Related papers: LINX: A Fast, Differentiable, and Extensible Big B…
Big Bang nucleosynthesis (BBN) theory predicts the abundances of the light elements D, $^3$He, $^4$He and $^7$Li produced in the early universe. The primordial abundances of D and $^4$He inferred from observational data are in good…
As the early universe expands and cools the rates of the weak interactions that keep neutrinos in thermal equilibrium with the matter and the related rates of the reactions that inter-convert neutrons and protons decrease. Eventually, these…
The physics of the standard hot big bang cosmology ensures that the early Universe was a primordial nuclear reactor, synthesizing the light nuclides (D, 3He, 4He, and 7Li) in the first 20 minutes of its evolution. After an overview of…
We improve Standard Big Bang Nucleosynthesis (SBBN) calculations taking into account new nuclear physics analyses (Descouvemont et al. 2003). Using a Monte-Carlo technique, we calculate the abundances of light nuclei versus the baryon to…
We calculate primordial abundances of light elements produced during big bang nucleosynthesis when the fine structure constant and/or the cosmic expansion rate take non-standard values. We compare them with the recent values of observed D,…
We consider the big bang nucleosynthesis (BBN) in the Brane world scenario, where all matter fields are confined on our 3-brane and the radion of the Brane evolves cosmologically. In the Einstein frame fundamental fermion masses vary and…
We discuss nuclear reactions which could play a role in Big Bang Nucleosynthesis (BBN). Most of these reactions involve lithium and beryllium isotopes and the rates for some of these have not previously been included in BBN calculations.…
In this paper we present an algorithm for rapid Bayesian analysis that combines the benefits of nested sampling and artificial neural networks. The blind accelerated multimodal Bayesian inference (BAMBI) algorithm implements the MultiNest…
Laser-induced breakdown spectroscopy (LIBS) is a laser based diagnostics used to study atomic emission from the expanding plasma plume formed during the laser-matter interaction. It provides valuable information about the composition of the…
It has been suggested that relic long-lived strongly interacting massive particles (SIMPs, or $X$ particles) existed in the early universe. We study effects of such long-lived unstable SIMPs on big bang nucleosynthesis (BBN) assuming that…
Within the first 20 minutes of the evolution of the hot, dense, early Universe, astrophysically interesting abundances of deuterium, helium-3, helium-4, and lithium-7 were synthesized by the cosmic nuclear reactor. The primordial abundances…
The high inference demands of transformer-based Large Language Models (LLMs) pose substantial challenges in their deployment. To this end, we introduce Neural Block Linearization (NBL), a novel framework for accelerating transformer model…
We introduce Larq Compute Engine, the world's fastest Binarized Neural Network (BNN) inference engine, and use this framework to investigate several important questions about the efficiency of BNNs and to design a new state-of-the-art BNN…
We present updated and improved Big Bang Nucleosynthesis (BBN) constraints on heavy, long-lived beyond the Standard Model (BSM) relics $\phi$ decaying into pairs of Standard Model particles, covering a comprehensive set of two-body decay…
We present msmJAX, a Python package implementing the multilevel summation method with B-spline interpolation, a linear-scaling algorithm for efficiently evaluating electrostatic and other long-range interactions in particle-based…
JAX-bandflux is a JAX implementation of critical supernova modelling functionality for cosmological analysis. The codebase implements key components of the established library SNCosmo in a differentiable framework, offering efficient…
Large language models have high compute, latency, and memory requirements. While specialized accelerators such as GPUs and TPUs typically run these workloads, CPUs are more widely available and consume less energy. Accelerating LLMs with…
Image hash codes are produced by binarizing the embeddings of convolutional neural networks (CNN) trained for either classification or retrieval. While proxy embeddings achieve good performance on both tasks, they are non-trivial to…
We update the Big Bang Nucleosynthesis (BBN) calculations on the basis of the recent NACRE compilation of reaction rates. We estimate the uncertainties related to the nuclear reaction rates on the abundances of D, 3He, 4He, 6Li, 7Li, 9Be,…
We present a simple method for determining the (correlated) uncertainties of the light element abundances expected from big bang nucleosynthesis, which avoids the need for lengthy Monte Carlo simulations. Our approach helps to clarify the…