arXiv subjects

Matthew Johnson

Publications and source records attributed to Matthew Johnson.

At least 37 records · Page 2Linked to original sources

arXiv2022

AutoMat: Accelerated Computational Electrochemical systems Discovery

Large-scale electrification is vital to addressing the climate crisis, but several scientific and technological challenges remain to fully electrify both the chemical industry and transportation. In both of these areas, new electrochemical materials will be critical, but their development currently relies heavily on human-time-intensive experimental trial and error and computationally expensive first-principles, meso-scale and continuum simulations. We present an automated workflow, AutoMat, that accelerates these computational steps by introducing both automated input generation and management of simulations across scales from first principles to continuum device modeling. Furthermore, we show how to seamlessly integrate multi-fidelity predictions such as machine learning surrogates or automated robotic experiments "in-the-loop". The automated framework is implemented with design space search techniques to dramatically accelerate the overall materials discovery pipeline by implicitly learning design features that optimize device performance across several metrics. We discuss the benefits of AutoMat using examples in electrocatalysis and energy storage and highlight lessons learned.

cond-mat.mtrl-sci↗

arXiv2022

Snowmass2021 Cosmic Frontier: Cosmic Microwave Background Measurements White Paper

This is a solicited whitepaper for the Snowmass 2021 community planning exercise. The paper focuses on measurements and science with the Cosmic Microwave Background (CMB). The CMB is foundational to our understanding of modern physics and continues to be a powerful tool driving our understanding of cosmology and particle physics. In this paper, we outline the broad and unique impact of CMB science for the High Energy Cosmic Frontier in the upcoming decade. We also describe the progression of ground-based CMB experiments, which shows that the community is prepared to develop the key capabilities and facilities needed to achieve these transformative CMB measurements.

astro-ph.CO↗

arXiv2022

Snowmass2021 CMB-HD White Paper

CMB-HD is a proposed millimeter-wave survey over half the sky that would be ultra-deep (0.5 uK-arcmin) and have unprecedented resolution (15 arcseconds at 150 GHz). Such a survey would answer many outstanding questions about the fundamental physics of the Universe. Major advances would be 1.) the use of gravitational lensing of the primordial microwave background to map the distribution of matter on small scales (k~10 h Mpc^(-1)), which probes dark matter particle properties. It will also allow 2.) measurements of the thermal and kinetic Sunyaev-Zel'dovich effects on small scales to map the gas density and velocity, another probe of cosmic structure. In addition, CMB-HD would allow us to cross critical thresholds: 3.) ruling out or detecting any new, light (< 0.1 eV) particles that were in thermal equilibrium with known particles in the early Universe, 4.) testing a wide class of multi-field models that could explain an epoch of inflation in the early Universe, and 5.) ruling out or detecting inflationary magnetic fields. CMB-HD would also provide world-leading constraints on 6.) axion-like particles, 7.) cosmic birefringence, 8.) the sum of the neutrino masses, and 9.) the dark energy equation of state. The CMB-HD survey would be delivered in 7.5 years of observing 20,000 square degrees of sky, using two new 30-meter-class off-axis crossed Dragone telescopes to be located at Cerro Toco in the Atacama Desert. Each telescope would field 800,000 detectors (200,000 pixels), for a total of 1.6 million detectors.

astro-ph.CO↗

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↗

arXiv2022

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

Computing Subset Transversals in $H$-Free Graphs

We study the computational complexity of two well-known graph transversal problems, namely Subset Feedback Vertex Set and Subset Odd Cycle Transversal, by restricting the input to $H$-free graphs, that is, to graphs that do not contain some fixed graph~$H$ as an induced subgraph. By combining known and new results, we determine the computational complexity of both problems on $H$-free graphs for every graph $H$ except when $H=sP_1+P_4$ for some $s\geq 1$. As part of our approach, we introduce the Subset Vertex Cover problem and prove that it is polynomial-time solvable for $(sP_1+P_4)$-free graphs for every $s\geq 1$.

cs.DS↗

arXiv2021

Fake It Till You Make It: Face analysis in the wild using synthetic data alone

We demonstrate that it is possible to perform face-related computer vision in the wild using synthetic data alone. The community has long enjoyed the benefits of synthesizing training data with graphics, but the domain gap between real and synthetic data has remained a problem, especially for human faces. Researchers have tried to bridge this gap with data mixing, domain adaptation, and domain-adversarial training, but we show that it is possible to synthesize data with minimal domain gap, so that models trained on synthetic data generalize to real in-the-wild datasets. We describe how to combine a procedurally-generated parametric 3D face model with a comprehensive library of hand-crafted assets to render training images with unprecedented realism and diversity. We train machine learning systems for face-related tasks such as landmark localization and face parsing, showing that synthetic data can both match real data in accuracy as well as open up new approaches where manual labelling would be impossible.

cs.CV↗

arXiv2021

What Graphs are 2-Dot Product Graphs?

Let $d \geq 1$ be an integer. From a set of $d$-dimensional vectors, we obtain a $d$-\dpg\ by letting each vector $\va^u$ correspond to a vertex $u$ and by adding an edge between two vertices $u$ and $v$ if and only if their dot product $\va^{u} \cdot \va^{v} \geq t$, for some fixed, positive threshold~$t$. Dot product graphs can be used to model social networks. Recognizing a $d$-dot product graph is known to be \NP-hard for all fixed $d\geq 2$. To understand the position of $d$-dot product graphs in the landscape of graph classes, we consider the case $d=2$, and investigate how $2$-dot product graphs relate to a number of other known graph classes including a number of well-known classes of intersection graphs.

math.CO↗

arXiv2021

Computing Weighted Subset Transversals in $H$-Free Graphs

For the Odd Cycle Transversal problem, the task is to find a small set $S$ of vertices in a graph that intersects every cycle of odd length. The Subset Odd Cycle Transversal problem requires S to intersect only those odd cycles that include a vertex of a distinguished vertex subset $T$. If we are given weights for the vertices, we ask instead that $S$ has small weight: this is the problem Weighted Subset Odd Cycle Transversal. We prove an almost-complete complexity dichotomy for Weighted Subset Odd Cycle Transversal for graphs that do not contain a graph $H$ as an induced subgraph. Our general approach can also be used for Weighted Subset Feedback Vertex Set, which enables us to generalize a recent result of Papadopoulos and Tzimas.

cs.DS↗

arXiv2021

FastNeRF: High-Fidelity Neural Rendering at 200FPS

Recent work on Neural Radiance Fields (NeRF) showed how neural networks can be used to encode complex 3D environments that can be rendered photorealistically from novel viewpoints. Rendering these images is very computationally demanding and recent improvements are still a long way from enabling interactive rates, even on high-end hardware. Motivated by scenarios on mobile and mixed reality devices, we propose FastNeRF, the first NeRF-based system capable of rendering high fidelity photorealistic images at 200Hz on a high-end consumer GPU. The core of our method is a graphics-inspired factorization that allows for (i) compactly caching a deep radiance map at each position in space, (ii) efficiently querying that map using ray directions to estimate the pixel values in the rendered image. Extensive experiments show that the proposed method is 3000 times faster than the original NeRF algorithm and at least an order of magnitude faster than existing work on accelerating NeRF, while maintaining visual quality and extensibility.

cs.CV↗

arXiv2021

Getting to the Point. Index Sets and Parallelism-Preserving Autodiff for Pointful Array Programming

We present a novel programming language design that attempts to combine the clarity and safety of high-level functional languages with the efficiency and parallelism of low-level numerical languages. We treat arrays as eagerly-memoized functions on typed index sets, allowing abstract function manipulations, such as currying, to work on arrays. In contrast to composing primitive bulk-array operations, we argue for an explicit nested indexing style that mirrors application of functions to arguments. We also introduce a fine-grained typed effects system which affords concise and automatically-parallelized in-place updates. Specifically, an associative accumulation effect allows reverse-mode automatic differentiation of in-place updates in a way that preserves parallelism. Empirically, we benchmark against the Futhark array programming language, and demonstrate that aggressive inlining and type-driven compilation allows array programs to be written in an expressive, "pointful" style with little performance penalty.

cs.PL↗

arXiv2020

NDRIO White Paper: Envisioning Digital Research Infrastructure for the Simons Observatory

Observations of the cosmic microwave background (CMB) are an incredibly fertile source of information for studying the origins and evolution of the Universe. Canadian digital research infrastructure (DRI) has played a key role in reducing ever-larger quantities of raw data into maps of the CMB suitable for scientific analysis, as exemplified by the many scientific results produced by the Atacama Cosmology Telescope (ACT) over the past decade. The Simons Observatory (SO), due to start observing in 2023, will be able to measure the CMB with about an order of magnitude more sensitivity than ACT and other current telescopes. In this White Paper we outline how Canadian DRI under the New Digital Research Infrastructure Organization (NDRIO) could build upon the legacy of ACT and play a pivotal role in processing SO data, helping to produce data products that will be central to the cosmology community for years to come. We include estimates of DRI resources required for this work to indicate what kind of advanced research computing (ARC) would best support an SO-like project. Finally, we comment on how ARC allocations could be structured for large collaborations like SO and propose a research data management (RDM) system that makes public data releases available not only for download but also for direct analysis on Canadian DRI.

astro-ph.IM↗

arXiv2020

Steiner Trees for Hereditary Graph Classes: a Treewidth Perspective

We consider the classical problems (Edge) Steiner Tree and Vertex Steiner Tree after restricting the input to some class of graphs characterized by a small set of forbidden induced subgraphs. We show a dichotomy for the former problem restricted to $(H_1,H_2)$-free graphs and a dichotomy for the latter problem restricted to $H$-free graphs. We find that there exists an infinite family of graphs $H$ such that Vertex Steiner Tree is polynomial-time solvable for $H$-free graphs, whereas there exist only two graphs $H$ for which this holds for Edge Steiner Tree. We also find that Edge Steiner Tree is polynomial-time solvable for $(H_1,H_2)$-free graphs if and only if the treewidth of the class of $(H_1,H_2)$-free graphs is bounded (subject to P $\neq$ NP). To obtain the latter result, we determine all pairs $(H_1,H_2)$ for which the class of $(H_1,H_2)$-free graphs has bounded treewidth.

cs.DS↗

arXiv2020

CONFIG: Controllable Neural Face Image Generation

Our ability to sample realistic natural images, particularly faces, has advanced by leaps and bounds in recent years, yet our ability to exert fine-tuned control over the generative process has lagged behind. If this new technology is to find practical uses, we need to achieve a level of control over generative networks which, without sacrificing realism, is on par with that seen in computer graphics and character animation. To this end we propose ConfigNet, a neural face model that allows for controlling individual aspects of output images in semantically meaningful ways and that is a significant step on the path towards finely-controllable neural rendering. ConfigNet is trained on real face images as well as synthetic face renders. Our novel method uses synthetic data to factorize the latent space into elements that correspond to the inputs of a traditional rendering pipeline, separating aspects such as head pose, facial expression, hair style, illumination, and many others which are very hard to annotate in real data. The real images, which are presented to the network without labels, extend the variety of the generated images and encourage realism. Finally, we propose an evaluation criterion using an attribute detection network combined with a user study and demonstrate state-of-the-art individual control over attributes in the output images.

cs.CV↗

arXiv2020

A high fidelity synthetic face framework for computer vision

Analysis of faces is one of the core applications of computer vision, with tasks ranging from landmark alignment, head pose estimation, expression recognition, and face recognition among others. However, building reliable methods requires time-consuming data collection and often even more time-consuming manual annotation, which can be unreliable. In our work we propose synthesizing such facial data, including ground truth annotations that would be almost impossible to acquire through manual annotation at the consistency and scale possible through use of synthetic data. We use a parametric face model together with hand crafted assets which enable us to generate training data with unprecedented quality and diversity (varying shape, texture, expression, pose, lighting, and hair).

cs.CV↗

arXiv2020

High Resolution Zero-Shot Domain Adaptation of Synthetically Rendered Face Images

Generating photorealistic images of human faces at scale remains a prohibitively difficult task using computer graphics approaches. This is because these require the simulation of light to be photorealistic, which in turn requires physically accurate modelling of geometry, materials, and light sources, for both the head and the surrounding scene. Non-photorealistic renders however are increasingly easy to produce. In contrast to computer graphics approaches, generative models learned from more readily available 2D image data have been shown to produce samples of human faces that are hard to distinguish from real data. The process of learning usually corresponds to a loss of control over the shape and appearance of the generated images. For instance, even simple disentangling tasks such as modifying the hair independently of the face, which is trivial to accomplish in a computer graphics approach, remains an open research question. In this work, we propose an algorithm that matches a non-photorealistic, synthetically generated image to a latent vector of a pretrained StyleGAN2 model which, in turn, maps the vector to a photorealistic image of a person of the same pose, expression, hair, and lighting. In contrast to most previous work, we require no synthetic training data. To the best of our knowledge, this is the first algorithm of its kind to work at a resolution of 1K and represents a significant leap forward in visual realism.

cs.CV↗

arXiv2020

CMB-HD: Astro2020 RFI Response

CMB-HD is a proposed ultra-deep (0.5 uk-arcmin), high-resolution (15 arcseconds) millimeter-wave survey over half the sky that would answer many outstanding questions in both fundamental physics of the Universe and astrophysics. This survey would be delivered in 7.5 years of observing 20,000 square degrees, using two new 30-meter-class off-axis cross-Dragone telescopes to be located at Cerro Toco in the Atacama Desert. Each telescope would field 800,000 detectors (200,000 pixels), for a total of 1.6 million detectors.

astro-ph.CO↗

arXiv2019

On the Price of Independence for Vertex Cover, Feedback Vertex Set and Odd Cycle Transversal

Let $vc(G)$, $fvs(G)$ and $oct(G)$, respectively, denote the size of a minimum vertex cover, minimum feedback vertex set and minimum odd cycle transversal in a graph $G$. One can ask, when looking for these sets in a graph, how much bigger might they be if we require that they are independent; that is, what is the price of independence? If $G$ has a vertex cover, feedback vertex set or odd cycle transversal that is an independent set, then we let $ivc(G)$, $ifvs(G)$ or $ioct(G)$, respectively, denote the minimum size of such a set. Similar to a recent study on the price of connectivity (Hartinger et al. EuJC 2016), we investigate for which graphs $H$ the values of $ivc(G)$, $ifvs(G)$ and $ioct(G)$ are bounded in terms of $vc(G)$, $fvs(G)$ and $oct(G)$, respectively, when the graph $G$ belongs to the class of $H$-free graphs. We find complete classifications for vertex cover and feedback vertex set and an almost complete classification for odd cycle transversal (subject to three non-equivalent open cases). We also investigate for which graphs $H$ the values of $ivc(G)$, $ifvs(G)$ and $ioct(G)$ are equal to $vc(G)$, $fvs(G)$ and $oct(G)$, respectively, when the graph $G$ belongs to the class of $H$-free graphs. We find a complete classification for vertex cover and almost complete classifications for feedback vertex set (subject to one open case) and odd cycle transversal (subject to three open cases).

math.CO↗