal solution for DL models through relative cost function minimization. Therefore, selecting the appropriate LR is essential to the performance of DL models. The existing cosine annealing LR, despite its demonstration for fast model convergence, lacks complete loss landscape exploration of the flat minima, hence limiting its ability to model better generalization. To address this, the paper proposes a period-shift Cosine Annealing Learning Rate with warm-up epochs (ps-CALR) to perturb the LR update. Six different publicly available datasets were used to benchmark the proposed LR method by experimenting with custom DL (multilayer perceptron and convolutional neural networks) and pre-trained DL models. The proposed ps-CARL enhances model generalization and convergence, pushing the solution to notably better performance compared with fixed LR and the existing cosine annealing method.

8 April 2016 in Communications in Statistics - Theory and Methods
Verified
Journal Article
Closed access
MLMing Ha Lee
Ming Ha Lee
MKMichael B. C. Khoo
Michael B. C. Khoo
The multivariate synthetic generalized sample variance |S| (synthetic |S|) chart is a combination of the |S| sub-chart and the conforming run length sub-chart. A procedure for optimal designs of the synthetic |S| chart, based on the median run length (MRL), for both zero and steady-state modes are provided by minimizing the out-of-control MRL. The comparative results show that the synthetic |S| chart performs better than the standard |S| chart for detecting shifts in the covariance matrix of a multivariate normally distributed process, in terms of the MRL. An example is given to illustrate the operation of the synthetic |S| chart.

3 April 2020 in IEEE Access
Verified
Journal Article
Open access
MLMing Ha Lee
Ming Ha Lee
MKMichael B. C. Khoo
Michael B. C. Khoo
XCXinYing Chew
XinYing Chew
PTPatrick Hang Hui Then
Patrick Hang Hui Then
This paper investigates the performance of the coefficient of variation chart in the presence of measurement errors for finite production horizon. We study a two-sided Shewhart coefficient of variation chart with measurement errors for detecting both increase and decrease in the coefficient of variation for short run processes using an error model with linear covariate. The performance of the coefficient of variation chart is evaluated by the truncated average run length and the expected value of the truncated average run length. The numerical results indicate that the precision error and the accuracy error have negative effect of the measurement errors on the performance of the coefficient of variation chart. In addition, the constant coefficient B in the linear covariate error model reduces the negative effect of the measurement errors on the performance of the coefficient of variation chart. However, taking multiple measurements per item in each sample is not an effective method to enhance the performance of the coefficient of variation chart. An example is provided to illustrate the implementation of the coefficient of variation chart. In addition, the economic criterion is also added to study the effect of measurement errors on the expected inspection cost. The result shows that an increase in the precision error reduces the expected inspection cost.

4 September 2017 in Communications in Statistics - Simulation and Computation
Verified
Journal Article
Closed access
MLMing Ha Lee
Ming Ha Lee
MKMichael B. C. Khoo
Michael B. C. Khoo
The economic-statistical design of charts with 2-of-2 and 2-of-3 runs rules for correlated data within a sample is developed in this study, in which the Markov chain approach is adopted to derive the statistical properties of the charts with runs rules. Lorenzen–Vance cost model is modified to determine the optimal design parameters and the generalized reduced gradient algorithm is used to solve the model. The results of the sensitivity analysis show that changing the value of correlation coefficient from low to high level increases the values of the control limit coefficient and the out-of-control average time to signal.

16 April 2019 in Communications in Statistics - Simulation and Computation
Verified
Journal Article
Closed access
MLMing Ha Lee
Ming Ha Lee
MKMichael B. C. Khoo
Michael B. C. Khoo
This study presents a double sampling (DS) np chart when the process parameter is estimated. The DS np chart is evaluated in terms of the average of the average run length and the standard deviation of the average run length. The results show that a large number of Phase-I samples is required for designing the DS np chart. Since it is unrealistic in practice to obtain a large amount of Phase-I data, a combination of new design parameters for the DS np chart with estimated process parameter is computed by taking the number of Phase-I samples into consideration.

16 May 2017 in Communications in Statistics - Theory and Methods
Verified
Journal Article
Closed access
MLMing Ha Lee
Ming Ha Lee
MKMichael B. C. Khoo
Michael B. C. Khoo
The synthetic np chart is a combination of the np sub-chart and the conforming run length sub-chart. A procedure for the optimal design of the synthetic np chart is provided by minimizing the out-of-control median run length (MRL). The comparative results show that the synthetic np chart performs better than the corresponding standard np chart for detecting process shifts in the fraction non conforming, in terms of the MRL. An example is given to illustrate the operation of the synthetic np chart.