Bayesian deep learning for incomplete information

Summary

This project will feature a synergy of deep learning, modular and multi-task learning with Bayesian methods to address the problem of decision making given incomplete information. This will enable the development of robust deep learning architectures and learning algorithms in applications that include pattern recognition, time series prediction, and computer vision.

Supervisor(s)

Professor Sally Cripps, Dr Rohitash Chandra

Research Location

School of Mathematics and Statistics

Program Type

PHD

Synopsis

Incomplete information in problems and datasets is becoming a growing challenge for machine learning. Incomplete information could arise from 1) limited data availability, 2) datasets where certain features are missing in various locations, or 3) when the nature of the problem dynamically changes that requires decision making given absence of certain groups of features in the input space. In order to develop robust learning algorithms, it is important to take into account modular learning in order to utilise knowledge as building blocks. Traditional learning algorithms such as stochastic gradient descent provide point estimates or single solution for the weights that represent knowledge learnt in deep neural networks. As a result, these networks make predictions that do not account for uncertainty in the parameters. However, in several cases, it is desirable to produce uncertainty estimates in decision-making that could be addressed through Bayesian methods. It is important to develop robust learning algorithms and network architectures that can adapt with dynamic problems, environment, and inconsistent features in datasets.

Additional Information

The student will be based at the Centre for Translational Data Science.

HDR Inherent Requirements

In addition to the academic requirements set out in the Science Postgraduate Handbook, you may be required to satisfy a number of inherent requirements to complete this degree. Example of inherent requirement may include:

- Confidential disclosure and registration of a disability that may hinder your performance in your degree;
- Confidential disclosure of a pre-existing or current medical condition that may hinder your performance in your degree (e.g. heart disease, pace-maker, significant immune suppression, diabetes, vertigo, etc.);
- Ability to perform independently and/or with minimal supervision;
- Ability to undertake certain physical tasks (e.g. heavy lifting);
- Ability to undertake observatory, sensory and communication tasks;
- Ability to spend time at remote sites (e.g. One Tree Island, Narrabri and Camden);
- Ability to work in confined spaces or at heights;
- Ability to operate heavy machinery (e.g. farming equipment);
- Hold or acquire an Australian driver’s licence;
- Hold a current scuba diving license;
- Hold a current Working with Children Check;
- Meet initial and ongoing immunisation requirements (e.g. Q-Fever, Vaccinia virus, Hepatitis, etc.)

You must consult with your nominated supervisor regarding any identified inherent requirements before completing your application.

Want to find out more?

Contact us to find out what’s involved in applying for a PhD. Domestic students and International students

Contact Research Expert to find out more about participating in this opportunity.

Browse for other opportunities within the School of Mathematics and Statistics .

Keywords

deep learning, Bayesian methods, Machine learning

Opportunity ID

The opportunity ID for this research opportunity is: 2301

Other opportunities with Dr Rohitash Chandra