David Tien

This paper introduces our Agnostic Interpretable Machine Learning for Knowledge Discovery (AIM4KD) framework. Our approach to emotional stress detection is innovative, as it integrates dataset preprocessing, classification, knowledge discovery techniques, and visualisation methods to augment traditional machine learning methodologies. The key innovation is our departure from conventional machine learning strategies, adopting an agnostic approach that involves preprocessing multiple dataset versions and employing multiple classification algorithms with dynamic configuration settings. This strategy has yielded high-quality logic rules, ensuring that the most significant patterns surface for effective knowledge discovery. We validated the framework through extensive experiments using the DEAP dataset, a crucial resource in Affective Computing for examining physiological changes during emotional episodes. When benchmarked against traditional preprocessing and classification methods using tree and forest algorithms, AIM4KD produced, on average, an impressive 22-fold increase in high-quality rules. These experiments highlight the framework’s effectiveness in accurately detecting emotional stress states. Furthermore, the findings demonstrate AIM4KD’s capacity to handle complex multimodal datasets and underscore the critical role of interpretability in machine learning. This research emphasises AIM4KD’s potential to improve decision-making processes in physiological data analysis and its applicability across various domains featuring intricate time-series data. For instance, experts in health or finance can utilise AIM4KD to understand the underlying time series dynamics. To encourage broader adoption and facilitate future research, we will release the AIM4KD framework code and the DEAP dataset preprocessing script publicly at https://github.com/CyberNaut-AU/AIM4KD upon the paper’s final acceptance.

PurposeAlzheimer’s disease (AD) is becoming more prevalent among the elderly, with projections indicating that it will affect a significantly large population in the future. Regardless of substantial research efforts and investments focused on exploring the underlying biological factors, a definitive cure has yet to be discovered. The currently available treatments are only effective in slowing disease progression if it is identified in the early stages of the disease. Therefore, early diagnosis has become critical in treating AD.ApproachRecently, the use of deep learning techniques has demonstrated remarkable improvement in enhancing the precision and speed of automatic AD diagnosis through medical image analysis. We propose a hybrid model that integrates a convolutional variational auto-encoder (CVAE) with a vision transformer (ViT). During the encoding phase, the CVAE captures key features from the MRI scans, whereas the decoding phase reduces irrelevant details in MRIs. These refined inputs enhance the ViT’s ability to analyze complex patterns through its multihead attention mechanism.ResultsThe model was trained and evaluated using 14,000 structural MRI samples from the ADNI and SCAN databases. Compared with three benchmark methods and previous studies with Alzheimer’s classification techniques, our approach achieved a significant improvement, with a test accuracy of 93.3%.ConclusionsThrough this research, we identified the potential of the CVAE-ViT hybrid approach in detecting minor structural abnormalities related to AD. Integrating unsupervised feature extraction via CVAE can significantly enhance transformer-based models in distinguishing between stages of cognitive impairment, thereby identifying early indicators of AD.

A method to solve the convex problems of nondifferentiable optimization relying on the basic philosophy of the method of conjugate gradients and coinciding with it in the case of quadratic functions was presented. Its basic distinction from the earlier counterparts lies in the a priori fixed constraint on the memory size which is independent of the accuracy of the resulting solution. Numerical experiments suggest practically linear rate of convergence of this algorithm.

Automatic detection of text on road signs can help drivers keep aware of the traffic situation and surrounding environments by reminding them of the signs ahead. Current systems can only detect constrained road signs or produce unsatisfying performance when dealing with complex scenes in practical use. This paper firstly reviews the existing techniques used for text detection from natural scene. A novel system which detects text on road signs from natural scene video is then proposed. Our detailed approaches and methodology give a promising solution to this problem in order to reduce the running time and improve the recognition rate

Multi-carrier direct-sequence code division multiple access (MC-DS-CDMA) system is very attractive in high-bit-rate wireless communication because of the good bandwidth efficiency, robustness to the frequency selective multi-path fading. However, MC-DS-CDMA inherently has a major drawback, which is the high peak-to-average-power ratio (PAPR) problem. In this paper, we propose a new efficient PAPR reduction method based on the spectrum phase control (SPC). The proposed method rotates the phase of each DS-CDMA spectrum for reducing the PAPR. Also, it uses the disjointed-spectral characteristic of MC-DS-CDMA signal in order to insert the tones for conveying the side information of SPC. The simulation results show the PAPR reduction capability and the improvement in nonlinear environment. Proposed method can reduce more than 4.6dB PAPR reduction of MC-DS-CDMA signal, and total power degradation can be improved by about 3.42 dB when TWTA (traveling wave tube amplifier) is used.

In this paper we analyze a new language called game Pidgin language (GPL) for speech interactive computer games. We use two formulas namely entropy and perfect information of Information Theory to perform the analysis. We establish that GPL is an efficient language for communication in real time speech interactive systems.