Machine Learning (ML) technologies face prolonged challenges with low user acceptance of delivered solutions as well as seeing system misuse, disuse, or even failure. These fundamental challenges can be attributed to the nature of the “black-box” of machine learning methods with trust issues when offering ML-based solutions.
Dr. Zhou is a leading senior researcher in trustworthy and transparent machine learning, and has done pioneering research in the area of linking human and machine learning. He also works with industries in advanced data analytics for transforming data into actionable operations particularly by incorporating human user aspects into machine learning and translate machine learning into impacts in real world applications.
- IEEE Senior Member
- Guest Editor: Journal of Visual Languages and Computing (Elsevier)
- Associate Editor: Behaviour & Information Technology (Taylor&Francis)
Can supervise: YES
- Data analytics
- Explainable machine learning
- Trustworthy machine learning
- Behaviour analytics
- Cognitive computing
- Human-computer interaction
- Visual analytics
This is the first book of its kind to systematically understand the current active research activities and outcomes related to human and machine learning.
Chen, F, Zhou, J, Wang, Y, Yu, K, Arshad, SZ, Khawaji, A & Conway, D 2016, Robust Multimodal Cognitive Load Measurement, Springer.
This book explores robust multimodal cognitive load measurement with physiological and behavioural modalities, which involve the eye, Galvanic Skin Response, speech, language, pen input, mouse movement and multimodality fusions.
Zhou, J 2012, Gaining insights into volumetric data visualization: a semi-automatic transfer function generation approach using contour tree analyses, LAP Lambert Academic Publishing, Germany.
Zhou, J & Xiao, C 2007, Computer Graphics Theory and OpenGL Programming (In Chinese), South China University of Technology Press, Guangzhou, China.
Zhou, J 1999, Mastering AutoCAD 2000 (In Chinese), Chongqing University Press, Chongqing, China.
Cui, H, Wang, X, Zhou, J, Gong, G, Eberl, S, Yin, Y, Wang, L, Feng, D & Fulham, M 2018, 'A topo-graph model for indistinct target boundary definition from anatomical images.', Computer methods and programs in biomedicine, vol. 159, pp. 211-222.View/Download from: UTS OPUS or Publisher's site
BACKGROUND AND OBJECTIVE:It can be challenging to delineate the target object in anatomical imaging when the object boundaries are difficult to discern due to the low contrast or overlapping intensity distributions from adjacent tissues. METHODS:We propose a topo-graph model to address this issue. The first step is to extract a topographic representation that reflects multiple levels of topographic information in an input image. We then define two types of node connections - nesting branches (NBs) and geodesic edges (GEs). NBs connect nodes corresponding to initial topographic regions and GEs link the nodes at a detailed level. The weights for NBs are defined to measure the similarity of regional appearance, and weights for GEs are defined with geodesic and local constraints. NBs contribute to the separation of topographic regions and the GEs assist the delineation of uncertain boundaries. Final segmentation is achieved by calculating the relevance of the unlabeled nodes to the labels by the optimization of a graph-based energy function. We test our model on 47 low contrast CT studies of patients with non-small cell lung cancer (NSCLC), 10 contrast-enhanced CT liver cases and 50 breast and abdominal ultrasound images. The validation criteria are the Dice's similarity coefficient and the Hausdorff distance. RESULTS:Student's t-test show that our model outperformed the graph models with pixel-only, pixel and regional, neighboring and radial connections (p-values <0.05). CONCLUSIONS:Our findings show that the topographic representation and topo-graph model provides improved delineation and separation of objects from adjacent tissues compared to the tested models.
Oviatt, S, Hang, K, Zhou, J, Yu, K & Chen, F 2018, 'Dynamic handwriting signal features predict domain expertise', ACM Transactions on Interactive Intelligent Systems, vol. 8, no. 3.View/Download from: UTS OPUS or Publisher's site
© 2018 ACM. As commercial pen-centric systems proliferate, they create a parallel need for analytic techniques based on dynamic writing.Within educational applications, recent empirical research has shown that signal-level features of students' writing, such as stroke distance, pressure and duration, are adapted to conserve total energy expenditure as they consolidate expertise in a domain. The present research examined how accurately three different machine-learning algorithms could automatically classify users' domain expertise based on signal features of their writing, without any content analysis. Compared with an unguided machine-learning classification accuracy of 71%, hybrid methods using empirical-statistical guidance correctly classified 79-92% of students by their domain expertise level. In addition to improved accuracy, the hybrid approach contributed a causal understanding of prediction success and generalization to new data. These novel findings open up opportunities to design new automated learning analytic systems and student-adaptive educational technologies for the rapidly expanding sector of commercial pen systems.
Wang, X, Cui, H, Gong, G, Fu, Z, Zhou, J, Gu, J, Yin, Y & Feng, D 2018, 'Computational delineation and quantitative heterogeneity analysis of lung tumor on 18F-FDG PET for radiation dose-escalation.', Scientific reports, vol. 8, no. 1.View/Download from: UTS OPUS or Publisher's site
Quantitative measurement and analysis of tumor metabolic activities could provide a more optimal solution to personalized accurate dose painting. We collected PET images of 58 lung cancer patients, in which the tumor exhibits heterogeneous FDG uptake. We design an automated delineation and quantitative heterogeneity measurement of the lung tumor for dose-escalation. For tumor delineation, our algorithm firstly separates the tumor from its adjacent high-uptake tissues using 3D projection masks; then the tumor boundary is delineated with our stopping criterion of joint gradient and intensity affinities. For dose-escalation, tumor sub-volumes with low, moderate and high metabolic activities are extracted and measured. Based on our quantitative heterogeneity measurement, a sub-volume oriented dose-escalation plan is implemented in intensity modulated radiation therapy (IMRT) planning system. With respect to manual tumor delineations by two radiation oncologists, the paired t-test demonstrated our model outperformed the other computational methods in comparison (p < 0.05) and reduced the variability between inter-observers. Compared to standard uniform dose prescription, the dosimetry results demonstrated that the dose-escalation plan statistically boosted the dose delivered to high metabolic tumor sub-volumes (p < 0.05). Meanwhile, the doses received by organs-at-risk (OAR) including the heart, ipsilateral lung and contralateral lung were not statistically different (p > 0.05).
Zhou, J & Chen, F 2018, 'DecisionMind: revealing human cognition states in data analytics-driven decision making with a multimodal interface', Journal on Multimodal User Interfaces, vol. 12, no. 2, pp. 67-76.View/Download from: UTS OPUS or Publisher's site
© 2017, Springer International Publishing AG. Despite the recognized value of machine learning (ML) techniques and high expectation of applying ML techniques within various applications, significant barriers to widespread adoption and local implementation of ML approaches still exist in the areas of trust (of ML results), comprehension (of ML processes) and related workload, as well as confidence (in decision making based on ML results) by users. This paper argues that the revealing of human cognition states with a multimodal interface during ML-based data analytics-driven decision making could provide a rich view for both ML researchers and domain experts to learn the effectiveness of ML technologies in applications. On the one hand, human cognition states could help understand to what degree users accept innovative technologies. On the other hand, through understanding human cognition states during data analytics-driven decision making, ML-based decision attributes and even ML models can be adaptively refined in order to make ML transparent. The paper also identifies examples of impact challenges and obstacles, as well as high-demand research directions in making ML transparent.
Zhou, J, Arshad, SZ, Wang, X, Li, Z, Feng, D & Chen, F 2018, 'End-User Development for Interactive Data Analytics: Uncertainty, Correlation and User Confidence', IEEE TRANSACTIONS ON AFFECTIVE COMPUTING, vol. 9, no. 3, pp. 383-395.View/Download from: Publisher's site
Chen, F, Zhou, J & Yu, K 2017, 'Multimodal and data-driven cognitive load measurement', Cognitive Load Measurement and Application: A Theoretical Framework for Meaningful Research and Practice, pp. 147-178.
Chen, Z, Zhou, J & Wang, X 2017, 'Visual Analytics of Movement Pattern Based on Time-Spatial Data: A Neural Net Approach', arXiv preprint arXiv:1707.02554.
Zheng, C, Wang, X, Zeng, S, Zhou, J, Yin, Y, Feng, D & Fulham, M 2017, 'Topology-guided deformable registration with local importance preservation for biomedical images', Physics in Medicine & Biology, vol. 63, pp. 015028-015028.
Zheng, C, Wang, X, Zeng, S, Zhou, J, Yin, Y, Feng, D & Fulham, M 2017, 'Topology-guided deformable registration with local importance preservation for biomedical images.', Physics in medicine and biology, vol. 63, no. 1, pp. 015028-015028.View/Download from: UTS OPUS or Publisher's site
The demons registration (DR) model is well recognized for its deformation capability. However, it might lead to misregistration due to erroneous diffusion direction when there are no overlaps between corresponding regions. We propose a novel registration energy function, introducing topology energy, and incorporating a local energy function into the DR in a progressive registration scheme, to address these shortcomings. The topology energy that is derived from the topological information of the images serves as a direction inference to guide diffusion transformation to retain the merits of DR. The local energy constrains the deformation disparity of neighbouring pixels to maintain important local texture and density features. The energy function is minimized in a progressive scheme steered by a topology tree graph and we refer to it as topology-guided deformable registration (TDR). We validated our TDR on 20 pairs of synthetic images with Gaussian noise, 20 phantom PET images with artificial deformations and 12 pairs of clinical PET-CT studies. We compared it to three methods: (1) free-form deformation registration method, (2) energy-based DR and (3) multi-resolution DR. The experimental results show that our TDR outperformed the other three methods in regard to structural correspondence and preservation of the local important information including texture and density, while retaining global correspondence.
Zhou, J, Sun, J, Wang, Y & Chen, F 2017, 'Wrapping practical problems into a machine learning framework: Using water pipe failure prediction as a case study', International Journal of Intelligent Systems Technologies and Applications, vol. 16, pp. 191-207.
Zhou, J, Sun, J, Wang, Y & Chen, F 2017, 'Wrapping practical problems into a machine learning framework: Using water pipe failure prediction as a case study', International Journal of Intelligent Systems Technologies and Applications, vol. 16, no. 3, pp. 191-207.View/Download from: Publisher's site
Copyright © 2017 Inderscience Enterprises Ltd. Despite the recognised value of machine learning (ML) techniques and high expectation of applying ML techniques within various applications, users often find it difficult to effectively apply ML techniques in practice because of complicated interfaces between ML algorithms and users. This paper presents a work flow of wrapping practical problems into an ML framework. The water pipe failure prediction is used as a case study to show that the applying process can be divided into various steps: obtain domain data, interview with domain experts, clean/pre-process and preview original domain data, extract ML features, set up ML models, explain ML results and make decisions, as well as make feedback to the system based on decision making. In this process, domain experts and ML developers need to collaborate closely in order to make this workflow more effective.
Cui, H, Wang, X, Lin, W, Zhou, J, Eberl, S, Feng, D & Fulham, M 2016, 'Primary lung tumor segmentation from PET-CT volumes with spatial-topological constraint', INTERNATIONAL JOURNAL OF COMPUTER ASSISTED RADIOLOGY AND SURGERY, vol. 11, no. 1, pp. 19-29.View/Download from: Publisher's site
Wei, WH, Zhou, JL, Tao, M & Yuan, HQ 2016, 'Constrained differential evolution using opposition-based learning', Tien Tzu Hsueh Pao/Acta Electronica Sinica, vol. 44, no. 2, pp. 426-436.View/Download from: Publisher's site
© 2016, Chinese Institute of Electronics. All right reserved. Differential evolution is a global heuristic algorithm, which is simple, easy-to-use and robust in practice. Combining with the constraint-handling techniques, it can solve constrained optimization problems. Machine learning often guides population to evolve in the evolution computation, and is widely applied to unconstrained differential evolution algorithm. However, machine learning is rarely applied to constrained differential evolution algorithm, so this paper proposed a constrained differential evolution algorithm framework using opposition-based learning. The algorithm can improve the diversity and convergence of differential evolution. At last, the proposed algorithm framework is applied to two popular constrained differential evolution variants, that is (μ+λ)-CDE and ECHT-DE. And 18 benchmark functions presented in CEC 2010 are chosen as the test suite, experimental results show that comparing with (μ+λ)-CDE and ECHT-DE, our algorithms are able to improve global search ability, convergence speed and accuracy in the majority of test cases.
Wei, W, Zhou, J, Chen, F & Yuan, H 2016, 'Constrained differential evolution using generalized opposition-based learning', SOFT COMPUTING, vol. 20, no. 11, pp. 4413-4437.View/Download from: Publisher's site
Zhou, J, Wang, X, Cui, H, Gong, P, Miao, X, Miao, Y, Xiao, C, Chen, F & Feng, D 2016, 'Topology-aware illumination design for volume rendering', BMC BIOINFORMATICS, vol. 17.View/Download from: UTS OPUS or Publisher's site
Cui, H, Wang, X, Zhou, J, Eberl, S, Feng, D & Fulham, M 2015, 'Improved segmentation accuracy for thoracic PET-CT in patients with NSCLC using a multi-graph model (MGM)', Journal of Nuclear Medicine, vol. 56, pp. 2527-2527.
Cui, H, Wang, X, Zhou, J, Eberl, S, Yin, Y, Feng, D & Fulham, M 2015, 'Topology polymorphism graph for lung tumor segmentation in PET-CT images', PHYSICS IN MEDICINE AND BIOLOGY, vol. 60, no. 12, pp. 4893-4914.View/Download from: Publisher's site
Uemura, Y, Kajiwara, Y, Zhou, J, Chen, F & Shimakawa, H 2015, 'Estimating Human Physical States from Chronological Gait Features Acquired with RFID Technology', Sensors & Transducers, vol. 194, pp. 76-76.
Copyright © 2015 Inderscience Enterprises Ltd. Despite the recognised value of machine learning (ML) techniques and high expectation of applying ML techniques within various applications, users often find it difficult to effectively apply ML techniques in practise because of complicated interfaces between ML algorithms and users. This paper focuses on investigating making ML useable from the point of view of how human-computer interaction (HCI) techniques benefit ML in order to simplify the interface between users and ML algorithms. We formulate possible research directions in making ML useable based on human factors, decision making and trust in ML. We strongly believe that a trustworthy decision making based on ML results, which is the ultimate goal of ML-based applications, contributes to the overall application performance and makes ML more useable. Two case studies of measurable decision making and revealing internal states of ML process are presented to show how HCI techniques are used to make ML useable.
Zhou, J, Sun, J, Chen, F, Wang, Y, Taib, R, Khawaji, A & Li, Z 2015, 'Measurable Decision Making with GSR and Pupillary Analysis for Intelligent User Interface', ACM TRANSACTIONS ON COMPUTER-HUMAN INTERACTION, vol. 21, no. 6.View/Download from: Publisher's site
Peng, H, Tang, J, Xiao, H, Bria, A, Zhou, J, Butler, V, Zhou, Z, Gonzalez-Bellido, PT, Oh, SW, Chen, J, Mitra, A, Tsien, RW, Zeng, H, Ascoli, GA, Iannello, G, Hawrylycz, M, Myers, E & Long, F 2014, 'Virtual finger boosts three-dimensional imaging and microsurgery as well as terabyte volume image visualization and analysis', NATURE COMMUNICATIONS, vol. 5.View/Download from: UTS OPUS or Publisher's site
Zhou, J, Xiao, C & Takatsuka, M 2013, 'A multi-dimensional importance metric for contour tree simplification', JOURNAL OF VISUALIZATION, vol. 16, no. 4, pp. 341-349.View/Download from: Publisher's site
Long, F, Zhou, J & Peng, H 2012, 'Visualization and analysis of 3D microscopic images', PLoS Computational Biology, vol. 8.
Zhou, J, Lee, I, Thomas, B, Menassa, R, Farrant, A & Sansome, A 2012, 'In-Situ Support for Automotive Manufacturing Using Spatial Augmented Reality', International Journal of Virtual Reality, vol. 11, pp. 33-41.
Zhou, J, Lee, I, Thomas, BH, Sansome, A & Menassa, R 2011, 'Facilitating Collaboration with Laser Projector-Based Spatial Augmented Reality in Industrial Applications', Recent Trends of Mobile Collaborative Augmented Reality Systems, pp. 161-173.
XIAO, C & ZHOU, J-L 2010, 'Optimized modulation of rendering parameters in volume rendering based on parallel coordinates', Application Research of Computers, pp. 12-12.
Xiao, C, Zhou, J-L & Wang, Z-Y 2010, 'Depiction of structural relationships between objects in volumetric data', Jisuanji Yingyong/ Journal of Computer Applications, vol. 30, pp. 3288-3291.
Zhou, J, Wang, Z & Xiao, C 2008, 'Focal region based volume rendering by texture mapping and GPU based approach', Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal of Computer-Aided Design and Computer Graphics, vol. 20, no. 1, pp. 32-37.
This paper presents an approach of using texture mapping and GPU based method for realizing focal region based volume rendering. The proposed method first labels the volumetric data into three parts based on the mechanism of stencil buffer test. Then the data model is rendered using texture mapping method. The approach also adopts GPU-based method to render the context region with volume silhouettes, and also realize the multiple transfer function specification process based on pixel transfer function. The proposed method shows its good performance in the implementation and function extension.
Zhou, J, Xiao, C, Wang, Z & Takatsuka, M 2008, 'A concept of volume rendering guided search process to analyze medical data set', Computerized Medical Imaging and Graphics, vol. 32, pp. 140-149.
Miao, X, Miao, Y, Wang, Q & Zhou, J 2005, 'A method of realizing the visual repository base on SVG standard', Academic Journal of Xi'an Jiaotong University, vol. 17, no. 1, pp. 21-24.
This paper demonstrates that the visual and graphical repository can be established through extending the SVG standard, introducing C# script and customizing component into repository. The graphical system, which uses the technique of Visual Studio. Net, supports the secondary object-oriented development. It allows the user to not only add all kinds of component objects, but also add properties and functions for the object. In addition, the graphical system has a very good property of customization. It supports the C# script and the zoom in/zoom out of pictures. The file formats used in this system are XML and SVG.
Miao, Y, Miao, X, Bian, Z & Zhou, J 2005, 'The recognition system of moving machine printed Mark/Numeral', Academic Journal of Xi'an Jiaotong University, vol. 17, no. 1, pp. 15-20.
This paper presents a recognition system for the automatic quality control in industrial applications. The purpose of the system is to collect the product information (e. g. expiry-date, production identification) and verify these information for quality control. The main difficulties of the system are to make an effcient preprocessing for the acquired low resolution image and to create a simple and fast recognition method to get the product information. In this paper, we propose an effcient recognition method based on the endpoint features and structure characteristics of the numerals. The experimental results show that the proposed method is effcient, robust and reliable for recognizing machine printed numerals. The system is currently successfully working with a real application with required specifications.
Zhou, J, Döring, A & Tönnies, KD 2005, 'Control of object visibility in volume rendering—a distance-based approach', International Journal of Image and Graphics, vol. 5, pp. 699-714.
Zhou, J & Tonnies, KD 2003, 'State of the art for volume rendering', Simulation, pp. 1-29.
Yu, K, Berkovsky, S, Conway, D, Taib, R, Zhou, J & Chen, F 2018, 'Do I trust a machine? Differences in user trust based on system performance' in Human and Machine Learning, Springer, Switzerland, pp. 245-264.View/Download from: UTS OPUS or Publisher's site
Trust plays an important role in various user-facing systems and applications. It is particularly important in the context of decision support systems, where the system's output serves as one of the inputs for the users' decision making processes. In this chapter, we study the dynamics of explicit and implicit user trust in a simulated automated quality monitoring system, as a function of the system accuracy. We establish that users correctly perceive the accuracy of the system and adjust their trust accordingly. The results also show notable differences between two groups of users and indicate a possible threshold in the acceptance of the system. This important learning can be leveraged by designers of practical systems for sustaining the desired level of user trust.
Zhou, J & Chen, F 2018, '2D Transparency Space—Bring Domain Users and Machine Learning Experts Together' in Human and Machine Learning, Springer, Germany, pp. 3-19.View/Download from: UTS OPUS or Publisher's site
Machine Learning (ML) is currently facing prolonged challenges with the user acceptance of delivered solutions as well as seeing system misuse, disuse, or even failure. These fundamental challenges can be attributed to the nature of the 'black-box' of ML methods for domain users when offering ML-based solutions. That is, transparency of ML is essential for domain users to trust and use ML confidently in their practices. This chapter argues for a change in how we view the relationship between human and machine learning to translate ML results into impact. We present a two-dimensional transparency space which integrates domain users and ML experts together to make ML transparent. We identify typical Transparent ML (TML) challenges and discuss key obstacles to TML, which aim to inspire active discussions of making ML transparent with a systematic view in this timely field.
Zhou, J, Yu, K & Chen, F 2018, 'Revealing User Confidence in Machine Learning-Based Decision Making' in Human and Machine Learning: Visible, Explainable, Trustworthy and Transparent, Springer, Switzerland, pp. 225-244.View/Download from: UTS OPUS or Publisher's site
Zhang, B, Guo, T, Zhang, L, Lin, P, Wang, Y, Zhou, J & Chen, F 2018, 'Water pipe failure prediction: A machine learning approach enhanced by domain knowledge' in Human and Machine Learning: Visible, Explainable, Trustworthy and Transparent, Springer, Switzerland, pp. 363-383.View/Download from: UTS OPUS or Publisher's site
Zhou, J, Yu, K, Chen, F, Wang, Y & Arshad, SZ 2018, 'Multimodal behavioral and physiological signals as indicators of cognitive load' in Oviatt, S, Schuller, B, Cohen, P, Sonntag, D, Potamianos, G & Krueger, A (eds), The Handbook of Multimodal-Multisensor Interfaces, Association for Computing Machinery and Morgan & Claypool, New York, NY, USA, pp. 287-329.View/Download from: UTS OPUS or Publisher's site
Chen, F, Zhou, J & Yu, K 2017, 'Multimodal and data-driven cognitive load measurement' in Cognitive Load Measurement and Application: A Theoretical Framework for Meaningful Research and Practice, pp. 147-178.View/Download from: Publisher's site
© 2018 Taylor & Francis. Cognitive load (CL), or mental workload, is an important issue in various application areas such as human-computer interaction (HCI), adaptive automation and training, traffic control, performance prediction, driving safety, and military command and control (Byrne & Parasuraman, 1996; Coyne, Baldwin, Cole, Sibley, & Roberts, 2009; Grootjen, Neerincx, Weert, & Truong, 2007). Many definitions exist for cognitive load, and one of the most widely accepted definitions is that it is a multidimensional construct representing the load imposed on the working memory during performance of a cognitive task (Paas & van Merriënboer, 1994; Paas, Tuovinen, Tabbers, & Van Gerven, 2003). The term working memory has been extensively used since its first appearance in the classic work on information processing capacity, which to some extent quantifies the capability of the human brain (Baddeley, 1986; Baddeley, Thomson, & Buchanan, 1975). The concept of working memory set the foundation for the cognitive load analysis thereafter and, to many researchers, cognitive load research is the examination of how and to what extent the working memory is deployed and utilized during a specific cognitive task.
Yu, K, Berkovsky, S, Taib, R, Zhou, J & Chen, F 2019, 'Do I trust my machine teammate?: an investigation from perception to decision', Proceedings of the 24th International Conference on Intelligent User Interfaces, ACM, pp. 460-468.View/Download from: UTS OPUS
Zhou, J, Yu, K, Chen, F, Wang, Y & Arshad, SZ 2018, 'Multimodal behavioral and physiological signals as indicators of cognitive load', The Handbook of Multimodal-Multisensor Interfaces, Association for Computing Machinery and Morgan & Claypool, pp. 287-329.
Cui, H, Wang, X, Zhou, J, Gong, G, Yin, Y & Feng, D 2017, 'COMPUTATIONAL BOUNDARY DEFINITON BY GEODESIC GRAPH MODEL', 2017 IEEE 14TH INTERNATIONAL SYMPOSIUM ON BIOMEDICAL IMAGING (ISBI 2017), IEEE 14th International Symposium on Biomedical Imaging (ISBI) - From Nano to Macro, IEEE, Melbourne, AUSTRALIA, pp. 1201-1204.
Zhou, J, Li, Z, Zhi, W, Liang, B, Moses, D & Dawes, L 2017, 'Using Convolutional Neural Networks and Transfer Learning for Bone Age Classification', DICTA 2017 - 2017 International Conference on Digital Image Computing: Techniques and Applications, pp. 1-8.View/Download from: Publisher's site
© 2017 IEEE. The bone age of a child indicates the skeletal and biological maturity of an individual. The most commonly applied clinical methods for Bone Age Assessment (BAA) are based on the visual examination of ossification of individual bones in radiographs of the left hand and the wrist by comparing with standard hand atlas. This kind of method is highly subjective and the performance extremely depends on practitioners' experiences. This paper investigates the use of Deep Convolutional Neural Networks (DCNNs) for the automatic bone age assessment. As there exists no large-scale annotated medical image dataset comparable to ImageNet for medical image analysis, this paper uses transfer learning within DCNNs to perform bone age classifications making full use of advantages of DCNNs. We define various Regions of Interest (ROIs) based on domain knowledge, for each of which a local bone age classification model is achieved by fine-tuning the pre-trained VGGNet with corresponding ROI patches. A final bone age classification is obtained by fusing multiple regional models. The results show that the proposed approach outperforms the current state-of-the-art classification methods in BAA with small dataset.
Luo, S, Chu, VW, Zhou, J, Chen, F, Wong, RK & Huang, W 2017, 'A Multivariate Clustering Approach for Infrastructure Failure Predictions', 2017 IEEE 6TH INTERNATIONAL CONGRESS ON BIG DATA (BIGDATA CONGRESS 2017), IEEE 6th International Congress on Big Data (BigData Congress), IEEE, Honolulu, HI, pp. 274-281.View/Download from: Publisher's site
Zhou, J, Arshad, SZ, Luo, S & Chen, F 2017, 'Effects of Uncertainty and Cognitive Load on User Trust in Predictive Decision Making', HUMAN-COMPUTER INTERACTION - INTERACT 2017, PT IV, 16th IFIP TC 13 International Conference on Human-Computer Interaction (INTERACT), SPRINGER INTERNATIONAL PUBLISHING AG, Indian Inst Technol, Mumbai, INDIA, pp. 23-39.View/Download from: Publisher's site
Luo, S, Duh, HBL, Zhou, J & Chen, F 2017, 'BVP signal feature analysis for intelligent user interface', Conference on Human Factors in Computing Systems - Proceedings, pp. 1861-1868.View/Download from: Publisher's site
Copyright © 2017 by the Association for Computing Machinery, Inc. (ACM). The Blood Volume Pulse (BVP) sensor has been becoming increasingly common in devices such as smart phones and smart watches. These devices often use BVP to monitor the heart rate of an individual. There has been a large amount of research linking the mental and emotional changes with the physiological changes. The BVP sensor measures one of these physiological changes known as Heart Rate Variability (HRV). HRV is known to be closely related to Respiratory Sinus Arrhythmia (RSA) which can be used as a measurement to quantify the activity of the parasympathetic activity. However, the BVP sensor is highly susceptible to noise and therefore BVP signals often contain a large number of artefacts which make it difficult to extract meaningful features from the BVP signals. This paper proposes a new algorithm to filter artefacts from BVP signals. The algorithm is comprised of two stages. The first stage is to detect the corrupt signal using a Short Term Fourier Transform (STFT). The second stage uses Lomb-Scargle Periodogram (LSP) to approximate the Power Spectral Density (PSD) of the BVP signal. The algorithm has shown to be effective in removing artefacts which disrupt the signal for a short period of time. This algorithm provides the capability for BVP signals to be analysed for frequency based features in HRV which traditionally could be done from the cleaner signals from electrocardiogram (ECG) in medical applications.
Chen, Z, Zhou, J, Wang, X, Swanson, J, Chen, F & Feng, D 2017, 'Neural Net-Based and Safety-Oriented Visual Analytics for Time-Spatial Data', 2017 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), International Joint Conference on Neural Networks (IJCNN), IEEE, Anchorage, AK, pp. 1133-1140.
Zhou, J, Arshad, SZ, Luo, S, Yu, K, Berkovsky, S & Chen, F 2017, 'Indexing cognitive load using blood volume pulse features', Conference on Human Factors in Computing Systems - Proceedings, pp. 2269-2275.View/Download from: Publisher's site
Copyright © 2017 by the Association for Computing Machinery, Inc. (ACM). Physiological responses contain rich affective information even when humans are not expressing any external signs. In this paper, we investigate the use of the Blood Volume Pulse (BVP) signals for indexing cognitive load. An experiment, which introduced cognitive load as a secondary task in a decision making context was conducted in the study. BVP signals were analyzed in order to establish relationships between BVP and cognitive load levels. A set of features (e.g. peak and max features) was found to be significantly distinctive across different cognitive load levels. The identified BVP features can be used to set up machine learning models for the automatic classification of CL levels in intelligent systems.
Yu, K, Conway, D, Berkovsky, S, Zhou, J, Taib, R & Chen, F 2017, 'User Trust Dynamics: An Investigation Driven by Differences in System Performance', IUI'17: PROCEEDINGS OF THE 22ND INTERNATIONAL CONFERENCE ON INTELLIGENT USER INTERFACES, 22nd International Conference on Intelligent User Interfaces (IUI), ASSOC COMPUTING MACHINERY, Limassol, CYPRUS, pp. 307-317.View/Download from: Publisher's site
Cui, H, Wang, X, Zhou, J, Gong, G, Yin, Y, Zheng, F & Feng, D 2016, 'LEARNING MULTI-MODALITY LOCAL AND GLOBAL AFFINITIES IN GRAPH BASED RANKING FOR AUTOMATED LUNG TUMOR DELINEATION', 2016 IEEE 13TH INTERNATIONAL SYMPOSIUM ON BIOMEDICAL IMAGING (ISBI), IEEE 13th International Symposium on Biomedical Imaging (ISBI), IEEE, Prague, CZECH REPUBLIC, pp. 948-951.View/Download from: Publisher's site
Zhou, J, Arshad, SZ, Yu, K & Chen, F 2016, 'Correlation for User Confidence in Predictive Decision Making', PROCEEDINGS OF THE 28TH AUSTRALIAN COMPUTER-HUMAN INTERACTION CONFERENCE (OZCHI 2016), 28th Australian Computer-Human Interaction Conference (OzCHI), ASSOC COMPUTING MACHINERY, Univ Tasmania, Hobart, AUSTRALIA.View/Download from: Publisher's site
Conway, D, Chen, F, Yu, K, Zhou, J & Morris, R 2016, 'Misplaced trust: A bias in human-machine trust attribution - In contradiction to learning theory', Conference on Human Factors in Computing Systems - Proceedings, pp. 3035-3041.View/Download from: Publisher's site
© 2016 Authors. Human-machine trust is a critical mitigating factor in many HCI instances. Lack of trust in a system can lead to system disuse whilst over-trust can lead to inappropriate use. Whilst human-machine trust has been examined extensively from within a technicosocial framework, few efforts have been made to link the dynamics of trust within a steady-state operatormachine environment to the existing literature of the psychology of learning. We set out to recreate a commonly reported learning phenomenon within a trust acquisition environment: Users learning which algorithms can and cannot be trusted to reduce traffic in a city. We failed to replicate (after repeated efforts) the learning phenomena of 'blocking', resulting in a finding that people consistently make a very specific error in trust assignment to cues in conditions of uncertainty. This error can be seen as a cognitive bias and has important implications for HCI.
Zhou, J, Li, Z, Zhang, Z, Liang, B & Chen, F 2016, 'Visual Analytics of Relations of Multi-Attributes in Big Infrastructure Data', 2016 INTERNATIONAL SYMPOSIUM ON BIG DATA VISUAL ANALYTICS (BDVA), International Symposium on Big Data Visual Analytics (BDVA), IEEE, Sydney, AUSTRALIA, pp. 31-32.
Zhou, J, Asif Khawaja, M, Li, Z, Sun, J, Wang, Y & Chen, F 2016, 'Making machine learning useable by revealing internal states update-a transparent approach', International Journal of Computational Science and Engineering, pp. 378-389.View/Download from: Publisher's site
© 2016 Inderscience Enterprises Ltd. Machine learning (ML) techniques are often found difficult to apply effectively in practice because of their complexities. Therefore, making ML useable is emerging as one of active research fields recently. Furthermore, an ML algorithm is still a 'black-box'. This 'black-box' approach makes it difficult for users to understand complicated ML models. As a result, the user is uncertain about the usefulness of ML results and this affects the effectiveness of ML methods. This paper focuses on making a 'black-box' ML process transparent by presenting real-time internal status update of the ML process to users explicitly. A user study was performed to investigate the impact of revealing internal status update to users on the easiness of understanding data analysis process, meaningfulness of real-time status update, and convincingness of ML results. The study showed that revealing of the internal states of ML process can help improve easiness of understanding the data analysis process, make real-time status update more meaningful, and make ML results more convincing.
Yu, K, Taib, R, Berkovsky, S, Zhou, J, Conway, D & Chen, F 2016, 'Trust and Reliance based on system accuracy', UMAP 2016 - Proceedings of the 2016 Conference on User Modeling Adaptation and Personalization, pp. 223-227.View/Download from: Publisher's site
© 2016 ACM. Trust plays an important role in various user-facing systems and applications. It is particularly important in the context of decision support systems, where the system's output serves as one of the inputs for the users' decision making processes. In this work, we study the dynamics of explicit and implicit user trust in a simulated automated quality monitoring system, as a function of the system accuracy. We establish that users correctly perceive the accuracy of the system and adjust their trust accordingly.
Zhou, J, Bridon, C, Chen, F, Khawaji, A & Wang, Y 2015, 'Be informed and be involved: Effects of uncertainty and correlation on user's confidence in decision making', Conference on Human Factors in Computing Systems - Proceedings, pp. 923-928.View/Download from: Publisher's site
User's confidence in machine learning (ML) based decision making significantly affects acceptability of ML techniques. In this work, we investigate how uncertainty/correlation affects user's confidence in order to design effective user interface for ML-based intelligent systems. A user study was performed and we found that revealing of correlation helped users better understand uncertainty and thus increased confidence in model output. When correlation had the same trend with performance, correlation but not uncertainty helped users more confident in their decisions.
Zhou, J, Jung, JY & Chen, F 2015, 'Dynamic Workload Adjustments in Human-Machine Systems Based on GSR Features', HUMAN-COMPUTER INTERACTION, PT I, 15th IFIP TC.13 International Conference on Human-Computer Interaction (INTERACT), SPRINGER-VERLAG BERLIN, Bamberg, GERMANY, pp. 550-558.View/Download from: Publisher's site
Zhou, J, Sun, J, Chen, F, Wang, X & Miao, X 2015, 'Safety-Oriented Visual Analytics of People Movement', 2015 IEEE CONFERENCE ON VISUAL ANALYTICS SCIENCE AND TECHNOLOGY, 10th IEEE Conference on Visual Analytics Science and Technology (VAST), IEEE, Chicago, IL, pp. 181-182.
Oviatt, S, Hang, K, Zhou, J & Chen, F 2015, 'Spoken Interruptions Signal Productive Problem Solving and Domain Expertise in Mathematics', ICMI'15: PROCEEDINGS OF THE 2015 ACM INTERNATIONAL CONFERENCE ON MULTIMODAL INTERACTION, 2015 ACM International Conference on Multimodal Interaction, ASSOC COMPUTING MACHINERY, Seattle, WA, pp. 311-318.View/Download from: Publisher's site
Chen, F, Marcus, N, Khawaji, A & Zhou, J 2015, 'Using galvanic skin response (GSR) to measure trust and cognitive load in the text-chat environment', Conference on Human Factors in Computing Systems - Proceedings, pp. 1989-1994.View/Download from: Publisher's site
Exchanging text messages via software on smart phones and computers has recently become one of the most popular ways for people to communicate and accomplish their tasks. However, there are negative aspects to using this kind of software, for example, it has been found that people communicating in the text-chat environment may experience a lack of trust and may face different levels of cognitive load [1, 11]. This study examines a novel way to measure interpersonal trust and cognitive load when they overlap with each other in the text-chat environment. We used Galvanic Skin Response (GSR), a physiological measurement, to collect data from twenty-eight subjects at four gradients and overlapping conditions between trust and cognitive load. The findings show that the GSR signals were significantly affected by both trust and cognitive load and provide promising evidence that GSR can be used as a tool for measuring interpersonal trust when cognitive load is low and also for measuring cognitive load when trust is high.
Arshad, SZ, Zhou, J, Bridon, C, Chen, F & Wang, Y 2015, 'Investigating user confidence for uncertainty presentation in predictive decision making', OzCHI 2015: Being Human - Conference Proceedings, pp. 352-360.View/Download from: Publisher's site
Copyright © 2015 ACM. Machine Learning (ML) based decision support systems are often like a black box to non-expert users. Here user's confidence becomes critical for effective decision making and maintaining trust in the system. We find that user confidence varies significantly depending on supplementary material presented on screen. We investigate change in user confidence (in the context of ML based decision making) by varying level of uncertainty presented (in an online water-pipe failure prediction case study) and find that all 26 subjects rated higher uncertainty task to be most difficult and had lowest user confidence in predictive decisions of the same. This agrees with our expectation that increased uncertainty would reduce user confidence in predictive decision making. However, ML-researchers subgroup reported being most confident when uncertainty with known probability was presented, whereas other subgroups (viz. general staff and non-ML researchers) appeared most confident when uncertainty was not at all presented. This is an original research to improve understanding of user's decision making confidence with respect to uncertainty presented in machine learning context.
Cui, H, Wang, X, Zhou, J, Fulham, M, Eberl, S & Feng, D 2014, 'TOPOLOGY CONSTRAINT GRAPH-BASED MODEL FOR NON-SMALL-CELL LUNG TUMOR SEGMENTATION FROM PET VOLUMES', 2014 IEEE 11th International Symposium on Biomedical Imaging (ISBI), 11th IEEE International Symposium on Biomedical Imaging (ISBI), IEEE, Beijing, PEOPLES R CHINA, pp. 1243-1246.
Zhou, J, Wang, X & Feng, D 2014, 'Importance-Aware Lighting Design in Volume Visualization', 2014 13TH INTERNATIONAL CONFERENCE ON CONTROL AUTOMATION ROBOTICS & VISION (ICARCV), 13th International Conference on Control Automation Robotics & Vision (ICARCV), IEEE, Singapore, SINGAPORE, pp. 839-843.
Khawaji, A, Chen, F, Zhou, J & Marcus, N 2014, 'Trust and cognitive load in the text-chat environment: The role of mouse movement', Proceedings of the 26th Australian Computer-Human Interaction Conference, OzCHI 2014, pp. 324-327.
Copyright 2014 ACM. This paper examines how different levels of cognitive load can affect trust in the text-chat environment. It also examines how the mouse movements of participants can indicate the level of cognitive load when they chat with each other. We designed two chat systems: one in which subjects chat under low mental load and the other in which subjects chat under high mental load. Twenty subjects participated in the study and the results showed significant differences in the level of trust between subjects under different cognitive loads; that is, subjects who chatted under low mental load showed more trust in their partners. Moreover, the mouse data obtained proved to be effective in indicating the level of cognitive load existing between the subjects. However, this work suggests that to establish trust in the chat environment, it is better to communicate under a low cognitive load. Our findings also show the ability of designed systems to measure cognitive load via tracking mouse events for the purpose of providing assistance to communicators.
Zhou, J, Hang, K, Oviatt, S, Yu, K & Chen, F 2014, 'Combining empirical and machine learning techniques to predict math expertise using pen signal features', MLA 2014 - Proceedings of the 2014 ACM Multimodal Learning Analytics Workshop and Grand Challenge, Co-located with ICMI 2014, pp. 29-36.View/Download from: Publisher's site
© 2014 ACM. Multimodal learning analytics aims to automatically analyze students' natural communication patterns based on speech, writing, and other modalities during learning activities. This research used the Math Data Corpus, which contains timesynchronized multimodal data from collaborating students as they jointly solved problems varying in difficulty. The aim was to investigate how reliably pen signal features, which were extracted as students wrote with digital pens and paper, could identify which student in a group was the dominant domain expert. An additional aim was to improve prediction of expertise based on joint bootstrapping of empirical science and machine learning techniques. To accomplish this, empirical analyses first identified which data partitioning and pen signal features were most reliably associated with expertise. Then alternative machine learning techniques compared classification accuracies based on all pen features, versus empirically selected ones. The best unguided classification accuracy was 70.8%, which improved to 83.3% with empirical guidance. These results demonstrate that handwriting signal features can predict domain expertise in math with high reliability. Hybrid methods also can outperform blackbox machine learning in both accuracy and transparency.
Zhou, J, Li, Z, Wang, Y & Chen, F 2013, 'Transparent Machine Learning—Revealing Internal States of Machine Learning', Proceedings of International Conference on Intelligent User Interfaces 2013 Workshop on Interactive Machine Learning.
Khawaji, A, Chen, F, Marcus, N & Zhou, J 2013, 'Trust and cooperation in text-based computer-mediated communication', Proceedings of the 25th Australian Computer-Human Interaction Conference: Augmentation, Application, Innovation, Collaboration, OzCHI 2013, pp. 37-40.View/Download from: Publisher's site
This study examines how different behaviours can affect trust in the text-chat environment. We designed two automated chat systems: one behaves cooperatively and the other behaves competitively. Thirty subjects participated in this study and the results revealed that the trust of subjects who chatted with a cooperative partner was significantly higher than the trust of subjects who chatted with a competitive partner. This study also examines the chat contents and the results show that subjects behave differently when they trust their partner, using more assent and positive emotion words.
Zhou, J & Xiao, C 2012, 'Similarity-driven multi-level partial contour tree simplification', Proceedings - 2012 9th International Conference on Fuzzy Systems and Knowledge Discovery, FSKD 2012, pp. 1949-1953.View/Download from: Publisher's site
A contour tree is a topological abstraction of a scalar field. Contour tree simplification (CTS) removes branches corresponding to noise, while making the size of the tree small enough for maintaining essential structure of data. This paper proposes a similarity-driven multi-level partial CTS (PCTS) approach. The PCTS preserves branches corresponding to structures of interest or specific objects specified by users, while removing other branches of the contour tree. A clustering method (e.g. k-means clustering) is used to cluster branch nodes into groups based on their similarities (i.e., similar locations) in the attribute space. As a result, the contour tree is simplified with multi-levels based on different clustering groups. Furthermore, various interfaces and rendering windows are provided and synchronized, which makes the simplification process more meaningful and efficient compared with traditional simplification methods using slide-bar based approaches. The proposed approach can be generalized to process branches with more than three measures. © 2012 IEEE.
Zhou, J & Takatsuka, M 2011, 'Importance driven contour tree simplification', Proceedings - 2011 International Conference on Internet Computing and Information Services, ICICIS 2011, pp. 265-268.View/Download from: Publisher's site
Real-world data sets produce uumanageably large contour trees because of uoise and artifacts. It makes the contour tree impractical in data analysis and visualization. This paper proposes an importance-driven contour tree simplification approach which combines different measures of importance through an importance triangle to maximize advantages of each measure of importance. Extended Gaussian image, map projection, and K-Means clustering are used to manipulate importance measure vectors, which makes the simplification more meaningful and efficient. The proposed approach can be generalized to process branches with more than three measures. © 2011 IEEE.
Zhou, J, Lee, I, Thomas, B, Menassa, R, Farrant, A & Sansome, A 2011, 'Applying Spatial Augmented Reality to facilitate in-situ support for automotive spot welding inspection', Proceedings of VRCAI 2011: ACM SIGGRAPH Conference on Virtual-Reality Continuum and its Applications to Industry, pp. 195-200.View/Download from: Publisher's site
In automotive manufacturing, the quality of spot welding on car bodies needs to be inspected frequently. Operators often only check different subsets of spots on different car bodies with a predetermined sequence. Currently, spot welding inspections rely on a printed drawing of the testing body, with the inspection points marked on this drawing. Operators have to locate the matching spot on the drawing and the body manually to perform the inspection. The manual inspection process suffers from inefficiencies and potential mistakes. This paper describes a system that projects visual data onto arbitrary surfaces for providing just-in-time information to a user in-situ within a physical work-cell. Spatial Augmented Reality (SAR) is the key technology utilized in our system. SAR facilitates presentation of projected digital Augmented Reality (AR) information on surfaces of car bodies. Four types of digital AR information are projected onto the surfaces of car body parts in structured work environments: 1) Location of spot welds; 2) Inspection methods; 3) Operation Description Sheet (ODS) information; 4) Visualization of weld locating methods. Various visualization methods are used to indicate the position of spot welds and the method used for spot welding inspection. Dynamical visualizations are used to assist operators to locate spot welds more easily. The SAR approach does not require additional special models in finding spot welds, but only needs knowledge of location of spot welds on the part. Our system allows operators becoming more effective and efficient to in performing proper inspections, by providing them the required information at the required time without the need to refer to paper-based manuals or computer terminals. © 2011 ACM.
Zhou, J & Takatsuka, M 2009, 'Automatic Transfer Function Generation Using Contour Tree Controlled Residue Flow Model and Color Harmonics', IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS, IEEE Information Visualization Conference/IEEE Visualization Conference, IEEE COMPUTER SOC, Atlantic City, NJ, pp. 1481-1488.View/Download from: Publisher's site
Zhou, J & Takatsuka, M 2009, 'Structural Relationship Preservation in Volume Rendering', COMPUTER AND INFORMATION SCIENCE 2009, 8th IEEE/ACIS International Conference on Computer and Information Science, SPRINGER, Shanghai, PEOPLES R CHINA, pp. 229-238.
Zhou, J, Wang, Z & Xiao, C 2007, 'Perceptive factors for volume visualization in medical image analysis', 2007 IEEE/ICME INTERNATIONAL CONFERENCE ON COMPLEX MEDICAL ENGINEERING, VOLS 1-4, IEEE/ICME International Conference on Complex Medical Engineering, IEEE, Beijing, PEOPLES R CHINA, pp. 579-585.View/Download from: Publisher's site
Zhou, J & Wang, Z 2006, 'Focal region-based volume rendering', INTERNATIONAL JOURNAL OF PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE, WORLD SCIENTIFIC PUBL CO PTE LTD, pp. 665-677.View/Download from: Publisher's site
Zhou, J 2005, 'The roles of perception for volume visualization and designing volume visualization methods based on perceptual factors', Proceedings - APGV 2005: 2nd Symposium on Applied Perception in Graphics and Visualization, p. 177.
Zhou, J, Döring, A & Tönnies, KD 2004, 'Distance Based Enhancement for Focal Region Based Volume Rendering', Bildverarbeitung für die Medizin 2004, Algorithmen - Systeme - Anwendungen, Proceedings des Workshops vom 29. bis 30. März 2003 in Berlin, pp. 199-203.
Zhou, J, Döring, A & Tönnies, KD 2004, 'Distance transfer function based rendering', Technical Report, University of Magdeburg, Germany.
Zhou, J, Hinz, M & Tönnies, KD 2002, 'Hybrid Focal Region-Based Volume Rendering of', Bildverarbeitung für die Medizin 2002: Algorithmen—Systeme—Anwendungen Proceedings des Workshops vom 10.–12. März 2002 in Leipzig, Springer-Verlag, pp. 113-113.
Zhou, JL, Hinz, M & Tonnies, KD 2002, 'Focal region-guided feature-based volume rendering', FIRST INTERNATIONAL SYMPOSIUM ON 3D DATA PROCESSING VISUALIZATION AND TRANSMISSION, 1st International Symposium on 3D Data Processing Visualization and Transmission, IEEE COMPUTER SOC, PADUA, ITALY, pp. 87-90.View/Download from: Publisher's site
Zhou, J & Takatsuka, M 2008, 'Contour tree simplification based on a combined approach', School of Information Technologies, University of Sydney.
Zhou, J & Toennies, K 2002, 'Investigation of Volume Rendering Algorithms'.