Topic: machine-learning methods

CIERA recruits researchers for autumn 2027 appointments in theoretical astronomy and astrophysics. Areas of interest include gravitational-wave analysis, stellar and compact-object evolution, interacting binaries, statistical and machine-learning methods with Vicky Kalogera, and protoplanetary and accretion-disk theory with Yoram Lithwick. Researchers receive access to dedicated computing allocations, GPU and high-memory resources, and computational support. Applications received by 15 December 2026 receive full consideration; later applications may be considered until positions are filled.

Marking a century of Richardson’s work on turbulent dispersion, this programme compares Lagrangian and Eulerian approaches, stochastic and deterministic models, particle dynamics and continuum descriptions. Topics include multiparticle dispersion, experimental particle tracking, geophysical transport, spontaneous stochasticity, mixing, and machine-learning methods for closure and inference. The programme connects theory, simulation and experiment with environmental and climate prediction. It targets doctoral students, postdocs and established researchers in turbulence and transport; exceptiona

Develop machine-learning methods that integrate multimodal human-brain data and characterise Alzheimer’s disease trajectories. Contribute to mechanistic analysis of disease using large-scale molecular, cellular and neuropathology datasets.

Develop machine-learning methods for sparse three-dimensional reconstruction of virtual histology from X-ray phase-contrast tomography. Work includes imaging breast biopsy specimens, constructing data pipelines, designing reconstruction algorithms and comparing results with standard histology. The doctoral researcher joins academic and clinical collaborators in studies intended to improve cancer diagnosis workflows, and contributes publications and teaching.

Develop climate projections that make extreme-weather risks useful for impact assessment. The postdoctoral work links climate data, statistical or machine-learning methods and applications relevant to society.

Yolanne Lee, Georg Maierhofer and Chris Budd report on a two-day Maths4DL hackathon investigating machine-learning methods for differential equations. They discuss how such systems can be taught to solve equations, how proposed methods should be evaluated, and the collaborative process of moving from an ambitious research question to workable tests.

Friedemann Zenke discusses learning in spiking networks and the challenges of translating machine-learning methods into biologically plausible rules.