Dan MacKinlay — The Dan MacKinlay stable of variably-well-consider’d enterprises

You may have come here looking for one of my other projects, or what I am up to right now.

Or maybe you did come here looking for me? That is sweet of you.

Hi, I’m Dan MacKinlay.

1 Research interests

I’m a statistician and musician based in Melbourne, Australia. Musician should be clear. Statistician, though, what’s that? A statistician is the exact same thing as a data scientist or machine learning researcher, but on average about one standard deviation snarkier.

My specialty is the application of statistical inference to AI Safety and the process of scientific discovery, especially in geospatial settings. My methods of interest include Bayesian neural nets, sparse coding, Gaussian processes, sequential Monte Carlo methods, factor graphs, scientific machine learning… actually quite a lot of stuff.

I’m currently employed at CSIRO researching topics in hybrid machine learning methods for physical sciences.

1.1 Publications

Davies et al. (2025): Inference in spaces where you are not sure how many dimensions the answer has. The trick is the combine reparameterization trick with Bayesian Optimization to be efficient. And many other things, we learned a lot of trick to write this paper, many of whcih are not in the body text any longer.
Hasan et al. (2025): PC-SRGAN is a paper that demonstrates astonishingly good performance in super resolution in adversarial + Physics-loss training in super-resolution physical models. Astonishing because this was, to my mind, a nifty trick that wouldn’t work, on the way to something much harder. Props to Rakibul Hasan, Pouria Behnoudfar and Thomas Poulet whose optimism about this technique was vindicated.
Dan MacKinlay (2025): Gaussian process models for geospatial data and physics (See the blog post)
Dan MacKinlay et al. (2025): A short useful note connecting Ensemble Kalman filters to pathwise empirical Gaussian processes; this one is a bit technical but useful for the right audience.
Dabrowski et al. (2023): Bayesian inference for bushfire risk using PINNs
Pagendam et al. (2023): geospatial neural nets for hydrology
Takamoto et al. (2022): benchmarking PDE solvers based on neural nets; see pdebench/PDEBench: An Extensive Benchmark for Scientific Machine Learning
PhD thesis (Rare event simulation and audio synthesis, together at last)
MSc thesis (Predicting virality of Youtube videos using a mathematical model of sparse branching dynamics).
others older and more recondite (Ben Rached et al. 2020; Botev, MacKinlay, and Chen 2017; Daniel MacKinlay and Botev 2019; Dan MacKinlay et al. 2021)

Or check my Google Scholar.

2 Previously

I’ve been based at various times in places such as

Zürich, Switzerland, where I did my MSc in statistics under Professors Sara van de Geer and Didier Sornette at the Swiss Federal Institute of Technology,
Sydney, Australia, where I visualised data for the Powerhouse Museum under Seb Chan,
Bandung, Indonesia, where I worked on interactive music with Common Room with Gustaff Harriman Iskandar and Robi Rusdiana,
University of New South Wales Sydney, where I undertook graduate studies with Zdravko Botev,
Western Australia, a giant introvert colony

Minimalist CV here.

3 Contact me

If you have projects with a budget for data science, statistics or machine learning, AI, or whatever they call it this week, let me know. My current contract will not last forever.

If you do not have a budget but believe your cause is righteous, you are welcome to pitch to me for a slice of my volunteer time. Send a short paragraph making a case for what that time will help (I tend to favour AI, climate and conflict risk mitigation causes) and why you think it will be a high-leverage use of my time towards that goal.

Contact form here.

4 Follow me

Automatic updates in your feed reader:

or subscribe to emails:

You can find me on hypothes.is, Google Scholar, LinkedIn, …

orcid.org/0000-0001-6077-2684 .

5 Projects

What I am doing right now.

What I am doing in general:

The most obvious project is this very site, concerning whatever shiny thing distracted me and needed notes taken.
Lotekno, a Sundanese pop/electronic dance music crossover collaboration that I co-founded.
Bodywerk, my slammin’ DJ project.
Synestizer, an audiovisual synthesizer for which I was technical lead.
Efek Samping, my solo abstract electronic art music project.
My Github hosts various software I wrote.

6 Values and metrics

I aspire to be constructively deviant and truth-seeking.

By the ClearerThinking.org’s Intrinsic Values Test, my intrinsic values are

That I increase my understanding of reality beyond my current understanding
That humanity does not go extinct
That I feel connected to other people
That I believe true rather than false things
That beautiful things continue to come into existence (e.g. art or music)
That I get to experience a wide variety of different things during my life

I think that this fails to capture my community orientation; I value doing things with people more than this “feeling connected” metric captures, I think.

7 My face

Thanks to Abdelwahed Khamis for that portrait.

8 References

Ben Rached, MacKinlay, Botev, et al. 2020. “A Universal Splitting Estimator for the Performance Evaluation of Wireless Communications Systems.” IEEE Transactions on Wireless Communications.

Botev, MacKinlay, and Chen. 2017. “Logarithmically Efficient Estimation of the Tail of the Multivariate Normal Distribution.” In 2017 Winter Simulation Conference (WSC).

Dabrowski, Pagendam, Hilton, et al. 2023. “Bayesian Physics Informed Neural Networks for Data Assimilation and Spatio-Temporal Modelling of Wildfires.” Spatial Statistics.

Davies, Mackinlay, Oliveira, et al. 2025. “Amortized Variational Transdimensional Inference.”

Hasan, Behnoudfar, MacKinlay, et al. 2025. “PC-SRGAN: Physically Consistent Super-Resolution Generative Adversarial Network for General Transient Simulations.”

MacKinlay, Dan. 2025. “The Ensemble Kalman Update Is an Empirical Matheron Update.”

MacKinlay, Daniel, and Botev. 2019. “Mosaic Style Transfer Using Sparse Autocorrelograms.” In Proceedings of the 20th Conference of the International Society for Music Information Retrieval.

MacKinlay, Dan, Pagendam, Kuhnert, et al. 2021. “Model Inversion for Spatio-Temporal Processes Using the Fourier Neural Operator.” In.

MacKinlay, Dan, Tsuchida, Pagendam, et al. 2025. “Gaussian Ensemble Belief Propagation for Efficient Inference in High-Dimensional Systems.” In Proceedings of the International Conference on Learning Representations (ICLR).

Pagendam, Janardhanan, Dabrowski, et al. 2023. “A Log-Additive Neural Model for Spatio-Temporal Prediction of Groundwater Levels.” Spatial Statistics.

Takamoto, Praditia, Leiteritz, et al. 2022. “PDEBench: An Extensive Benchmark for Scientific Machine Learning.” In.