Rixon Crane

About Me

I'm a researcher in optimization and machine learning, with a focus on developing methods that are efficient, scalable, and reliable. I enjoy challenging problems where mathematical theory, algorithm design, modeling, and real-world applications come together. I'm especially interested in understanding why a method works, how efficiently it can be implemented, and how reliably it performs as problems grow in scale and complexity. I've published at ICML and NeurIPS and have experience with Python, PyTorch, and JAX.

I'm currently a CERC Postdoctoral Fellow in Robotics at CSIRO in Brisbane, where I've been working on point-cloud registration, neural network models, and simultaneous localization and mapping (SLAM). I led the development of MMD-Reg, a scalable and differentiable registration method based on maximum mean discrepancy, published at ICML 2026.

Before joining CSIRO, I completed a PhD in mathematics at The University of Queensland. My thesis was on efficient second-order optimization methods for large-scale machine learning. As part of this research, I developed the distributed Newton-type methods DINGO, published at NeurIPS 2019, and DINO, published at ICML 2020. I also worked on an invexifying regularization for non-linear least-squares problems. During my PhD, I was part of the 2018 cohort of the Global Change Scholars Program. I was also a conference organizer for INFORMS APS 2019 and a workshop organizer for ICML 2021.

Before my PhD, I studied at The University of Queensland, where I completed a Bachelor of Advanced Science (Honours) with a major in mathematics and graduated with First Class Honours. My honours thesis, Plane Partitions in Number Theory and Algebra, also gave me experience in pure mathematics.

Research Interests

Artificial Intelligence · Machine Learning · Optimization · Robotics · Distributed Computing

Papers

• Scalable and Differentiable Point-Cloud Registration Using Maximum Mean Discrepancy
  • Rixon Crane, Fahira Afzal Maken, Nicholas Lawrance, Stanislav Funiak, Kasra Khosoussi, Ming Xu, Russell Tsuchida
  • ICML 2026
  • Paper · arXiv · Code
• Efficient second-order optimisation methods for large scale machine learning
  • Rixon Crane
  • PhD Thesis, The University of Queensland, 2022
  • Paper
• Invexifying Regularization of Non-Linear Least-Squares Problems
  • Rixon Crane, Fred Roosta
  • arXiv 2021
  • arXiv · Code
• DINO: Distributed Newton-Type Optimization Method
  • Rixon Crane, Fred Roosta
  • ICML 2020
  • Paper · arXiv · Code
• DINGO: Distributed Newton-Type Method for Gradient-Norm Optimization
  • Rixon Crane, Fred Roosta
  • NeurIPS 2019
  • Paper · arXiv · Code