The James Webb Space Telescope (JWST) hosts a non-redundant Aperture Masking Interferometer (AMI) in its Near Infrared Imager and Slitless Spectrograph (NIRISS) instrument, providing the only dedicated interferometric facility aboard — magnitudes more precise than any interferometric experiment previously flown. However, the performance of AMI (and other high resolution approaches such as kernel phase) in recovery of structure at high contrasts has not met design expectations. A major contributing factor has been the presence of uncorrected detector systematics, notably charge migration effects in the H2RG sensor, and insufficiently accurate mask metrology. Here we present AMIGO, a data-driven calibration framework and analysis pipeline that forward-models the full JWST AMI system — including its optics, detector physics, and readout electronics — using an end-to-end differentiable architecture implemented in the JAX framework and in particular exploiting the ∂LUX optical modelling package. AMIGO directly models the generation of up-the-ramp detector reads, using an embedded neural sub-module to capture non-linear charge redistribution effects, enabling the optimal extraction of robust observables, for example kernel amplitudes and phases, while mitigating systematics such as the brighter-fatter effect. We demonstrate AMIGO’s capabilities by recovering the AB Dor AC binary from commissioning data with high-precision astrometry, and detecting both HD 206893 B and the inner substellar companion HD 206893 c: a benchmark requiring contrasts approaching 10 magnitudes at separations of only 100mas. These results exceed outcomes from all published pipelines, and re-establish AMI as a viable competitor for imaging at high contrast at the diffraction limit. AMIGO is publicly available as open-source software community resource.

Flying on board the James Webb Space Telescope (JWST) above Earth’s turbulent atmosphere, the Aperture Masking Interferometer (AMI) on the NIRISS instrument is the highest-resolution infrared interferometer ever placed in space. However, its performance was found to be limited by non-linear detector systematics, particularly charge migration – or the Brighter-Fatter Effect. Conventional interferometric Fourier observables are degraded by non-linear transformations in the image plane, with the consequence that the inner working angle and contrast limits of AMI were seriously compromised. Building on the end-to-end differentiable model & calibration code , we here present a regularised maximum-likelihood image reconstruction framework , which can deconvolve AMI images either in the image plane or from calibrated Fourier observables, achieving high angular resolution and contrast over a wider field of view than conventional interferometric limits. This modular code by default includes regularisation by maximum entropy, and total variation defined with $l_1$ or $l_2$ metrics. We present imaging results from dorito for three benchmark imaging datasets: the volcanoes of Jupiter’s moon Io, the colliding-wind binary dust nebula WR 137 and the archetypal Seyfert 2 active galactic nucleus NGC 1068. In all three cases, we recover images consistent with the literature at diffraction-limited resolutions. The performance, limitations, and future opportunities enabled by amigo for AMI imaging (and beyond) are discussed.

Deep-learning algorithms have gained much popularity in the past decade. However, their supervised nature makes them fully dependent on the quality and completeness of the data samples used in the training processes. And when it comes to galaxy spectra for instance, there is simply no suited training sample available yet. This is where unsupervised classification (clustering) can come to the rescue.

Using the discriminant latent mixture-model based algorithm Fisher-EM, we 1) demonstrate the discriminative capacity of the method and its robustness with respect to noise on a sample of galaxy spectra simulated with the code CIGALE, 2) manage to classify ∼700 000 spectra of galaxies from the SDSS, and 3) extend this classification to higher redshifts thanks to the VIPERS data (∼80 000 spectra up to a redshift of 1.2). Finally, it appears that FisherEM is efficient for images of galaxies as well.

Large area astronomical surveys will almost certainly contain new objects of a type that have never been seen before. The detection of ‘unknown unknowns’ by an algorithm is a difficult problem to solve, as unusual things are often easier for a human to spot than a machine. We use the concept of apparent complexity, previously applied to detect multi-component radio sources, to scan the radio continuum Evolutionary Map of the Universe (EMU) Pilot Survey data for complex and interesting objects in a fully automated and blind manner. Here we describe how the complexity is defined and measured, how we applied it to the Pilot Survey data, and how we calibrated the completeness and purity of these interesting objects using a crowd-sourced ‘zoo’. The results are also compared to unexpected and unusual sources already detected in the EMU Pilot Survey, including Odd Radio Circles, that were found by human inspection.

Machine Learning is a powerful tool for astrophysicists, which has already had significant uptake in the community. But there remain some barriers to entry, relating to proper understanding, the difficulty of interpretability, and the lack of cohesive training. In this discussion session we addressed some of these questions, and suggest how the field may move forward.

As the scale of cosmological surveys increases, so does the complexity in the analyses. This complexity can often make it difficult to derive the underlying principles, necessitating statistically rigorous testing to ensure the results of an analysis are consistent and reasonable. This is particularly important in multi-probe cosmological analyses like those used in the Dark Energy Survey (DES) and the upcoming Legacy Survey of Space and Time, where accurate uncertainties are vital. In this paper, we present a statistically rigorous method to test the consistency of contours produced in these analyses and apply this method to the Pippin cosmological pipeline used for type Ia supernova cosmology with the DES. We make use of the Neyman construction, a frequentist methodology that leverages extensive simulations to calculate confidence intervals, to perform this consistency check. A true Neyman construction is too computationally expensive for supernova cosmology, so we develop a method for approximating a Neyman construction with far fewer simulations. We find that for a simulated dataset, the 68% contour reported by the Pippin pipeline and the 68% confidence region produced by our approximate Neyman construction differ by less than a percent near the input cosmology; however, they show more significant differences far from the input cosmology, with a maximal difference of 0.05 in $\Omega_{M}$ and 0.07 in w. This divergence is most impactful for analyses of cosmological tensions, but its impact is mitigated when combining supernovae with other cross-cutting cosmological probes, such as the cosmic microwave background.

Faraday complexity describes whether a spectropolarimetric observation has simple or complex magnetic structure. Quickly determining the Faraday complexity of a spectropolarimetric observation is important for processing large, polarised radio surveys. Finding simple sources lets us build rotation measure grids, and finding complex sources lets us follow these sources up with slower analysis techniques or further observations. We introduce five features that can be used to train simple, interpretable machine learning classifiers for estimating Faraday complexity. We train logistic regression and extreme gradient boosted tree classifiers on simulated polarised spectra using our features, analyse their behaviour, and demonstrate our features are effective for both simulated and real data. This is the first application of machine learning methods to real spectropolarimetry data. With 95% accuracy on simulated ASKAP data and 90% accuracy on simulated ATCA data, our method performs comparably to state-of-the-art convolutional neural networks while being simpler and easier to interpret. Logistic regression trained with our features behaves sensibly on real data and its outputs are useful for sorting polarised sources by apparent Faraday complexity.

We report the results from our ongoing pilot investigation of the use of deep learning techniques for forecasting the state of turbulent flows onto black holes. Deep neural networks seem to learn well black hole accretion physics and evolve the accretion flow orders of magnitude faster than traditional numerical solvers, while maintaining a reasonable accuracy for a long time.