Design of Neural Networks Using Randomised Search Techniques

Taking inspiration from the human brain, (artificial) neural networks provide highly flexible ways for a computer to learn patterns in data. In recent years it has become clear that both the sparsity and structure of a neural network are essential for fast and efficient learning. The aim of this project is to investigate the efficient design of deep learning networks for various classes of regression and classification problems via randomised search strategies. The idea is that such designs will be optimal in the sense that the noisy optimisation problem of training the network is efficient and that they are flexible in their capabilities.

Jonathan Wilton

University of Queensland

Jonathan is a student at the University of Queensland completing his third year of a Bachelor of Mathematics with a major in Statistics. His academic interests lie predominantly in Probability Theory, Mathematical Statistics and Machine Learning. Throughout his degree, Jonathan has had experience as a Teacher’s Assistant (Tutor) for The University of Queensland in the School of Mathematics and Physics. The AMSI Vacation Research Scholarship will provide excellent experience for continuing his studies into an honours project in the coming year.

You may be interested in

Gyu Hwan Park

Gyu Hwan Park

Variational Inference for Bayesian Nonnegative Matrix Factorisation
Meg Tully

Meg Tully

Statistical Modelling of Malaria Parasite Clearance
Kulunu Dharmakeerthi

Kulunu Dharmakeerthi

Generating New Instance Spaces for Stress-Testing Classification Algorithm
Abhishek Varghese

Abhishek Varghese

Adaptive Tolerance Selection for SMC ABC
Contact Us

We're not around right now. But you can send us an email and we'll get back to you, asap.

Not readable? Change text.