arXiv subjects

Jordan Hoffmann

Publications and source records attributed to Jordan Hoffmann.

16 recordsLinked to original sources

arXiv2026

I can see your halo: Constraining the Milky Way halo DM with FRB population studies

Fast radio bursts (FRBs) probe the electron column density along the line of sight and hence can be used to probe foreground structures. One such structure is the Galactic halo. In this work, we use a total of 98 high Galactic latitude ($|b| > 20^\circ$) FRBs detected by ASKAP, Parkes, DSA and FAST with 32 associated redshifts to constrain the dispersion measure (DM) contribution from the Galactic halo. We simultaneously fit unknown FRB population parameters, which show correlations with the Galactic halo but are not completely degenerate. We primarily use an isotropic model for the halo, but find no evidence favouring a particular halo model. We find DM$_{\rm MW,halo}$=$68^{+27}_{-24}$pc/cm$^3$, which is in agreement with other results within the literature. Previous constraints on DM$_{\rm MW,halo}$ with FRBs have used a few, low-DM FRBs. However, this is highly subject to fluctuations between different lines of sight, and hence using a larger number of sightlines as we do is more likely to be representative of the true average contribution. Nevertheless, we show that individual FRBs can still skew the data significantly and hence will be important in the future for more precise results.

astro-ph.GA↗

arXiv2025

A fast radio burst from the first 3 billion years of the Universe

Fast radio bursts (FRBs) are enigmatic millisecond-duration signals which encode otherwise unattainable information on the plasma which permeates our Universe, providing insights into magnetic fields and gas distributions. Here we report the discovery of FRB 20240304B originating at redshift 2.148 +/- 0.001 corresponding to just 3 billion years after the Big Bang. FRB 2024030 was detected with the MeerKAT radio telescope and localized to a low-mass, clumpy, star forming galaxy using the James Webb Space Telescope. This discovery doubles the redshift reach of localized FRBs and probes ionized baryons across ~80% of cosmic history. Its sightline, intersecting the Virgo Cluster and a foreground group, reveals magnetic field complexity over many gigaparsec scales. Our observations establish FRB activity during the peak of cosmic star formation and demonstrate that FRBs can probe galaxy formation during the most active era in cosmic time.

astro-ph.HE↗

arXiv2024

The impact of the FREDDA dedispersion algorithm on $H_0$ estimations with FRBs

Fast radio bursts (FRBs) are transient radio signals of extragalactic origins that are subjected to propagation effects such as dispersion and scattering. It follows then that these signals hold information regarding the medium they have traversed and are hence useful as cosmological probes of the Universe. Recently, FRBs were used to make an independent measure of the Hubble Constant $H_0$, promising to resolve the Hubble tension given a sufficient number of detected FRBs. Such cosmological studies are dependent on FRB population statistics, cosmological parameters and detection biases, and thus it is important to accurately characterise each of these. In this work, we empirically characterise the sensitivity of the Fast Real-time Engine for Dedispersing Amplitudes (FREDDA) which is the current detection system for the Australian Square Kilometer Array Pathfinder (ASKAP). We coherently redisperse high-time resolution data of 13 ASKAP-detected FRBs and inject them into FREDDA to determine the recovered signal-to-noise ratios as a function of dispersion measure (DM). We find that for 11 of the 13 FRBs, these results are consistent with injecting idealised pulses. Approximating this sensitivity function with theoretical predictions results in a systematic error of 0.3$\,$km$\,$s$^{-1}\,$Mpc$^{-1}$ on $H_0$ when it is the only free parameter. Allowing additional parameters to vary could increase this systematic by up to $\sim1\,$km$\,$s$^{-1}\,$Mpc$^{-1}$. We estimate that this systematic will not be relevant until $\sim$400 localised FRBs have been detected, but will likely be significant in resolving the Hubble tension.

astro-ph.HE↗

arXiv2024

Modelling DSA, FAST and CRAFT surveys in a z-DM analysis and constraining a minimum FRB energy

Fast radio burst (FRB) science primarily revolves around two facets: the origin of these bursts and their use in cosmological studies. This work follows from previous redshift-dispersion measure ($z$-DM) analyses in which we model instrumental biases and simultaneously fit population parameters and cosmological parameters to the observed population of FRBs. This sheds light on both the progenitors of FRBs and cosmological questions. Previously, we have completed similar analyses with data from the Australian Square Kilometer Array Pathfinder (ASKAP) and the Murriyang (Parkes) Multibeam system. With this manuscript, we additionally incorporate data from the Deep Synoptic Array (DSA) and the Five-hundred-meter Aperture Spherical Telescope (FAST), invoke a Markov chain Monte Carlo (MCMC) sampler and implement uncertainty in the Galactic DM contributions. The latter leads to larger uncertainties in derived model parameters than previous estimates despite the additional data. We provide refined constraints on FRB population parameters and derive a new constraint on the minimum FRB energy of log$\,E_{\mathrm{min}}$(erg)=39.49$^{+0.39}_{-1.48}$ which is significantly higher than bursts detected from strong repeaters. This result may indicate a low-energy turnover in the luminosity function or may suggest that strong repeaters have a different luminosity function to single bursts. We also predict that FAST will detect 25-41% of their FRBs at $z \gtrsim 2$ and DSA will detect 2-12% of their FRBs at $z \gtrsim 1$.

astro-ph.CO↗

arXiv2023

Policy composition in reinforcement learning via multi-objective policy optimization

We enable reinforcement learning agents to learn successful behavior policies by utilizing relevant pre-existing teacher policies. The teacher policies are introduced as objectives, in addition to the task objective, in a multi-objective policy optimization setting. Using the Multi-Objective Maximum a Posteriori Policy Optimization algorithm (Abdolmaleki et al. 2020), we show that teacher policies can help speed up learning, particularly in the absence of shaping rewards. In two domains with continuous observation and action spaces, our agents successfully compose teacher policies in sequence and in parallel, and are also able to further extend the policies of the teachers in order to solve the task. Depending on the specified combination of task and teacher(s), teacher(s) may naturally act to limit the final performance of an agent. The extent to which agents are required to adhere to teacher policies are determined by hyperparameters which determine both the effect of teachers on learning speed and the eventual performance of the agent on the task. In the humanoid domain (Tassa et al. 2018), we also equip agents with the ability to control the selection of teachers. With this ability, agents are able to meaningfully compose from the teacher policies to achieve a superior task reward on the walk task than in cases without access to the teacher policies. We show the resemblance of composed task policies with the corresponding teacher policies through videos.

cs.LG↗

arXiv2022

Training Compute-Optimal Large Language Models

We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4$\times$ more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher.

cs.CL↗

arXiv2022

Unified Scaling Laws for Routed Language Models

The performance of a language model has been shown to be effectively modeled as a power-law in its parameter count. Here we study the scaling behaviors of Routing Networks: architectures that conditionally use only a subset of their parameters while processing an input. For these models, parameter count and computational requirement form two independent axes along which an increase leads to better performance. In this work we derive and justify scaling laws defined on these two variables which generalize those known for standard language models and describe the performance of a wide range of routing architectures trained via three different techniques. Afterwards we provide two applications of these laws: first deriving an Effective Parameter Count along which all models scale at the same rate, and then using the scaling coefficients to give a quantitative comparison of the three routing techniques considered. Our analysis derives from an extensive evaluation of Routing Networks across five orders of magnitude of size, including models with hundreds of experts and hundreds of billions of parameters.

cs.CL↗

arXiv2021

Improving language models by retrieving from trillions of tokens

We enhance auto-regressive language models by conditioning on document chunks retrieved from a large corpus, based on local similarity with preceding tokens. With a $2$ trillion token database, our Retrieval-Enhanced Transformer (RETRO) obtains comparable performance to GPT-3 and Jurassic-1 on the Pile, despite using 25$\times$ fewer parameters. After fine-tuning, RETRO performance translates to downstream knowledge-intensive tasks such as question answering. RETRO combines a frozen Bert retriever, a differentiable encoder and a chunked cross-attention mechanism to predict tokens based on an order of magnitude more data than what is typically consumed during training. We typically train RETRO from scratch, yet can also rapidly RETROfit pre-trained transformers with retrieval and still achieve good performance. Our work opens up new avenues for improving language models through explicit memory at unprecedented scale.

cs.CL↗

arXiv2021

Scaling Language Models: Methods, Analysis & Insights from Training Gopher

Language modelling provides a step towards intelligent communication systems by harnessing large repositories of written human knowledge to better predict and understand the world. In this paper, we present an analysis of Transformer-based language model performance across a wide range of model scales -- from models with tens of millions of parameters up to a 280 billion parameter model called Gopher. These models are evaluated on 152 diverse tasks, achieving state-of-the-art performance across the majority. Gains from scale are largest in areas such as reading comprehension, fact-checking, and the identification of toxic language, but logical and mathematical reasoning see less benefit. We provide a holistic analysis of the training dataset and model's behaviour, covering the intersection of model scale with bias and toxicity. Finally we discuss the application of language models to AI safety and the mitigation of downstream harms.

cs.CL↗

arXiv2021

A Systematic Investigation of Commonsense Knowledge in Large Language Models

Language models (LMs) trained on large amounts of data have shown impressive performance on many NLP tasks under the zero-shot and few-shot setup. Here we aim to better understand the extent to which such models learn commonsense knowledge -- a critical component of many NLP applications. We conduct a systematic and rigorous zero-shot and few-shot commonsense evaluation of large pre-trained LMs, where we: (i) carefully control for the LMs' ability to exploit potential surface cues and annotation artefacts, and (ii) account for variations in performance that arise from factors that are not related to commonsense knowledge. Our findings highlight the limitations of pre-trained LMs in acquiring commonsense knowledge without task-specific supervision; furthermore, using larger models or few-shot evaluation are insufficient to achieve human-level commonsense performance.

cs.CL↗

arXiv2020

AlgebraNets

Neural networks have historically been built layerwise from the set of functions in ${f: \mathbb{R}^n \to \mathbb{R}^m }$, i.e. with activations and weights/parameters represented by real numbers, $\mathbb{R}$. Our work considers a richer set of objects for activations and weights, and undertakes a comprehensive study of alternative algebras as number representations by studying their performance on two challenging problems: large-scale image classification using the ImageNet dataset and language modeling using the enwiki8 and WikiText-103 datasets. We denote this broader class of models as AlgebraNets. Our findings indicate that the conclusions of prior work, which explored neural networks constructed from $\mathbb{C}$ (complex numbers) and $\mathbb{H}$ (quaternions) on smaller datasets, do not always transfer to these challenging settings. However, our results demonstrate that there are alternative algebras which deliver better parameter and computational efficiency compared with $\mathbb{R}$. We consider $\mathbb{C}$, $\mathbb{H}$, $M_{2}(\mathbb{R})$ (the set of $2\times2$ real-valued matrices), $M_{2}(\mathbb{C})$, $M_{3}(\mathbb{R})$ and $M_{4}(\mathbb{R})$. Additionally, we note that multiplication in these algebras has higher compute density than real multiplication, a useful property in situations with inherently limited parameter reuse such as auto-regressive inference and sparse neural networks. We therefore investigate how to induce sparsity within AlgebraNets. We hope that our strong results on large-scale, practical benchmarks will spur further exploration of these unconventional architectures which challenge the default choice of using real numbers for neural network weights and activations.

cs.LG↗

arXiv2019

Recurrent Independent Mechanisms

Learning modular structures which reflect the dynamics of the environment can lead to better generalization and robustness to changes which only affect a few of the underlying causes. We propose Recurrent Independent Mechanisms (RIMs), a new recurrent architecture in which multiple groups of recurrent cells operate with nearly independent transition dynamics, communicate only sparingly through the bottleneck of attention, and are only updated at time steps where they are most relevant. We show that this leads to specialization amongst the RIMs, which in turn allows for dramatically improved generalization on tasks where some factors of variation differ systematically between training and evaluation.

cs.LG↗

arXiv2019

Data-Driven Approach to Encoding and Decoding 3-D Crystal Structures

Generative models have achieved impressive results in many domains including image and text generation. In the natural sciences, generative models have led to rapid progress in automated drug discovery. Many of the current methods focus on either 1-D or 2-D representations of typically small, drug-like molecules. However, many molecules require 3-D descriptors and exceed the chemical complexity of commonly used dataset. We present a method to encode and decode the position of atoms in 3-D molecules from a dataset of nearly 50,000 stable crystal unit cells that vary from containing 1 to over 100 atoms. We construct a smooth and continuous 3-D density representation of each crystal based on the positions of different atoms. Two different neural networks were trained on a dataset of over 120,000 three-dimensional samples of single and repeating crystal structures, made by rotating the single unit cells. The first, an Encoder-Decoder pair, constructs a compressed latent space representation of each molecule and then decodes this description into an accurate reconstruction of the input. The second network segments the resulting output into atoms and assigns each atom an atomic number. By generating compressed, continuous latent spaces representations of molecules we are able to decode random samples, interpolate between two molecules, and alter known molecules.

cs.LG↗

arXiv2019

InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization

This paper studies learning the representations of whole graphs in both unsupervised and semi-supervised scenarios. Graph-level representations are critical in a variety of real-world applications such as predicting the properties of molecules and community analysis in social networks. Traditional graph kernel based methods are simple, yet effective for obtaining fixed-length representations for graphs but they suffer from poor generalization due to hand-crafted designs. There are also some recent methods based on language models (e.g. graph2vec) but they tend to only consider certain substructures (e.g. subtrees) as graph representatives. Inspired by recent progress of unsupervised representation learning, in this paper we proposed a novel method called InfoGraph for learning graph-level representations. We maximize the mutual information between the graph-level representation and the representations of substructures of different scales (e.g., nodes, edges, triangles). By doing so, the graph-level representations encode aspects of the data that are shared across different scales of substructures. Furthermore, we further propose InfoGraph*, an extension of InfoGraph for semi-supervised scenarios. InfoGraph* maximizes the mutual information between unsupervised graph representations learned by InfoGraph and the representations learned by existing supervised methods. As a result, the supervised encoder learns from unlabeled data while preserving the latent semantic space favored by the current supervised task. Experimental results on the tasks of graph classification and molecular property prediction show that InfoGraph is superior to state-of-the-art baselines and InfoGraph* can achieve performance competitive with state-of-the-art semi-supervised models.

cs.LG↗

arXiv2019

vGraph: A Generative Model for Joint Community Detection and Node Representation Learning

This paper focuses on two fundamental tasks of graph analysis: community detection and node representation learning, which capture the global and local structures of graphs, respectively. In the current literature, these two tasks are usually independently studied while they are actually highly correlated. We propose a probabilistic generative model called vGraph to learn community membership and node representation collaboratively. Specifically, we assume that each node can be represented as a mixture of communities, and each community is defined as a multinomial distribution over nodes. Both the mixing coefficients and the community distribution are parameterized by the low-dimensional representations of the nodes and communities. We designed an effective variational inference algorithm which regularizes the community membership of neighboring nodes to be similar in the latent space. Experimental results on multiple real-world graphs show that vGraph is very effective in both community detection and node representation learning, outperforming many competitive baselines in both tasks. We show that the framework of vGraph is quite flexible and can be easily extended to detect hierarchical communities.

cs.SI↗

arXiv2018

Machine Learning in a data-limited regime: Augmenting experiments with synthetic data uncovers order in crumpled sheets

Machine learning has gained widespread attention as a powerful tool to identify structure in complex, high-dimensional data. However, these techniques are ostensibly inapplicable for experimental systems where data is scarce or expensive to obtain. Here we introduce a strategy to resolve this impasse by augmenting the experimental dataset with synthetically generated data of a much simpler sister system. Specifically, we study spontaneously emerging local order in crease networks of crumpled thin sheets, a paradigmatic example of spatial complexity, and show that machine learning techniques can be effective even in a data-limited regime. This is achieved by augmenting the scarce experimental dataset with inexhaustible amounts of simulated data of rigid flat-folded sheets, which are simple to simulate and share common statistical properties. This significantly improves the predictive power in a test problem of pattern completion and demonstrates the usefulness of machine learning in bench-top experiments where data is good but scarce.

cond-mat.soft↗