Project Number: NML 1 Project Title: Dynamic Workload Adjustment in Human- Machine Systems Name of Supervisor: Jianlong Zhou (Jianlong.zhou@nicta.com.au) Name of Co- Supervisor: Dr. Fang Chen (Fang.Chen@nicta.com.au) Workload (also refers as cognitive load) is found to be a critical factor driving human behaviour in human- machine interactions (HMI) with automation systems in modern complex high- risk domains such as aviation, and the military command and control. Dynamic workload adjustments can be used in human- machine systems in order to improve user s engagement and performance. The objectives of this project include: 1) Investigate critical factors affecting real- time workload classifications, which include automatic feature selection, as well as the development of effective ML classifiers for real- time workload classifications. 2) Develop dynamic workload adaptation model to control how workload can be adjusted in the pipeline. This is the core of dynamic workload adjustment in this study. 3) Explore factors that affect the generalization of workload measurement and propose effective approaches for reliable workload measurement in complex environments with variations of humans in different aspects. 4) carry out a small scale user experiment. Approaches used in this project include user analysis, information visualization techniques, as well as ML for sensor data analysis. Techniques developed in this project will facilitate dynamic workload adaptation in a feedback loop. Students involved in this project will gain comprehensive knowledge on HCI techniques, computing skills as well as interesting Android- based developments. The topic is available to 1 or 2 students with strong interest in HCI and visualization. The selected student will work with a team of researchers in the field of machine learning and human- computer interaction, at the NICTA headquarters at the Australian Technology Park. The NICTA team is a world research leader in machine learning, and cognitive load measurement through physiological and behavioural features. The student should implement a framework for experiment and data analysis with high level languages (e.g. Matlab, Python, C#). A research report will be expected at the end of the project. 1. Making Machine Learning Transparent http://www.nicta.com.au/category/research/machine- learning/projects/making- machine- learning- transparent/
Project Number: NML 2 Project Title: Visual Analytics of Time Series Data Name of Supervisor: Jianlong Zhou (Jianlong.zhou@nicta.com.au ) Name of Co- Supervisor: Dr. Fang Chen (Fang.Chen@nicta.com.au) With the advances of technologies in various areas, large amount of data are increasingly recorded. Besides the management of large amount data, the analyses and getting insight of large amount data are active research topics in recent years. Two approaches are commonly used in data analysis: machine learning (ML) and visualization. ML tries to get insights from data while visualization aims to present original data as well as analysis results meaningfully to end users. This research project focuses on the visualization of time series data. The objective of this project include: 1) investigate approaches that are used to present our specific time series data; 2) investigate approaches for presenting ML analysis results from our specific time series data; 3) implement the proposed mechanism; 4) carry out a small scale user experiment. The ultimate goal of the project is to set up a framework of visual analytics for time series data, which presents analysis results from other approaches (e.g. ML) with the original data meaningfully. Approaches used in this project include information visualization, human computer interaction (HCI) techniques, as well as effective presentation methods for ML results. Techniques developed in this project will facilitate the effective presentation of data sets with ML results, which helps users understand data sets as well as analysis results easily. Students involved in this project will gain comprehensive knowledge on visualization, HCI techniques, computing skills as well as interesting domain knowledge. The topic is available to 1 or 2 students with strong interest in HCI and information visualization. The selected student will work with a team of researchers in the field of machine learning and human- computer interaction, at the NICTA headquarters at the Australian Technology Park. The NICTA team is a world research leader in machine learning, and cognitive load measurement through physiological and behavioural features. The student should implement a framework for information visualization with some high level languages (e.g. JavaScript, matlab, python, or others). A research report will be expected at the end of the project. 1. Making Machine Learning Transparent http://www.nicta.com.au/category/research/machine- learning/projects/making- machine- learning- transparent/
Project Number: NML 3 Project Title: Uncertainty Visual Analytics in Machine Learning Name of Supervisor: Jianlong Zhou (Jianlong.zhou@nicta.com.au ) Name of Co- Supervisor: Dr. Fang Chen (Jinjun.Sun@nicta.com.au, Fang.Chen@nicta.com.au) As machine learning (ML) techniques become more widely used in various fields, adequate methods have to be provided to allow users to employ ML techniques effectively. However, data is inherently uncertain and often incomplete and contradictory. Meanwhile, transformations with ML techniques also introduce uncertainty into data analysis pipeline. Therefore, a visual quantitative analysis of uncertainty in ML based data analysis process is helpful for users to make more informative decisions. This research project focuses on the visual analysis of uncertainty in an ML based data analysis process. The objective of this project include: 1) uncertainty modeling; 2) design a mechanism to visualize uncertainty at each stage of ML process; 3) implement the proposed mechanism; 4) carry out a small scale user experiment. Approaches used in this project include data modeling, human computer interaction (HCI) techniques, as well as effective presentation methods for ML results. Techniques developed in this project will facilitate the decision making effectiveness in ML and improve impact of ML in real- world applications. Students involved in this project will gain comprehensive knowledge on uncertainty, HCI techniques, computing skills as well as various interesting machine learning techniques. The topic is available to 1 or 2 students with strong interest in HCI and ML. The selected student will work with a team of researchers in the field of machine learning and human- computer interaction, at the NICTA headquarters at the Australian Technology Park. The NICTA team is a world research leader in machine learning, and cognitive load measurement through physiological and behavioural features. The student should implement a framework for uncertainty visual analytics in ML with some high level languages (e.g. Matlab, Python). A research report will be expected at the end of the project. 1. Making Machine Learning Transparent http://nicta.com.au/research/machine_learning/projects
Project Number: NML 4 Project Title: Data analytics for intelligent water networks Name of Supervisor: Ronnie Taib (ronnie.taib@nicta.com.au) Name of Co- Supervisor: Dr. Fang Chen (Fang.Chen@nicta.com.au) The supply of clean and safe drinking water is paramount to human development, and comes with many issues ranging from long- term supply planning, to infrastructure construction, chemical processing and demand management. NICTA has been working with Sydney Water and other water utilities in Australia and around the world to help improve water supply. This research project focuses on optimising water distribution through the analysis of network topology, demand patterns and energy costs. The main objective is to use machine learning methods to predict water needs, chemical dosing, and plan distribution accordingly. The student will analyse real- life data, identify specific geographical areas of interest where correlations between the parameters can be established, and then generalised. The student will also contribute to the software implementation of a simple geographic information system (GIS) allowing to visualise the results of applying various distribution options. The selected student will work with a team of researchers in the field of machine learning at the NICTA headquarters at the Australian Technology Park. The NICTA team is a world research leader in machine learning and is bringing efficiency improvements to the water industry through its partnerships with worldwide utilities. The student will summarise the data analysis in a report, and implement a GIS system to visualise the results. A presentation to the research group will be expected at the end of the project. 1. Advanced Data Analytics for Water Solutions http://www.nicta.com.au/category/research/machine- learning/projects/advanced- data- analytics- for- water- solutions/
Project Number: NML 5 Project Title: Trust investigations in Human- Machine Interaction Name of Supervisor: Kun Yu (kun.yu@nicta.com.au ) Name of Co- Supervisor: Dr. Fang Chen (Fang.Chen@nicta.com.au ) Trust has been attracting increasing research efforts since it is considered as an impacting factor in human machine interactions. It is a mental construct that relates to the preference, experience, cognitive status of human, and the competence, reliability of the machine. Understanding the role of trust in human machine interaction provides the potential to maximize the efficiency of human- machine cooperation, minimize the human efforts, and reduce the operational risks. This research project explores the relationship between trust and human decision making, and seeks answers to the following questions: - To what extent human decision is affected by the trust level on a machine? - What factors may affect trust, or distrust, or their conversion? - How can trust be effectively manipulated in an experimental environment? Experiments involving questionnaires, data collection and signal analysis will be the major part of this project. Via this process students involved in this project are capable of - Getting involved in an interesting research topic; - Viewing human machine interaction from another dimension trust; - Gaining a better understanding of the trustworthiness of a machine; - Knowing the impact when a machine has failed us. The topic is available to 1 or 2 students with strong interest in HCI, machine learning and psychological studies. The selected student will work with a team of researchers in the field of machine learning and human- computer interaction at NICTA headquarters in the Australian Technology Park. The NICTA team is a world research leader in machine learning, and cognitive load measurement through physiological and behavioural features. The student is expected to conduct user studies and corresponding data analysis. Programming skills with Matlab/Python are preferred. A research report/presentation will be required at the end of the project.
Project Number: NML 6 Project Title: Big data analysis for understanding customer behaviour Name of Supervisor: Dr. Bang Zhang (bang.zhang@nicta.com.au) Name of Co- Supervisor: Dr. Yang Wang and Prof. Fang Chen (yang.wang@nicta.com.au fang.chen@nicta.com.au) In this project, you will get hands- on experience on real- world customer consumption data. The aim of the project is to build a system to understand customer behaviour, e.g., customer segmentation, customer demand analysis, abnormally detection. The project consists of many components including data analysis, visualization and human computer interfacing. Successful candidates can select one of the components to make contributions. Your work will contribute to the system that makes real impacts on millions of residents in Sydney. The selected student will work with a team of researchers in the field of machine learning and data mining, at the NICTA headquarter - - Australian Technology Park. This project will design and develop algorithms that predict customer demand. This project will provide you an opportunity to learn about machine learning and data mining. We will develop algorithms to discover the critical factors that influence customer s behaviour.
Project Number: NML 7 Project Title: Trend Prediction on Time Series Name of Supervisor: Zhidong LI (ZIDONG.LI@nicta.com.au ) Many applications can be described by time series, such as stock market, customer demand, and change of economic indicators. Traditionally, the trend is estimated by the data from near history. However, the trend will eventually change even it can be closely fitted. The change can cause huge deviation on predictions. Even if the deviation can be corrected in short time, the loss cannot be neglected. The objective of this project includes: 1) Investigate the traditional methods of trend estimation, which provides the baseline work and the solution of in- trend components for trend prediction. 2) Collect and analyse factors that can cause trend volatility, which can include factor selection and visualization. 3) Find a suitable way to model the out of trend changes. Approaches including machine learning and data mining methods will be used in this project. The topic is available to 1 or 2 students with computer science background and analytical skills. The selected student will work with a team of researchers in the field of data science and machine learning, at the NICTA headquarters at the Australian Technology Park. The NICTA team is a world research leader in machine learning. The student can obtain some experience about data science. At the end of the project, the deliverable results include an implementation of a framework to show the evaluated results of trend prediction, together with discussions about how the work had been done and a research report. Project Number: NML 8 Project Title: Structural Health Monitoring Using Tensor Analysis Name of Supervisor: Dr. Khoa Nguyen (khoa.nguyen@nicta.com.au ) Structural health monitoring (SHM) is a technology to monitor civil infrastructure such as bridges and buildings using sensors. Data obtained from sensors can be used for damage detection in structures. NICTA and Roads and Maritime Services (RMS) have conducted a SHM project for the Sydney Harbour Bridge to monitor hundreds of joints underneath the bus lane. In SHM, the data are usually highly redundant and correlated. The measured variables are not only correlated with each other at a certain time but also are autocorrelated themselves over time. Matrix- based two- way analysis, which is usually used in SHM, cannot capture all these relationships and correlations together. Tensor analysis allows us to analyse the vibration data in more than two dimensions at the same time.
Students will work with real data obtained from several sensors installed on the bridge, investigate a method for damage identification using incremental tensor analysis and implement it using a programming language. The topic is available to a student with strong interest machine learning and data mining, a solid statistical background, and a strong programming skill. The selected student will work with a team of researchers in the field of machine learning, at the NICTA headquarter - - the Australian Technology Park. The NICTA team is a world research leader in machine learning. This project will investigate a method on tensor learning for SHM. The student will try the method with data from the bridge and other data as well. The student should exhibit autonomy and must have good programming skills in some high level language (e.g. Matlab, Java, C++). A research report will be expected at the end of the project. SHM project: https://www.nicta.com.au/category/industry- engagement/infrastructure- transport- and- logistics/projects/structural- health- monitoring/ Project Number: NML 9 Project Title: Feature Extraction for Structural Health Monitoring Name of Supervisor: Dr. Khoa Nguyen (khoa.nguyen@nicta.com.au ) Structural health monitoring (SHM) is a technology to monitor civil infrastructure such as bridges and buildings using sensors. Data obtained from sensors can be used for damage detection in structures. NICTA and Roads and Maritime Services (RMS) have conducted a SHM project for the Sydney Harbour Bridge to monitor hundreds of joints underneath the bus lane. For damage detection, feature extraction from raw time series data is a very important step. The goal is to extract features which are sensitive to damage, but not to environmental and operational conditions. Students will work with real data obtained from several sensors installed on the bridge, investigate a method for to extract features for damage detection and implement it using a programming language. The topic is available to a student with strong interest machine learning and data mining, a solid statistical background, and a strong programming skill. The selected student will work with a team of researchers in the field of machine learning, at the NICTA headquarter - - the Australian Technology Park. The NICTA team is a world research leader in machine learning. This project will investigate different common features for damage detection and apply them with the bridge data and other data as well.
The student should exhibit autonomy and must have good programming skills in some high level language (e.g. Matlab, Java, C++). A research report will be expected at the end of the project. SHM project: https://www.nicta.com.au/category/industry- engagement/infrastructure- transport- and- logistics/projects/structural- health- monitoring/ Project Number: NML 10 Project Title: Large Inference for Stochastic Random Partition Name of Supervisor: Dr. Xuhui Fan (Xuhui.Fan@nicta.com.au) Name of Co- Supervisor: Dr. Bin Li and Dr. Yang Wang (Bin.Li@nicta.com.au yang.wang@nicta.com.au) In this project, you will get hands- on experience on the Stochastic Random Partition problems, which belongs to the Nonparametric- Bayesian literature. The aim of the project is to build a systematic understanding of the inference methods in the stochastic Random partition problems, including the MCMC method and variational inference. The project consists of many components including data analysis, problem definition and machine learning. Successful candidates can select one of the components to make contributions. Your work will attends the big families of the Bayesian machine learning communities. The selected student will work with a team of researchers in the field of machine learning and data mining, at the NICTA headquarter - - Australian Technology Park. This project will design and develop new inference algorithms for the Stochastic Random Partition problems. This project will provide you an opportunity to learn about machine learning and data mining. We will develop algorithms which enables the large- scale inference and learning.