Study Causal Inference Techniques for Data-Driven Personalised Decision-Making

With the rapid accumulation of big data, data-driven personalised decision-making is becoming a reality in various areas, such as personalised online recommendation, precision medicine targeting specific patient subgroups and personalised teaching and learning. In the area of causal inference, heterogeneous treatment effect estimation has been studied extensively, with the goal of identifying the different effects of a treatment on different subpopulations or individuals. Recently machine learning techniques have been introduced for heterogeneous treatment effect estimation to deal with large and observational data, e.g. gene expression for identifying patient subgroups characterised by their distinct genetic features which possibly have led to the heterogeneous effects of a cancer treatment in the different subgroups. The existing machine learning techniques, however, are facing two major challenges: how to accurately identify subgroups from observational data and how to efficiently deal with large scale and multiple sources of data. This project aims to develop new machine learning and causal inference techniques to tackle the challenges. The outcome of the project can be applied to various application areas, e.g. medicine, particularly cancer treatment, business intelligence, and government policy-making.

This project aims at studying how deep learning methods can help solve the causal inference problem. There are two aspects I want to explore for my Vacation Research Scholarship project:

a) Evaluate the algorithms presented in Fredrik D.Johansson, Uri Shalit, David Sontag, Learning Representations for Counterfactual Inference (for the first three weeks)
b) Study the semantic interpretations for representations (for the second three weeks)

Zhou Dai

University of New England

Zhou Dai is a third-year Mathematics Science student at the University of South Australia. His interests include the application of variety of Mathematics theories specially Machining Learning related techniques. He has been through several projects applying ML, one of which was to apply Naive Bayes classification to detect variations in the data sets supplied by the South Australia Government of Environment and Mining. Another project was to apply Hidden Markov Model to detect regimes for the stock market. Zhou believes everything is in the form of compositions of sinusoidal, which is similar to the situation that convergent functions could be approximately synthesised by sinusoidal on different frequencies. Although it seems not possible to predict for the future on historical data, but it is practicable to recognise patterns through the periodic history by applying advanced techniques such as Machine Learning or Artificial Intelligence.

You may be interested in

Jonah Klowss

Jonah Klowss

Finite Transition Times for Coupled Linear Diffusion Problems
James McCusker

James McCusker

Quarternions And Octonions
Nicholas Garrett

Nicholas Garrett

Curve Shortening Flow and Grayson's Theorem
Joshua Wilson

Joshua Wilson

Computational Modelling Using Exponential Integrators
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.