School of Electronics and Information Technology, Sun Yat-sen University, Guangzhou, China, Schoo... more School of Electronics and Information Technology, Sun Yat-sen University, Guangzhou, China, School of Engineering, Computer and Mathematical Sciences, Auckland University of Technology, Auckland, New Zealand, School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu, China, College of Information Engineering (College of Artificial Intelligence), Yangzhou University, Yangzhou, China, School of Computer Science and Electronic Engineering, University of Essex, Colchester, United Kingdom, Department of Electrical and Computer Engineering, New Jersey Institute of Technology, Newark, NJ, United States, Department of Electronics and Communication Engineering, Institute of Information Technology (IIIT-Delhi), New Delhi, India, Department of Computer and Communication Systems Engineering, Putra Malaysia University, Selangor Darul Ehsan, Malaysia
Attention mechanisms have been incorporated into many neural network-based natural language proce... more Attention mechanisms have been incorporated into many neural network-based natural language processing (NLP) models. They enhance the ability of these models to learn and reason with long input texts. A critical part of such mechanisms is the computation of attention similarity scores between two elements of the texts using a similarity score function. Given that these models have different architectures, it is difficult to comparatively evaluate the effectiveness of different similarity score functions. In this paper, we proposed a baseline model that captures the common components of recurrent neural network-based Question Answering (QA) systems found in the literature. By isolating the attention function, this baseline model allows us to study the effects of different similarity score functions on the performance of such systems. Experimental results show that a trilinear function produced the best results among the commonly used functions. Based on these insights, a new T-trilin...
2021 IEEE Global Communications Conference (GLOBECOM), 2021
In order to perform competitive privacy-guaranteed object detection, we propose an end-to-end mod... more In order to perform competitive privacy-guaranteed object detection, we propose an end-to-end model called Privacy-preserving Deep Transformation Self-attention (PPDTSA). This model ensures the privacy of the inference results. It has a low-complexity hierarchical structure with a relatively small number of hyper-parameters. Consistency of prediction is achieved through the encoding and decoding blocks of the self-attention mechanism which enables points of interest to be located. Focus loss is estimated based on foreground-background imbalance. The remaining dense blocks enable image details to be retained and the Region Of Interest to be expanded. At the same time, the objects detected in the image are protected through the privacy noise volume which is specified by the user. Experimental results demonstrate that PPDTSA achieves superior performance on the MOT20 dataset compared with three other state-of-the-art object detection models.
Practical Digital Signal Processing for Engineers and Technicians Edmund Lai PhD, BEng; Lai and A... more Practical Digital Signal Processing for Engineers and Technicians Edmund Lai PhD, BEng; Lai and Associates, Singapore . ... Newnes An imprint of Elsevier Linacre House, Jordan Hill, Oxford OX2 8DP 200 Wheeler Road, Burlington, MA 01803 First published 2003 Copyright ...
The analytic process of parametrising manifolds is well known to Mathematicians and Physicists, [... more The analytic process of parametrising manifolds is well known to Mathematicians and Physicists, [1] and is becoming increasingly familiar to Engi-neers concerned with, for example, robot manip-ulator trajectories [2]. The numerical equivalent does not appear to ...
2017 2nd Workshop on Recent Trends in Telecommunications Research (RTTR), 2017
In recent years, research into localization systems has become more popular as the proliferation ... more In recent years, research into localization systems has become more popular as the proliferation of Wireless Sensor Networks (WSNs) grows. Wireless Localization can refer to either an “Active” system which tracks a mobile transceiver, or “Passive” localization which tracks a transceiver free entity by measuring the changes it makes to the surrounding wireless environment. Recent work has seen both of these systems implemented with Received Signal Strength Indication (RSSI) values from transceivers. Many algorithms and channel models have been presented to increase the accuracy of a Received Signal Strength (RSS) based system. In this paper we experimentally check whether RSSI values map to the expected RSS values within an IEEE 802.15.4 network. Indoor experiments are repeated within an ideal outdoor environment, with multiple device platforms, to eliminate indoor multipath propagation as the cause for inconsistent behavior of RSSI. We identify 3 key issues with raw RSSI values and present either a possible solution or a mitigation strategy to reduce their effect. We conclude that using raw RSSI values is flawed, as the premise that they uniquely map to the distance between transceivers is incorrect. However they may be calibrated to increase their accuracy, and therefore viability.
Device-free localization (DFL) algorithms using the received signal strength indicator (RSSI) met... more Device-free localization (DFL) algorithms using the received signal strength indicator (RSSI) metrics have become a popular research focus in recent years as they allow for location-based service using commercial-off-the-shelf (COTS) wireless equipment. However, most existing DFL approaches have limited applicability in realistic smart home environments as they typically require extensive offline calibration, large node densities, or use technology that is not readily available in commercial smart homes. In this paper, we introduce SpringLoc and a DFL algorithm that relies on simple parameter tuning and does not require offline measurements. It localizes and tracks an entity using an adaptive spring relaxation approach. The anchor points of the artificial springs are placed in regions containing the links that are affected by the entity. The affected links are determined by comparing the kernel-based histogram distance of successive RSSI values. SpringLoc is benchmarked against existing algorithms in two diverse and realistic environments, showing significant improvement over the state-of-the-art, especially in situations with low-node deployment density. INDEX TERMS Device-free localization (DFL), histogram distance, indoor positioning systems (IPS), smart homes, spring-relaxation.
Device-free localization (DFL) systems that rely on the wireless received signal strength indicat... more Device-free localization (DFL) systems that rely on the wireless received signal strength indicator (RSSI) metric to localize targets with no device attached to them have been reported in the literature for almost a decade. Approaches using RSSI can be split into three main categories. Link-based approaches utilize weighted summation or probabilistic methods to infer location. Location-based approaches create a fingerprint map of an area. Radio Tomographic Imaging treat DFL as an imaging problem solved with a linear inverse. In this article, we implement and investigate the performance of all three major RSSI approaches in two test environments. We demonstrate how different environments and walking trajectories can have significant effects on the localization accuracy. The experimental results lead us to the conclusion that without implementing and testing within the same environment for the same target trajectories, the performance of various classes of DFL systems cannot be reliably evaluated. Relying on the stated accuracy from the literature for comparison is a flawed premise.
2012 IEEE International Symposium on Circuits and Systems, 2012
This paper presents the design of a variable linear phase finite impulse response filter based on... more This paper presents the design of a variable linear phase finite impulse response filter based on second order frequency transformations and coefficient decimation. The design of variable digital filters (VDFs) using first and second order frequency transformations have been proposed in literature. The VDF using second order transformation has better cut-off slope characteristics compared to the VDF using first order transformation. However, the former has the drawback of limited range (approximately 25% of the half of the sampling frequency) over which the cut-off frequency, fc, can be varied. It also fails to provide variable lowpass, highpass, bandpass or bandstop responses from a fixed-coefficient lowpass filter using the same architecture. The architecture proposed here overcomes the above mentioned disadvantages using coefficient decimation technique. The design example shows that the range over which fc can be varied is 2.65 times wider in the proposed VDF than the VDF in [7] and for a given frequency range, the proposed VDF offers a total gate count saving of 33% and 41% over the VDF in [11] and [7] respectively. Also, the proposed architecture provides variable lowpass, highpass, bandpass or bandstop responses from a fixed coefficient lowpass filter.
IEEE Transactions on Circuits and Systems II: Express Briefs, 2014
This brief presents a low complexity linear phase variable digital filter (VDF) design with tunab... more This brief presents a low complexity linear phase variable digital filter (VDF) design with tunable lowpass (LP), highpass (HP), bandpass (BP) and bandstop (BS) responses anywhere over entire Nyquist band. The spectral parameter approximation based VDFs (SPA-VDFs) designed using Farrow structure have advantages of linear phase, lower group delay and fewer variable multipliers. However, the total gate count and dynamic range of filter coefficient values of SPA-VDFs increase significantly with the tunable range of cutoff frequency which limits their usefulness in emerging signal processing and wireless communication applications. Also, existing VDFs need to update either filter coefficients or need parallel filter structures to obtain variable LP, HP, BP and BS responses. In this paper, a new VDF design is proposed by deftly integrating SPA-VDF with the modified coefficient decimation method (MCDM) and it shall be referred to as SPA-MCDM-VDF. The SPA-MCDM-VDF provides LP, HP, BP and BS responses with unabridged center frequency and bandwidth control over the entire Nyquist band without the need for hardware re-implementation or coefficient update. The complexity comparisons show that the SPA-MCDM-VDF offers substantial savings in gate count, group delay and number of variable multiplications over other linear phase VDFs.
Page 1. On Minimum Sampling Rates for Signals in Shift Invariant Spaces Beilei Huang, Edmund MK. ... more Page 1. On Minimum Sampling Rates for Signals in Shift Invariant Spaces Beilei Huang, Edmund MK. Lai Institute of Information Sciences & Technology Massey University, Wellington, New Zealand AP Vinod School of Computer ...
2017 27th International Telecommunication Networks and Applications Conference (ITNAC), 2017
One of the main challenges of operating a smart city (SC) is collecting the massive data generate... more One of the main challenges of operating a smart city (SC) is collecting the massive data generated from multiple data sources (DSs) and transmitting them to the control units (CUs) for further data processing and analysis. These ever-increasing data demands require not only more and more capacity of the transmission channels but also results in resource over-provision to meet the resilience requirements, thus the unavoidable waste as a result of the data fluctuations throughout the day. In addition, the high energy consumption (EC) and carbon discharge from these data transmissions posing serious issues to the environment we live in. Therefore, to overcome the issues of intensive EC and carbon emission (CE) of massive data dissemination in SCs, we propose an energy-efficient and carbon reduction approach by using the daily mobility of the existing vehicles as an alternative communications channel to accommodate the data dissemination in SCs. To illustrate the effectiveness and efficiency of our approach, we take the Auckland City in New Zealand as an example, assuming massive data generated by various sources geographically scattered throughout the Auckland region, to the control centres located in the city. Results obtained show that our proposed approach can provide up to four times faster transferring the large volume of data by using the existing daily vehicles' mobility, than the conventional transmission network. Moreover, our proposed approach offers about 32% less EC and CE than that of conventional network transmission approach.
2017 IEEE International Conference on Agents (ICA)
The effects of asynchronous updating of the state of the agents in a Cucker-Smale flocking model ... more The effects of asynchronous updating of the state of the agents in a Cucker-Smale flocking model is considered in this paper. The study of asynchronous update is important because in practical implementations the agents typically have internal clocks that are not synchronized. We considered how asynchronous update will affect the time it takes to achieve flocking (flocking time) as well as how close the agents in the flock are moving (flock diameter). These factors were largely ignored in most previous works as achieving asymptotic convergence was their main focus. Furthermore, previous simulations typically assume that the agents move with the same speed. We considered the effects of achieving consensus of both the speed and the heading. Through computer simulations, we showed that both the flocking time and flock diameter increase significantly with asynchronicity. Results also showed that the diameter of the flock is substantially larger when the agents start with different speeds. These results should be taken into account when designing flocking agents in practice.
2021 IEEE International Conference on Data Mining (ICDM), 2021
When applying supervised learning to estimate class distributions of unlabelled samples (so-calle... more When applying supervised learning to estimate class distributions of unlabelled samples (so-called quantification), dataset shift is an expected yet challenging problem. Existing quantification methods make strong assumptions on the nature of dataset shift that often will not hold in practice. We propose a novel Gain-Some-Lose-Some (GSLS) model that accounts for more general conditions of dataset shift. We present a method for fitting the GSLS model without any labelled instances from the target sample, and experimentally demonstrate that GSLS can produce reliable quantification prediction intervals under broader conditions of shift than existing quantification methods.
Testing and validation of the functionalities and safety of automated vehicles shifted from a dis... more Testing and validation of the functionalities and safety of automated vehicles shifted from a distance-based to a scenario-based method in the past decade. A number of domain-specific languages and systems were developed to support scenario-based testing. The aim of this paper is to review and compare the features and characteristics of the major scenario description languages and systems (SDLS). Each of them is designed for different purposes and with different goals; therefore, they have their strengths and weaknesses. Their characteristics are highlighted with an example nontrivial traffic scenario that we designed. We also discuss some directions for further development and research of these SDLS.
Centralised machine learning brings in side effect pertaining to privacy preservation, most of ma... more Centralised machine learning brings in side effect pertaining to privacy preservation, most of machine learning methods prone to using the frameworks without privacy protection, as current methods for privacy preservation will slow down model training and testing. In order to resolve this problem, we develop a new noise generating method based on information entropy by using differential privacy for betterment the privacy protection which owns the architecture of federated machine learning. Our experiments unveil that this solution effectively preserves privacy in the vein of centralized federated learning. The gained accuracy is promising which has a room to be uplifted.
The emerging smart city paradigm e.g., intelligent transport, smart grid and participatory sensin... more The emerging smart city paradigm e.g., intelligent transport, smart grid and participatory sensing etc. is to advance the quality, performance and experience of urban citizten services through greater connectivity. This paradigm needs to collect data from citizens, various devices and assets that could be monitored, processed and analysed for the city governers to make better decision and also more efficiently manage those assests and resources. While the telecommunication and Internet are progressively being over-burdened and congested by the growing data transmission demands. To keep expanding the telecommunications and Internet infrastructures to accomodate these intensive data demands is costly and also the associated energy consumptions and carbon emissions could at long last wind up genuinely hurting the environment. To face this issue in the coming era of big data, we envision it will be best to utilize the established urban transport and road infrastructure and existing dail...
School of Electronics and Information Technology, Sun Yat-sen University, Guangzhou, China, Schoo... more School of Electronics and Information Technology, Sun Yat-sen University, Guangzhou, China, School of Engineering, Computer and Mathematical Sciences, Auckland University of Technology, Auckland, New Zealand, School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu, China, College of Information Engineering (College of Artificial Intelligence), Yangzhou University, Yangzhou, China, School of Computer Science and Electronic Engineering, University of Essex, Colchester, United Kingdom, Department of Electrical and Computer Engineering, New Jersey Institute of Technology, Newark, NJ, United States, Department of Electronics and Communication Engineering, Institute of Information Technology (IIIT-Delhi), New Delhi, India, Department of Computer and Communication Systems Engineering, Putra Malaysia University, Selangor Darul Ehsan, Malaysia
Attention mechanisms have been incorporated into many neural network-based natural language proce... more Attention mechanisms have been incorporated into many neural network-based natural language processing (NLP) models. They enhance the ability of these models to learn and reason with long input texts. A critical part of such mechanisms is the computation of attention similarity scores between two elements of the texts using a similarity score function. Given that these models have different architectures, it is difficult to comparatively evaluate the effectiveness of different similarity score functions. In this paper, we proposed a baseline model that captures the common components of recurrent neural network-based Question Answering (QA) systems found in the literature. By isolating the attention function, this baseline model allows us to study the effects of different similarity score functions on the performance of such systems. Experimental results show that a trilinear function produced the best results among the commonly used functions. Based on these insights, a new T-trilin...
2021 IEEE Global Communications Conference (GLOBECOM), 2021
In order to perform competitive privacy-guaranteed object detection, we propose an end-to-end mod... more In order to perform competitive privacy-guaranteed object detection, we propose an end-to-end model called Privacy-preserving Deep Transformation Self-attention (PPDTSA). This model ensures the privacy of the inference results. It has a low-complexity hierarchical structure with a relatively small number of hyper-parameters. Consistency of prediction is achieved through the encoding and decoding blocks of the self-attention mechanism which enables points of interest to be located. Focus loss is estimated based on foreground-background imbalance. The remaining dense blocks enable image details to be retained and the Region Of Interest to be expanded. At the same time, the objects detected in the image are protected through the privacy noise volume which is specified by the user. Experimental results demonstrate that PPDTSA achieves superior performance on the MOT20 dataset compared with three other state-of-the-art object detection models.
Practical Digital Signal Processing for Engineers and Technicians Edmund Lai PhD, BEng; Lai and A... more Practical Digital Signal Processing for Engineers and Technicians Edmund Lai PhD, BEng; Lai and Associates, Singapore . ... Newnes An imprint of Elsevier Linacre House, Jordan Hill, Oxford OX2 8DP 200 Wheeler Road, Burlington, MA 01803 First published 2003 Copyright ...
The analytic process of parametrising manifolds is well known to Mathematicians and Physicists, [... more The analytic process of parametrising manifolds is well known to Mathematicians and Physicists, [1] and is becoming increasingly familiar to Engi-neers concerned with, for example, robot manip-ulator trajectories [2]. The numerical equivalent does not appear to ...
2017 2nd Workshop on Recent Trends in Telecommunications Research (RTTR), 2017
In recent years, research into localization systems has become more popular as the proliferation ... more In recent years, research into localization systems has become more popular as the proliferation of Wireless Sensor Networks (WSNs) grows. Wireless Localization can refer to either an “Active” system which tracks a mobile transceiver, or “Passive” localization which tracks a transceiver free entity by measuring the changes it makes to the surrounding wireless environment. Recent work has seen both of these systems implemented with Received Signal Strength Indication (RSSI) values from transceivers. Many algorithms and channel models have been presented to increase the accuracy of a Received Signal Strength (RSS) based system. In this paper we experimentally check whether RSSI values map to the expected RSS values within an IEEE 802.15.4 network. Indoor experiments are repeated within an ideal outdoor environment, with multiple device platforms, to eliminate indoor multipath propagation as the cause for inconsistent behavior of RSSI. We identify 3 key issues with raw RSSI values and present either a possible solution or a mitigation strategy to reduce their effect. We conclude that using raw RSSI values is flawed, as the premise that they uniquely map to the distance between transceivers is incorrect. However they may be calibrated to increase their accuracy, and therefore viability.
Device-free localization (DFL) algorithms using the received signal strength indicator (RSSI) met... more Device-free localization (DFL) algorithms using the received signal strength indicator (RSSI) metrics have become a popular research focus in recent years as they allow for location-based service using commercial-off-the-shelf (COTS) wireless equipment. However, most existing DFL approaches have limited applicability in realistic smart home environments as they typically require extensive offline calibration, large node densities, or use technology that is not readily available in commercial smart homes. In this paper, we introduce SpringLoc and a DFL algorithm that relies on simple parameter tuning and does not require offline measurements. It localizes and tracks an entity using an adaptive spring relaxation approach. The anchor points of the artificial springs are placed in regions containing the links that are affected by the entity. The affected links are determined by comparing the kernel-based histogram distance of successive RSSI values. SpringLoc is benchmarked against existing algorithms in two diverse and realistic environments, showing significant improvement over the state-of-the-art, especially in situations with low-node deployment density. INDEX TERMS Device-free localization (DFL), histogram distance, indoor positioning systems (IPS), smart homes, spring-relaxation.
Device-free localization (DFL) systems that rely on the wireless received signal strength indicat... more Device-free localization (DFL) systems that rely on the wireless received signal strength indicator (RSSI) metric to localize targets with no device attached to them have been reported in the literature for almost a decade. Approaches using RSSI can be split into three main categories. Link-based approaches utilize weighted summation or probabilistic methods to infer location. Location-based approaches create a fingerprint map of an area. Radio Tomographic Imaging treat DFL as an imaging problem solved with a linear inverse. In this article, we implement and investigate the performance of all three major RSSI approaches in two test environments. We demonstrate how different environments and walking trajectories can have significant effects on the localization accuracy. The experimental results lead us to the conclusion that without implementing and testing within the same environment for the same target trajectories, the performance of various classes of DFL systems cannot be reliably evaluated. Relying on the stated accuracy from the literature for comparison is a flawed premise.
2012 IEEE International Symposium on Circuits and Systems, 2012
This paper presents the design of a variable linear phase finite impulse response filter based on... more This paper presents the design of a variable linear phase finite impulse response filter based on second order frequency transformations and coefficient decimation. The design of variable digital filters (VDFs) using first and second order frequency transformations have been proposed in literature. The VDF using second order transformation has better cut-off slope characteristics compared to the VDF using first order transformation. However, the former has the drawback of limited range (approximately 25% of the half of the sampling frequency) over which the cut-off frequency, fc, can be varied. It also fails to provide variable lowpass, highpass, bandpass or bandstop responses from a fixed-coefficient lowpass filter using the same architecture. The architecture proposed here overcomes the above mentioned disadvantages using coefficient decimation technique. The design example shows that the range over which fc can be varied is 2.65 times wider in the proposed VDF than the VDF in [7] and for a given frequency range, the proposed VDF offers a total gate count saving of 33% and 41% over the VDF in [11] and [7] respectively. Also, the proposed architecture provides variable lowpass, highpass, bandpass or bandstop responses from a fixed coefficient lowpass filter.
IEEE Transactions on Circuits and Systems II: Express Briefs, 2014
This brief presents a low complexity linear phase variable digital filter (VDF) design with tunab... more This brief presents a low complexity linear phase variable digital filter (VDF) design with tunable lowpass (LP), highpass (HP), bandpass (BP) and bandstop (BS) responses anywhere over entire Nyquist band. The spectral parameter approximation based VDFs (SPA-VDFs) designed using Farrow structure have advantages of linear phase, lower group delay and fewer variable multipliers. However, the total gate count and dynamic range of filter coefficient values of SPA-VDFs increase significantly with the tunable range of cutoff frequency which limits their usefulness in emerging signal processing and wireless communication applications. Also, existing VDFs need to update either filter coefficients or need parallel filter structures to obtain variable LP, HP, BP and BS responses. In this paper, a new VDF design is proposed by deftly integrating SPA-VDF with the modified coefficient decimation method (MCDM) and it shall be referred to as SPA-MCDM-VDF. The SPA-MCDM-VDF provides LP, HP, BP and BS responses with unabridged center frequency and bandwidth control over the entire Nyquist band without the need for hardware re-implementation or coefficient update. The complexity comparisons show that the SPA-MCDM-VDF offers substantial savings in gate count, group delay and number of variable multiplications over other linear phase VDFs.
Page 1. On Minimum Sampling Rates for Signals in Shift Invariant Spaces Beilei Huang, Edmund MK. ... more Page 1. On Minimum Sampling Rates for Signals in Shift Invariant Spaces Beilei Huang, Edmund MK. Lai Institute of Information Sciences & Technology Massey University, Wellington, New Zealand AP Vinod School of Computer ...
2017 27th International Telecommunication Networks and Applications Conference (ITNAC), 2017
One of the main challenges of operating a smart city (SC) is collecting the massive data generate... more One of the main challenges of operating a smart city (SC) is collecting the massive data generated from multiple data sources (DSs) and transmitting them to the control units (CUs) for further data processing and analysis. These ever-increasing data demands require not only more and more capacity of the transmission channels but also results in resource over-provision to meet the resilience requirements, thus the unavoidable waste as a result of the data fluctuations throughout the day. In addition, the high energy consumption (EC) and carbon discharge from these data transmissions posing serious issues to the environment we live in. Therefore, to overcome the issues of intensive EC and carbon emission (CE) of massive data dissemination in SCs, we propose an energy-efficient and carbon reduction approach by using the daily mobility of the existing vehicles as an alternative communications channel to accommodate the data dissemination in SCs. To illustrate the effectiveness and efficiency of our approach, we take the Auckland City in New Zealand as an example, assuming massive data generated by various sources geographically scattered throughout the Auckland region, to the control centres located in the city. Results obtained show that our proposed approach can provide up to four times faster transferring the large volume of data by using the existing daily vehicles' mobility, than the conventional transmission network. Moreover, our proposed approach offers about 32% less EC and CE than that of conventional network transmission approach.
2017 IEEE International Conference on Agents (ICA)
The effects of asynchronous updating of the state of the agents in a Cucker-Smale flocking model ... more The effects of asynchronous updating of the state of the agents in a Cucker-Smale flocking model is considered in this paper. The study of asynchronous update is important because in practical implementations the agents typically have internal clocks that are not synchronized. We considered how asynchronous update will affect the time it takes to achieve flocking (flocking time) as well as how close the agents in the flock are moving (flock diameter). These factors were largely ignored in most previous works as achieving asymptotic convergence was their main focus. Furthermore, previous simulations typically assume that the agents move with the same speed. We considered the effects of achieving consensus of both the speed and the heading. Through computer simulations, we showed that both the flocking time and flock diameter increase significantly with asynchronicity. Results also showed that the diameter of the flock is substantially larger when the agents start with different speeds. These results should be taken into account when designing flocking agents in practice.
2021 IEEE International Conference on Data Mining (ICDM), 2021
When applying supervised learning to estimate class distributions of unlabelled samples (so-calle... more When applying supervised learning to estimate class distributions of unlabelled samples (so-called quantification), dataset shift is an expected yet challenging problem. Existing quantification methods make strong assumptions on the nature of dataset shift that often will not hold in practice. We propose a novel Gain-Some-Lose-Some (GSLS) model that accounts for more general conditions of dataset shift. We present a method for fitting the GSLS model without any labelled instances from the target sample, and experimentally demonstrate that GSLS can produce reliable quantification prediction intervals under broader conditions of shift than existing quantification methods.
Testing and validation of the functionalities and safety of automated vehicles shifted from a dis... more Testing and validation of the functionalities and safety of automated vehicles shifted from a distance-based to a scenario-based method in the past decade. A number of domain-specific languages and systems were developed to support scenario-based testing. The aim of this paper is to review and compare the features and characteristics of the major scenario description languages and systems (SDLS). Each of them is designed for different purposes and with different goals; therefore, they have their strengths and weaknesses. Their characteristics are highlighted with an example nontrivial traffic scenario that we designed. We also discuss some directions for further development and research of these SDLS.
Centralised machine learning brings in side effect pertaining to privacy preservation, most of ma... more Centralised machine learning brings in side effect pertaining to privacy preservation, most of machine learning methods prone to using the frameworks without privacy protection, as current methods for privacy preservation will slow down model training and testing. In order to resolve this problem, we develop a new noise generating method based on information entropy by using differential privacy for betterment the privacy protection which owns the architecture of federated machine learning. Our experiments unveil that this solution effectively preserves privacy in the vein of centralized federated learning. The gained accuracy is promising which has a room to be uplifted.
The emerging smart city paradigm e.g., intelligent transport, smart grid and participatory sensin... more The emerging smart city paradigm e.g., intelligent transport, smart grid and participatory sensing etc. is to advance the quality, performance and experience of urban citizten services through greater connectivity. This paradigm needs to collect data from citizens, various devices and assets that could be monitored, processed and analysed for the city governers to make better decision and also more efficiently manage those assests and resources. While the telecommunication and Internet are progressively being over-burdened and congested by the growing data transmission demands. To keep expanding the telecommunications and Internet infrastructures to accomodate these intensive data demands is costly and also the associated energy consumptions and carbon emissions could at long last wind up genuinely hurting the environment. To face this issue in the coming era of big data, we envision it will be best to utilize the established urban transport and road infrastructure and existing dail...
Uploads
Papers by Edmund M Lai