Geok See Ng

Followers

Following

Public Views

Interests

Uploads

Papers by Geok See Ng

Post nonlinear blind source separation by geometric linearization

Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005.

Download

A comprehensive approach, and a case study, for conducting attack detection experiments in Cyber–Physical Systems

Robotics and Autonomous Systems, Dec 1, 2017

h i g h l i g h t s • An approach for conducting robot cyber-security experiments is proposed. • ... more h i g h l i g h t s • An approach for conducting robot cyber-security experiments is proposed. • It includes vulnerability analysis and attack detection methods. • In addition, it provides guidelines for describing cyber-security experiments. • A study investigating the detection of robot sensor attacks was conducted. • Results highlight the usefulness of proposed approach for other empirical studies.

Download

ロボットのサイバー攻撃を検出するための方法の実験的評価【Powered by NICT】

IEEE Conference Proceedings, 2016

Undergraduate Students in a Computer Engineering Course: A Perspective of their Learning Approaches and Motivation Factors

Innovations in Education and Training International, Feb 1, 1997

SUMMARY This paper reports a study of the motivation and learning behaviour of first year compute... more SUMMARY This paper reports a study of the motivation and learning behaviour of first year computer engineering students at the School of Applied Science, Nanyang Technological University, Singapore. The Study Process Questionnaire (SPQ) (Biggs, 1987) is used to conduct the survey. The results of the survey are analyzed and recommendations are made for prospective first year computer engineering students.

Download

Complementary Learning Fuzzy Neural Network: An Approach to Imbalanced Dataset

IEEE ... International Conference on Neural Networks, Aug 1, 2007

Download

RLDDE: A novel reinforcement learning-based dimension and delay estimator for neural networks in time series prediction

Neurocomputing, Mar 1, 2007

Time series prediction is traditionally handled by linear models such as autoregressive and movin... more Time series prediction is traditionally handled by linear models such as autoregressive and moving-average. However they are unable to adequately deal with the non-linearity in the data. Neural networks are non-linear models that are suitable to handle the non-linearity in time series. When designing a neural network for prediction, two critical factors that affect the performance of the neural network predictor should be considered; they are namely: (1) the input dimension, and (2) the time delay. The former is the number of delayed values for prediction, while the latter is the time interval between two data. Prediction accuracy can be improved using suitable input dimension and time delay. A novel method, called reinforcement learning-based dimension and delay estimator (RLDDE), is proposed in this paper to simultaneously determine the input dimension and time delay. RLDDE is a meta-learner that tries to learn the selection policy of the dimension and delay under different distribution of the data. Two benchmarked datasets with different noise levels and one stock price are used to show the effectiveness of the proposed RLDDE together with the benchmarking against other methods.

Download

Comparison of Corrupted Sensor Data Detection Methods in Detecting Stealthy Attacks on Cyber-Physical Systems

2017 IEEE 22nd Pacific Rim International Symposium on Dependable Computing (PRDC), 2017

Effectiveness of seven methods for detecting stealthy attacks on Cyber Physical Systems (CPS) was... more Effectiveness of seven methods for detecting stealthy attacks on Cyber Physical Systems (CPS) was investigated using an experimental study. The Amigobot robot was used as the CPS. The experiments were conducted in simulation as well as on the physical robot. Three types of stealthy attacks were implemented: surge, bias, and geometric. Two variations of Cumulative Sum (CUSUM) method for detecting attacks were evaluated: partial and full physics. Four attack scenarios were implemented. Results from the experiments indicate that stealthy attacks could remain undetected by the CUSUM methods for some attack scenarios. In addition to the CUSUM-based methods, a set of five methods to complement CUSUM were implemented and their effectiveness assessed. While the additional methods do improve the effectiveness of CUSUM-based methods, some attacks remained undetected regardless of which method, or a combination of methods, was used for detection due to the amount of variation in sensor measurements between different runs in simulation and in the physical robot.

Leveraging the trade-off between accuracy and interpretability in a hybrid intelligent system

Neural Fuzzy Inference System (NFIS) is a widely adopted paradigm to develop a data-driven learni... more Neural Fuzzy Inference System (NFIS) is a widely adopted paradigm to develop a data-driven learning system. This hybrid system has been widely adopted due to its accurate reasoning procedure and comprehensible inference rules. Although most NFISs primarily focus on accuracy, we have observed an ever increasing demand on improving the interpretability of NFISs and other types of machine learning systems. In this paper, we illustrate how we leverage the trade-off between accuracy and interpretability in an NFIS called Genetic Algorithm and Rough Set Incorporated Neural Fuzzy Inference System (GARSINFIS). In a nutshell, GARSINFIS self-organizes its network structure with a small set of control parameters and constraints. Moreover, its autonomously generated inference rule base tries to achieve higher interpretability without sacrificing accuracy. Furthermore, we demonstrate different configuration options of GARSINFIS using well-known benchmarking datasets. The performance of GARSINFIS...

Empirical Assessment of Corrupt Sensor Data Detection Methods in a Robot

2016 IEEE 40th Annual Computer Software and Applications Conference (COMPSAC), 2016

An experiment was conducted to investigate the response of a robot to cyber attacks and the effec... more An experiment was conducted to investigate the response of a robot to cyber attacks and the effectiveness of methods to detect such attacks. The experiment was run in simulation as well as on an actual robot. To ensure validity of results, cyber attacks were implemented on three robots of the same make and model through their wireless control mechanisms. Attacks were launched to investigate their feasibility, impact, and the effectiveness of the detection methods. Analysis of experimental data indicates that, among the several methods examined, the one which compares sensor values to the average historical values, is the most effective. In some experiments, the effectiveness of various methods was found to be lower in actual robots as compared to that in simulation. Thus, when practically feasible, it is important to test security countermeasures in realistic environments. Furthermore, factors such as attack size and timing, were found to influence the attack detection effectiveness, and hence ought to be considered while designing security countermeasures.

Experimental Evaluation of Stealthy Attack Detection in a Robot

2015 IEEE 21st Pacific Rim International Symposium on Dependable Computing (PRDC), 2015

An experiment was conducted to investigate the effectiveness of the Cumulative Sum (CUSUM) approa... more An experiment was conducted to investigate the effectiveness of the Cumulative Sum (CUSUM) approach for detecting cyber attacks on Cyber Physical Systems (CPS). The Amigobot robot was used as the CPS in this study. Three types of stealthy attacks were considered, namely, surge, bias, and geometric. While a similar study has been reported earlier using a simulated chemical plant, the objective of the study reported here was to replicate the previous study in a realistic CPS environment and investigate whether the detection method performs differently. Cyber attacks were implemented on the Amigobot through its wireless control mechanism by changing the readings obtained from one of its sonar sensors. In addition, the investigation focused on understanding the impact of attack timing and duration on (a) detection effectiveness of the CUSUM method and (b) system safety. Analysis of experimental data indicates differences between results reported in the previous simulation-based study and those reported here.

An experimental study of the extended NRBF regression model and its enhancement for classification problem

Neurocomputing, 2008

As an extension of the traditional normalized radial basis function (NRBF) model, the extended no... more As an extension of the traditional normalized radial basis function (NRBF) model, the extended normalized RBF (ENRBF) model was proposed by Xu [RBF nets, mixture experts, and Bayesian Ying-Yang learning, Neurocomputing 19 (1998) 223–257]. In this paper, we perform a supplementary study on ENRBF with several properly designed experiments and some further theoretical discussions. It is shown that ENRBF is

Functional Prediction of Snake Neurotoxins

2006 9th International Conference on Control, Automation, Robotics and Vision, 2006

... Seng Hong SEAH1, Chee Keong KWOH1, Vladimir BRUSIC2, Meena Kishore SAKHARKAR3, Geok See NG1 1... more

Intelligent Fax Management System

Journal of Information & Knowledge Management, 2003

Fax machines have become a ubiquitous equipment in today's office environment due to its abil... more Fax machines have become a ubiquitous equipment in today's office environment due to its ability to send information around with speed and convenience. Increasingly many personal computers nowadays are able to play the role of a fax machine by adding in a fax card and the relevant software. This paper proposes a Fax Management System employing character recognition technique to enhance the basic capability of a PC-based fax system. The proposed enhancement includes Personal Fax Handling, Automatic Information Retrieval and Order Processing. This enhancement is envisaged to be of significant commercial potential as it addresses certain unfulfilled needs in current PC-based fax systems.

Download

Nonlinear Active Handwriting Models and Their Applications to Handwritten Chinese Radical Recognition

International Conference on Document Analysis and Recognition, 2003

This paper proposes active handwriting models, in which kernel principal component analysis is ap... more This paper proposes active handwriting models, in which kernel principal component analysis is applied to capture nonlinear handwriting variations. In the recogni- tion phase, the chamfer distance transform and a dynamic tunnelling algorithm (DTA) are employed to search for the optimal shape parameters. The proposed methodology is successfully applied to a novel radical decomposition ap- proach to the challenging problem

Download

RS-HeRR: a rough set-based Hebbian rule reduction neuro-fuzzy system

Neural Computing and Applications, 2020

Interpretabilty is one of the desired characteristics in various classification task. Rule-based ... more Interpretabilty is one of the desired characteristics in various classification task. Rule-based system and fuzzy logic can be used for interpretation in classification. The main drawback of rule-based system is that it may contain large complex rules for classification and sometimes it becomes very difficult in interpretation. Rule reduction is also difficult for various reasons. Removing important rules may effect in classification accuracy. This paper proposes a hybrid fuzzy-rough set approach named RS-HeRR for the generation of effective, interpretable and compact rule set. It combines a powerful rule generation and reduction fuzzy system, called Hebbian-based rule reduction algorithm (HeRR) and a novel rough-set-based attribute selection algorithm for rule reduction. The proposed hybridization leverages upon rule reduction through reduction in partial dependency as well as improvement in system performance to significantly reduce the problem of redundancy in HeRR, even while pr...

Download

Modified discrete clustering technique: a novel approach to represent membership functions of fuzzy sets

Control, Automation, Robotics and Vision …

In fuzzy neural network systems, fuzzy membership functions play a key role in making the fuzzy s... more In fuzzy neural network systems, fuzzy membership functions play a key role in making the fuzzy sets organize the input data knowledge in an appropriate and representative manner. Earlier clustering techniques exploit some uniform, convex algebraic functions, such as Gaussian, Triangular or Trapezoidal to represent the fuzzy sets. However, due to the irregularity of the input data, regular and uniform fuzzy sets may not be able to represent the exact feature information of input data. In order to address this issue, a clustering method called Modified Discrete Clustering Technique (MDCT) is proposed in this paper. MDCT represents non-uniform, and normal fuzzy sets with a set of irregular sampling points. The sampling points learn the knowledge of data feature in an irregular and flexible manner. Thus, the fuzzy membership functions generated using these sampling points can provide a better representation of the actual input data.

Download

An interpretable neural fuzzy inference system for predictions of underpricing in initial public offerings

Neurocomputing

Entropy Learning in Neural Network

ASEAN Journal on Science and Technology for Development

In this paper, entropy term is used in the learning phase of a neural network. As learning progr... more In this paper, entropy term is used in the learning phase of a neural network. As learning progresses, more hidden nodes get into saturation. The early creation of such hidden nodes may impair generalisation. Hence entropy approach is proposed to dampen the early creation of such nodes. The entropy learning also helps to increase the importance of relevant nodes while dampening the less important nodes. At the end of learning, the less important nodes can then be eliminated to reduce the memory requirements of the neural network.

Download

Clinical Diagnosis Using Proteomics and Complementary Learning

The 2006 IEEE International Joint Conference on Neural Network Proceedings, 2000

Accurate prediction of the Translation Initiation Sites (TIS) in eukaryotes is paramount for bett... more Accurate prediction of the Translation Initiation Sites (TIS) in eukaryotes is paramount for better understanding of the translation process, gene structure, as well as protein coding, and for more reliable amino acid prediction, etc. However, detecting TIS is not a simple task. Hence, computational biology is adopted to assist in the detection. Unfortunately, some computational biology tools do not provide means for facilitating the knowledge extraction or system validation. Also, they have neither biological interpretation nor human-like reasoning process. Realizing that, a novel Genetic Complementary Learning (GCL) fuzzy neural network, which based on gene selection process, is proposed. GCL inherits some advantageous traits from three worlds: the dynamics from genetic algorithm, the good pattern recognition performance from complementary learning, as well as the interpretable, autonomous, and human-like operations from fuzzy neural network. From experimental result, GCL demonstrates itself as a competent tool for TIS prediction.

Download

A Neuropsychology-inspired Learning System for Human Uncertainty Monitoring

2006 9th International Conference on Control Automation Robotics and Vision, 2006

Uncertainty exists in various complex problems. Yet, human is able to effectively handle these un... more Uncertainty exists in various complex problems. Yet, human is able to effectively handle these uncertainties and makes appropriate decision. Thus, modeling of human uncertainty process should improve the performance of learning system in uncertain environment. A mechanism for human uncertainty monitoring is the broad and narrow generalization in category learning. This can be modeled using upper and lower membership functions, which corresponds to the broad and narrow generalizations respectively. These upper and lower membership functions can be implemented using the Fuzzy Rough Set (FR) theory. A Complementary Learning Fuzzy Neural Network (CLFNN) is a functional model of human pattern recognition. It is integrated with the human uncertainty monitoring model and the resultant FRCLFNN offers good classification performance and better representation power as it captures input, linguistic, and rough uncertainties. Experimental result supports that FRCLFNN is a competent decision support system.

Download

Geok See Ng

Uploads

Papers by Geok See Ng

Log In