PhD projects

Human-in-The-Loop Analytics PhD projects

The Human-in-the-Loop Analytics (HiLA) program will provide up to 20 PhD scholarships. Each scholarship project will be listed here as they become available. These projects will continue to be defined over the course of the next 3 months (Q1 2019) and as they appear, applicants will be able to apply for the opportunities directly by selecting the "Request to Apply Button". This will take you directly to our request to apply form. Please make sure you select the topic you are interested in. You are also encouraged to apply for as many topics that you feel suited towards however you will need to fill out the form again for each topic.

Request to apply

Industry partner PhD projectSupervisorsApplications

Woodside Project 1 - Pipe Routing project

Developing efficient and (near) optimal algorithms for multiple 3D path routings is a significant challenge with a wide variety of applications: from routing the pipes in a chemical plant or a jet engine, to determining the paths of aerial drones. While significant research has been done in the 2D setting, there are still significant challenges whenever the paths are required to satisfy complex constraints. The 3D setting poses even more difficulties and has therefore not been studied. The aim of this PhD is to develop efficient and (near) optimal algorithms for multiple 3D path routings and integrate them into a visualisation system that allows users to query and modify the solutions proposed by the algorithm.

Prof Maria Garcia de la Banda

OPEN

Woodside Project 2 - Explainable optimisation project

A collaboration between Woodside Energy and Monash University has resulted in a system able to provide high quality plant layouts (including equipment positioning and pipe routing) for large chemical plants. This system works by optimising an objective function based on a calculated costs for the land footprint, piping and structural steel in the computed layout, while optimising a set of constraints based on safety, access and maintenance requirements for the plant. The optimisation is like a black box — the details of the equipment, connections, and other requirements are fed in, and solutions come out.  The relationship between the results of the optimisation and what was fed in, can sometimes be difficult for an Engineer to understand.

The aim of this PhD is to significantly improve the explainability of the plant layout optimisation process. This may include interrogation of solutions and their component costs, comparison of different solutions, detection of and recovery from infeasibility introduced by the user, and re-optimisation based on modified inputs suggested by the user.

Prof Maria Garcia de la Banda

OPEN

VIFM Project 1 - Immersive Visualisation of Medical Images

Medical imaging technology captures slices of the 3D internal structure of the human body (e.g. bones, tissues, organs) and creates 3D digital images (e.g. CT, MRI scans). Currently, doctors, medical and legal practitioners visualise these 3D data on 2D computer screens and explore the 3D slices in order to investigate a disease or an injury. Today, Augmented and Mixed Reality (AR/MR) technology allows us to visualise and interact with 3D data in an immersive way; for example, using a mixed-reality head-mounted display such as the Microsoft Hololens, a user can interact with 3D stereoscopic virtual graphics as if they were anchored in the 3D environment. While providing a more natural visualisation space for this type of data, interacting with a 3D scan in augmented reality remains challenging. For example, how can we browse the slices of a 3D scan? How do we extract useful information? How do we zoom into certain regions of the body? Promising ways to better interact with immersive 3D visualisations include:

  • Spatial interaction: for example using gestures such as pointing of swiping to select regions of interests in a 3D visualisation
  • Tangible interaction: for example using physical devices to support fine grained selections of CT scan slices

Assoc Prof Tim Dwyer and Dr Maxime Cordeil from the Faculty of Information Technology and Dr Matthew Dimmock from Faculty of Medicine, Nursing and Health Sciences.

CLOSED

VIFM Project 2 - Searching for an answer in incomplete forensic image-space

Forensic CT scans are acquired at the time of admission of a deceased person to the institute at which their autopsy will occur. This triage scan acts as important medico-legal evidence and a point of reference for the pathologist during the autopsy. This research project will investigate how to use state-of-the-art machine learning techniques to audit and search this database to match image queries resulting from pathologists, anthropologists and medico-legal professionals that require evidence to present in court. A variety of pipelines will be required to account for differences in the contents of the body bags due to variability in cause of death and the time to which the decedent is discovered by emergency services.

Dr David Albrecht from the Faculty of Information Technology and Dr Matthew Dimmock from Faculty of Medicine, Nursing and Health Sciences.

OPEN

Automated quality analysis of telecommunications audio

This project will explore the application of machine learning approaches with pre-labelled audio files, unlabelled audio and data-augmentation techniques to quickly and accurately assess the quality of audio and classify distortion characteristics. The candidate will research novel machine learning approaches with a methodology based on iterative experimentation through model development, implementation in software and testing with automated tools and user feedback.

Mean opinion score (MOS) is a measure used in telecommunications to measure the quality of audio (or video) in a system on a 5-point rating scale where 1 is bad (unintelligible) and 5 is excellent. These ratings are usually gathered in a subjective quality evaluation test or the use of algorithms
such as PSQM, PESQ, and POLQA. Enhancing this process through efficient machine learning algorithms has direct applicability in telecommunications and poses interesting research questions around automated data labelling, time-series classification and signal processing.

This project will also look at Identifying characteristic distortions that contribute to reduced MOS scores such as gaps and muffled audio caused by transcoders and accurately labelling these distortions temporally in audio streams.

As a Graduate Research Industry Partnership project, the candidate will have access to knowledge and resources from Monash University and Cyara, including state-of-the-art machine learning hardware (NVIDIA DGX-1) and access to large scale proprietary datasets. It also provides the opportunity for research to be tested and implemented in solving real-world problems during the project cycle.

This project would suit a candidate with a background in machine learning, especially in the areas of audio analysis and digital signal processing. Experience with deep learning frameworks such as PyTorch, Tensorflow or Caffe is desirable, but not required.

Prof Jon McCormack CLOSED

Audio Segmentation by Spoken Language

This is a Graduate Research Industry Partnership project run by Monash and Cyara that looks at the classification and segmentation of audio streams by spoken language content. Natural Language Recognition is a very active research area, with applications in telecommunications, robotics and portable computing.

An application of audio segmentation by spoken language will be explored with Cyara's Interactive Voice Response (IVR) systems that test speech interfaces, which can feature a mixture of languages and require fast classification for interactive responsiveness. For example, an American IVR might present a menu that says "For service in English, press 1. Par español, oprima 2", with the 2nd spoken phrase being in Spanish. The ability to segment and label audio from these systems by time offset, duration, language and confidence would assist automated testing and system improvement.

Machine learning techniques utilised in this area include support artificial neural networks, vector machines, multi-class logistic regression and Probabilistic Linear Discriminant Analysis and the candidate will be supported and encouraged to explore a range of techniques with an emphasis on responsiveness and accuracy for reduced language set classification ie. classifying the spoken language from a small pool of known language candidates.

As a Graduate Research Industry Partnership project, the candidate will have access to knowledge and resources from Monash University and Cyara, including state-of-the-art machine learning hardware (NVIDIA DGX-1) and access to large scale proprietary datasets. It also provides the opportunity for research to be tested and implemented in solving real-world problems during the project cycle.

This project would suit a candidate with a background in machine learning, especially in the area of speech processing. Experience with deep learning frameworks such as PyTorch, Tensorflow or Caffe is desirable, but not required.

Dr Patrick Hutchings CLOSED

Identification of Point of Sale Insights by Mining Social Media

This research project will explore machine learning techniques for mining social media data (text, photo and video) along with considering Patties Foods demographics and product data. Using machine learning techniques the PhD candidate will search for insights into questions such as "What makes food popular? What are the preferences of different demographics? Which elements most successfully drive sales?"

The intent is to find patterns in the data that may generate new insights into consumer preferences as applied to Patties product categories and Patties demographics.

As a Graduate Research Industry Partnership project, the candidate will have access to knowledge and resources from Monash University and our industry partner Patties Foods. The applicant will also have access large scale proprietary datasets. It also provides the opportunity for research to be tested and implemented in solving real-world problems during the project cycle.

Dr Xiaojun Chang CLOSED

Machine learning to support the civil construction industry to create a safer future for employees

Description: The highest priority on any work site especially, a civil construction site, is safety.

The Australian Work Health and Safety Strategy 2012-2022 describes the construction industry more broadly, as a priority industry for work health and safety. While many process exist that govern safe operations in a construction workplace any opportunity to further reduce the risk to workers is of immense value to society. This research will consider novel ways of classifying individual worker daily and intermediary safety checks, diary, project and company safety related data and operational scheduling data and discover insights that can help better predict safety risks to workers on construction sites whilst improving productivity gains.

Prof Fang Lee Cooke, Dr Lan Du, and Dr Paul Zhou CLOSED

Machine learning approaches in distributed optimisation and/or distributed learning using EDGE computing within a microgrid setting

This project will explore the application of machine learning approaches in either distributed optimisation and/or distributed learning using EDGE computing within a microgrid setting.

As a Graduate Research Industry Partnership project, the candidate will have access to knowledge and resources from Monash University and industry partner Engie, including state-of-the-art machine learning hardware (NVIDIA DGX 2), EDGE computing infrastructure and Monash’s state-of-the-art Microgrid. The applicant will also have access large scale proprietary datasets. It also provides the opportunity for research to be tested and implemented in solving real-world problems during the project cycle.

Assoc Prof Ariel Liebman OPEN

How to Apply

GRIP students are expected to begin in Semester 2, 2019.

The GRIP will then review all our applicants and successful candidates will be formally invited to "apply" to join Monash University. These successful candidates will then need to apply to enrol with Monash University. Further details will be provided along with the formal invitations.

Request to apply

Once you have completed the Request to Apply form, please email your academic transcripts to grip.hila@monash.edu

Expressions of Interest

If you do not yet see a suitable topic but are interested in the domain area's above, please visit our PhD Student Enquiries and complete our Expressions Of Interest (EOI) form, we will keep you notified as project topics become available.

The following table while not formal projects is indicative of the areas of research that may be potentially explored as part of the HiLA GRIP Program with Monash.

Research area Academic sponsors
Guidance in the Human-Machine Analysis Process Tim Dwyer (IT) Diane Cook (BusEco)
Building Human Trust in Machine Analytics and Optimisation: Helping users understand and interrogate the optimization results Tim Dwyer (IT) Maria Garcia de Lab Banda (IT) Nao Tsuchiya (MNHS)
Interactive Machine Learning Geoff Webb (IT) Nao Tsuchiya (MNHS)
Inductive Knowledge Refinement Geoff Webb (IT) Melissa Castan (Law)
Machine Learning for Decision Support Geoff Webb (IT) Diane Cooke (BusEco)
Human-centred AI, based on a deeper behavioural analytic understanding of users’ emotion, cognition, health & mental health status Sharon Oviatt (IT) Fang Lee Cooke (BusEco)
Multimodal learning analytics to identify students’ expertise & learning status Sharon Oviatt (IT) Kris Ryan (OLT)
Multimodal communication interfaces that stimulate cognition for native communicators of different world languages Sharon Oviatt (IT) Kris Ryan (OLT)
Causal network analysis of neural recordings (quantification of ‘integrated information’ and its relation to level and contents of consciousness) Nao Tsuchiya (MNHS) Tim Dwyer (IT)
Building/evolving artificial agents that can show generalization performance across massively different tasks and situations without learning (like conscious agents) Nao Tsuchiya (MNHS) Geoff Webb (IT)
Using category theory to define and analyse consciousness Nao Tsuchiya (MNHS) Tim Dwyer (IT)
Designing the human rights chip: ethical AI/ML systems Melissa Castan (Law) Tom Drummond (Eng.)
Analysis of Privacy by Design: regulatory and legal dimensions of data and information innovation (eg. cybersecurity, human services and public accountability) Melissa Castan (Law) Joanne Evans (IT)
Designing ethical algorithmic systems to adequately account for their decision-making, actions and impact on individuals and communities in, and through, time Joanne Evans (IT) Melissa Castan (Law)
Harnessing algorithmic capabilities to enfranchise individuals and communities in their interactions with institutional services and systems Joanne Evans (IT) Wray Buntine (IT) Fang Lee Cooke (BusEco)
Enabling multiple decision-makers to interact and reach joint decisions Fang Lee Cooke (BusEco) Maria de la Banda (IT)
Searching for an answer in incomplete forensic image-space  
Immersive Visualisation of Medical Images Assoc. Prof. Tim Dwyer, Prof. Richard Bassed and Dr Maxime Cordeil