Talks

Recent advancements have positioned Large Language Models (LLMs) as transformative tools for scientific research, capable of addressing complex tasks that require reasoning, problem-solving, and decision-making. Their exceptional capabilities suggest their potential as scientific research assistants, but also highlight the need for holistic, rigorous, and domain-specific evaluation to assess effectiveness in real-world scientific applications. This talk describes a multifaceted methodology for Evaluating AI models as scientific Research Assistants (EAIRA) developed at Argonne National Laboratory. This methodology incorporates four primary classes of evaluations. 1) Multiple Choice Questions to assess factual recall; 2) Open Response to evaluate advanced reasoning and problem-solving skills; 3) Lab-Style Experiments involving detailed analysis of capabilities as research assistants in controlled environments; and 4) Field-Style Experiments to capture researcher-LLM interactions at scale in a wide range of scientific domains and applications. These complementary methods enable a comprehensive analysis of LLM strengths and weaknesses with respect to their scientific knowledge, reasoning abilities, and adaptability. Recognizing the rapid pace of LLM advancements, we designed the methodology to evolve and adapt so as to ensure its continued relevance and applicability. This talk describes the current methodology's state. Although developed within a subset of scientific domains, the methodology is designed to be generalizable to a wide range of scientific domains.

Morphological summary statistics, such as Minkowski functionals, Betti numbers and Minkowski tensors, provide a route to carry out cosmological data analysis which is complementary to traditional mode decomposition based statistics such as the power spectrum. We summarize the properties of these statistics and demonstrate their information content by using the gravitational evolution of matter and 21-cm brightness temperature as physical examples.

The formation of the first stars in the Universe remains a key challenge in cosmology, requiring a multi-wavelength observational strategy across a range of facilities. Telescopes like the James Webb Space Telescope (JWST) enable direct observation of early galaxies, while upcoming radio telescopes such as the SKA will capture the hydrogen signal from the early Universe. Interpreting the extensive datasets from these facilities demands sophisticated theoretical models that balance detailed physics with computational efficiency, supported by advanced statistical methods to probe the parameter space. In this talk, we will discuss recent advancements in modelling the high-redshift Universe, along with the evolution of machine learning-driven statistical techniques. Together, these developments are essential for bridging the gap between theoretical predictions and observations, enabling a deeper understanding of cosmic evolution with next-generation observational capabilities.

In this talk I will describe recent work done by the MIST Global 21-cm experiment. The focus will be on sky observations conducted from the Canadian High Arctic and on efforts to calibrate these observations with high accuracy.

During the Epoch of Reionisation (EoR), the ultraviolet radiation from the first stars and galaxies ionised the neutral hydrogen of the intergalactic medium, which itself can emit radiation through the 21 cm hyperfine transition. Due to this, the 21 cm signal is a direct probe of the first stars in the early Universe and a key science goal for the future Square Kilometre Array (SKA). However, observing and interpreting this signal is a notoriously difficult task.

Another high-potential probe is the patchy kinetic Sunyaev-Zel'dovich effect (pkSZ). Induced by the scattering of Cosmic Microwave Background (CMB) photons with a medium of free electrons produced during the EoR, the effect altered the small-scale CMB temperature anisotropies, imprinting information on the growth of ionising bubbles from the first galaxies. While measurements of the pkSZ angular power spectrum by Reichardt et al. (2021) have reported a 3σ constraint of D^pkSZ (l=3000) = 3.0 ± 1.0 μK2, the results are also subject to modelling uncertainties.

In this talk, we propose a simple yet effective parametric model that establishes a formal connection between the 21 cm and pkSZ power spectra. Using this model to jointly fit mock 21 cm and pkSZ data points, we confirm that these two observables exhibit complementary characteristics, leading to significantly improved constraints on reionisation compared to analysing each data set separately. Our findings demonstrate that a few well-informed low-redshift (eg., z < 8) measurements of the 21 cm power spectrum at k ≈ 0.1 cMpc^-1 and pkSZ power spectra can precisely determine the reionisation history of the Universe.

Therefore, even in the early stages of observations with the SKA, we can begin to constrain cosmic reionisation by performing a combined analysis of the 21 cm power spectrum with the pkSZ observations.

The standard way to do this is to use the chain rule to backpropagate gradients through layers of neurons. I shall briefly review a few of the engineering successes of backpropagation and then describe a very different way of getting the gradients that, for a while, seemed a lot more plausible as a model of how the brain gets gradients. Consider a system composed of binary neurons that can be active or inactive with weighted pairwise couplings between pairs of neurons, including long range couplings. If the neurons represent pixels in a binary image, we can store a set of binary training images by adjusting the coupling weights so that the images are local minima of a Hopfield energy function which is minus the sum over all pairs of active neurons of their coupling weights. But this energy function can only capture pairwise correlations. It cannot represent the kinds of complicated higher-order correlations that occur in images. Now suppose that in addition to the "visible" neurons that represent the pixel intensities, we also have a large set of hidden neurons that have weighted couplings with each other and with the visible neurons. Suppose also that all of the neurons are asynchronous and stochastic: They adopt the active state with a log odds that is equal to the difference in the energy function when the neuron is inactive versus active. Given a set of training images, is there a simple way to set the weights on all of the couplings so that the training images are local minima of the free energy function obtained by integrating out the states of the hidden neurons? The Boltzmann machine learning algorithm solved this problem in an elegant way. It was proof of principle that learning in neural networks with hidden neurons was possible using only locally available information, contrary to what was generally believed at the time.