Stephen Tierney

Lecturer, data scientist and software engineer.

Currently at The University of Sydney where I teach computational statistics, data science and machine learning.

I am available for contract and consulting work.

Recent Activity

SCD4X Support in ESPHome

Upcoming - Code

I added support for the SCD4X CO2 sensor to ESPHome. Pull request on Github.

LTR390 Support in ESPHome

Upcoming - Code

I added support for the LTR390 UV and ambient light sensor to ESPHome. Pull request on Github.

Python + Ed: code for collaborative learning

16/09/2021 - In the Media

I was interviewed by the USYD Co-Design Research Group about our innovative approach to teaching online.

At the Turn of the Tide – Teaching Python in Real-Time at Scale

10/09/2021 - Talks

Alison Wong and I presented at the Education track of PyCon AU 2021 about our innovative approach to teaching online.

PMSA003I Support in ESPHome

18/08/2021 - Code

I added support for the PMSA003I particulate matter sensor to ESPHome. Pull request on Github.

Efficient Sparse Subspace Clustering by Nearest Neighbour Filtering

1/8/2021 - Papers - Signal Processing

KNN filtering reduces computation requirements for Subspace Clustering. Theorertical justification is provided and empirically validated. View on Google Scholar.

SHT4X Support in ESPHome

19/5/2021 - Code

I added support for the SHT4X temperature and humidity sensor to ESPHome. Pull request on Github.

Forecasting website

27/10/2020 - Talks

Andrey Vasnev and I presented at ISF 2020 about the Forecast Lab, which provides easily accessible economic and financial forecasts.

Top teachers should have a 50 per cent pay rise, expert says

10/10/2021 - In the Media

This Sydney Morning Herald article uses visualiations from my 'Teacher Pay Comparison' dashboard.


13/7/2020 - Code

I wrote a Python implementation of 'Better Estimates from Binned Income Data'. This technique allows distributions to be estimated from binned data.

Teacher Pay Comparison

16/6/2020 - Code

I made a supplemental dashboard to accompany submission to 'Valuing the teaching profession' inquiry. The dashboard uses Dash and the code is available on Github.

Robust Functional Manifold Clustering

2/4/2020 - Papers - IEEE Transactions on Neural Networks and Learning Systems

We extend subspace clustering to data that represents functions and curves. View on Google Scholar.

Too Many Guns

3/10/2019 - Code

I recreated this Greens NSW website using Dash. Code on Github.


21/4/2019 - Code

I wrote a wrapper for Dash to support multi page apps.


7/2/2019 - Code

I wrote a Python package for function caching that prevents re-entrant calls.

Selective Multi-Source Total Variation Image Restoration

23/11/2015 - Papers- DICTA 2015

We show that fusing multiple noisy images under a total variation penalty improves reconstruction. This idea can be selectively applied to individual pixels of each source image in the case of corrupted sections of images. View on Google Scholar.

Low Rank Sequential Subspace Clustering

12/7/2015 - Papers - IJCNN 2015

We show that forcing the affinity matrix of SpatSC to be low-rank improves performance. View on Google Scholar.

Affinity Pansharpening and Image Fusion

25/11/2014 - Papers - DICTA 2014

We show that the pixel affinity from a high resolution image can be used to enhance low resolution images of the same scene. This technique is general enough to support fusion between any image types. View on Google Scholar.

Subspace Clustering for Sequential Data

2014 - Papers - CVPR 2014

We show that subspace clustering on sequential data can be dramatically improved by incorporating spatial constraints. View on Google Scholar.

Image Matting for Sparse User Input by Iterative Refinement

26/11/2013 - Papers - DICTA 2013

We show that iteratively refining an alpha matte can lead to impressive results with very sparse labels. View on Google Scholar.

Natural Image Matting with Total Variation Regularisation

3/12/2012 - Papers

We show that total variation regularisation improves alpha mattes. View on Google Scholar.