Showing 1–50 of 66 results for author: Howard, P

SK-VQA: Synthetic Knowledge Generation at Scale for Training Context-Augmented Multimodal LLMs

Authors: Xin Su, Man Luo, Kris W Pan, Tien Pei Chou, Vasudev Lal, Phillip Howard

Abstract: Synthetic data generation has gained significant attention recently for its utility in training large vision and language models. However, the application of synthetic data to the training of multimodal context-augmented generation systems has been relatively unexplored. This gap in existing work is important because existing vision and language models (VLMs) are not trained specifically for conte… ▽ More Synthetic data generation has gained significant attention recently for its utility in training large vision and language models. However, the application of synthetic data to the training of multimodal context-augmented generation systems has been relatively unexplored. This gap in existing work is important because existing vision and language models (VLMs) are not trained specifically for context-augmented generation. Resources for adapting such models are therefore crucial for enabling their use in retrieval-augmented generation (RAG) settings, where a retriever is used to gather relevant information that is then subsequently provided to a generative model via context augmentation. To address this challenging problem, we generate SK-VQA: a large synthetic multimodal dataset containing over 2 million question-answer pairs which require external knowledge to determine the final answer. Our dataset is both larger and significantly more diverse than existing resources of its kind, possessing over 11x more unique questions and containing images from a greater variety of sources than previously-proposed datasets. Through extensive experiments, we demonstrate that our synthetic dataset can not only serve as a challenging benchmark, but is also highly effective for adapting existing generative multimodal models for context-augmented generation. △ Less

Submitted 27 June, 2024; originally announced June 2024.

Uncovering Bias in Large Vision-Language Models at Scale with Counterfactuals

Authors: Phillip Howard, Kathleen C. Fraser, Anahita Bhiwandiwalla, Svetlana Kiritchenko

Abstract: With the advent of Large Language Models (LLMs) possessing increasingly impressive capabilities, a number of Large Vision-Language Models (LVLMs) have been proposed to augment LLMs with visual inputs. Such models condition generated text on both an input image and a text prompt, enabling a variety of use cases such as visual question answering and multimodal chat. While prior studies have examined… ▽ More With the advent of Large Language Models (LLMs) possessing increasingly impressive capabilities, a number of Large Vision-Language Models (LVLMs) have been proposed to augment LLMs with visual inputs. Such models condition generated text on both an input image and a text prompt, enabling a variety of use cases such as visual question answering and multimodal chat. While prior studies have examined the social biases contained in text generated by LLMs, this topic has been relatively unexplored in LVLMs. Examining social biases in LVLMs is particularly challenging due to the confounding contributions of bias induced by information contained across the text and visual modalities. To address this challenging problem, we conduct a large-scale study of text generated by different LVLMs under counterfactual changes to input images. Specifically, we present LVLMs with identical open-ended text prompts while conditioning on images from different counterfactual sets, where each set contains images which are largely identical in their depiction of a common subject (e.g., a doctor), but vary only in terms of intersectional social attributes (e.g., race and gender). We comprehensively evaluate the text produced by different models under this counterfactual generation setting at scale, producing over 57 million responses from popular LVLMs. Our multi-dimensional analysis reveals that social attributes such as race, gender, and physical characteristics depicted in input images can significantly influence the generation of toxic content, competency-associated words, harmful stereotypes, and numerical ratings of depicted individuals. We additionally explore the relationship between social bias in LVLMs and their corresponding LLMs, as well as inference-time strategies to mitigate bias. △ Less

Submitted 30 May, 2024; originally announced May 2024.

Comparison of Coarsening Dynamics for the Cahn--Hilliard and Burgers--Cahn--Hilliard Equations

Authors: Peter Howard, Adam Larios, Quyuan Lin

Abstract: We consider coarsening dynamics associated with a Burgers--Cahn--Hilliard system modeling a two-phase flow in one space dimension. Our emphasis is on the effect that coupling between the phase and fluid dynamics has on coarsening rates, and on the mechanisms driving this effect. We start with a detailed examination of coarsening dynamics for the uncoupled Cahn--Hilliard equation, comparing numeric… ▽ More We consider coarsening dynamics associated with a Burgers--Cahn--Hilliard system modeling a two-phase flow in one space dimension. Our emphasis is on the effect that coupling between the phase and fluid dynamics has on coarsening rates, and on the mechanisms driving this effect. We start with a detailed examination of coarsening dynamics for the uncoupled Cahn--Hilliard equation, comparing numerically generated rates with two analytic methods, and then we consider how these dynamics are affected by appropriate coupling with a viscous Burgers equation. In order to keep the analysis as self-contained as possible, we establish the global well-posedness of the system under consideration. △ Less

Submitted 19 May, 2024; originally announced May 2024.

Comments: 55 pages, 14 figures

Maximal δ-separated sets in separable metric spaces and weak forms of choice

Authors: Michał Dybowski, Przemyslaw Górka, Paul Howard

Abstract: We show that the statement ``In every separable pseudometric space there is a maximal non-strictly δ-separated set.'' implies the axiom of choice for countable families of sets. This gives answers to a question of Dybowski and Górka in [M. Dybowski and P. Górka, The axiom of choice in metric measure spaces and maximal δ-separated sets, Archive for Mathematical Logic 62, 735-749, 2023.]. We also pr… ▽ More We show that the statement ``In every separable pseudometric space there is a maximal non-strictly δ-separated set.'' implies the axiom of choice for countable families of sets. This gives answers to a question of Dybowski and Górka in [M. Dybowski and P. Górka, The axiom of choice in metric measure spaces and maximal δ-separated sets, Archive for Mathematical Logic 62, 735-749, 2023.]. We also prove several related results. △ Less

Submitted 16 May, 2024; originally announced May 2024.

MSC Class: 03E25 (Primary) 03E35; 30L99 (Secondary)

Uncovering Bias in Large Vision-Language Models with Counterfactuals

Authors: Phillip Howard, Anahita Bhiwandiwalla, Kathleen C. Fraser, Svetlana Kiritchenko

Abstract: With the advent of Large Language Models (LLMs) possessing increasingly impressive capabilities, a number of Large Vision-Language Models (LVLMs) have been proposed to augment LLMs with visual inputs. Such models condition generated text on both an input image and a text prompt, enabling a variety of use cases such as visual question answering and multimodal chat. While prior studies have examined… ▽ More With the advent of Large Language Models (LLMs) possessing increasingly impressive capabilities, a number of Large Vision-Language Models (LVLMs) have been proposed to augment LLMs with visual inputs. Such models condition generated text on both an input image and a text prompt, enabling a variety of use cases such as visual question answering and multimodal chat. While prior studies have examined the social biases contained in text generated by LLMs, this topic has been relatively unexplored in LVLMs. Examining social biases in LVLMs is particularly challenging due to the confounding contributions of bias induced by information contained across the text and visual modalities. To address this challenging problem, we conduct a large-scale study of text generated by different LVLMs under counterfactual changes to input images. Specifically, we present LVLMs with identical open-ended text prompts while conditioning on images from different counterfactual sets, where each set contains images which are largely identical in their depiction of a common subject (e.g., a doctor), but vary only in terms of intersectional social attributes (e.g., race and gender). We comprehensively evaluate the text produced by different LVLMs under this counterfactual generation setting and find that social attributes such as race, gender, and physical characteristics depicted in input images can significantly influence toxicity and the generation of competency-associated words. △ Less

Submitted 7 June, 2024; v1 submitted 29 March, 2024; originally announced April 2024.

Comments: Accepted to the CVPR 2024 Responsible Generative AI (ReGenAI) Workshop

Mesoscale simulations of diffusion and sedimentation in shape-anisotropic nanoparticle suspensions

Authors: Yashraj M. Wani, Penelope Grace Kovakas, Arash Nikoubashman, Michael P. Howard

Abstract: We determine the long-time self-diffusion coefficient and sedimentation coefficient for suspensions of nanoparticles with anisotropic shapes (octahedra, cubes, tetrahedra, and spherocylinders) as a function of nanoparticle concentration using mesoscale simulations. We use a discrete particle model for the nanoparticles, and we account for solvent-mediated hydrodynamic interactions between nanopart… ▽ More We determine the long-time self-diffusion coefficient and sedimentation coefficient for suspensions of nanoparticles with anisotropic shapes (octahedra, cubes, tetrahedra, and spherocylinders) as a function of nanoparticle concentration using mesoscale simulations. We use a discrete particle model for the nanoparticles, and we account for solvent-mediated hydrodynamic interactions between nanoparticles using the multiparticle collision dynamics method. Our simulations are compared to theoretical predictions and experimental data from existing literature, demonstrating good agreement in the majority of cases. Further, we find that the self-diffusion coefficient of the regular polyhedral shapes can be estimated from that of a sphere whose diameter is average of their inscribed and circumscribed sphere diameters. △ Less

Submitted 5 March, 2024; originally announced March 2024.

Comments: 13 pages, 7 figures

SocialCounterfactuals: Probing and Mitigating Intersectional Social Biases in Vision-Language Models with Counterfactual Examples

Authors: Phillip Howard, Avinash Madasu, Tiep Le, Gustavo Lujan Moreno, Anahita Bhiwandiwalla, Vasudev Lal

Abstract: While vision-language models (VLMs) have achieved remarkable performance improvements recently, there is growing evidence that these models also posses harmful biases with respect to social attributes such as gender and race. Prior studies have primarily focused on probing such bias attributes individually while ignoring biases associated with intersections between social attributes. This could be… ▽ More While vision-language models (VLMs) have achieved remarkable performance improvements recently, there is growing evidence that these models also posses harmful biases with respect to social attributes such as gender and race. Prior studies have primarily focused on probing such bias attributes individually while ignoring biases associated with intersections between social attributes. This could be due to the difficulty of collecting an exhaustive set of image-text pairs for various combinations of social attributes. To address this challenge, we employ text-to-image diffusion models to produce counterfactual examples for probing intersectional social biases at scale. Our approach utilizes Stable Diffusion with cross attention control to produce sets of counterfactual image-text pairs that are highly similar in their depiction of a subject (e.g., a given occupation) while differing only in their depiction of intersectional social attributes (e.g., race & gender). Through our over-generate-then-filter methodology, we produce SocialCounterfactuals, a high-quality dataset containing 171k image-text pairs for probing intersectional biases related to gender, race, and physical characteristics. We conduct extensive experiments to demonstrate the usefulness of our generated dataset for probing and mitigating intersectional social biases in state-of-the-art VLMs. △ Less

Submitted 9 April, 2024; v1 submitted 30 November, 2023; originally announced December 2023.

Comments: Accepted to CVPR 2024. arXiv admin note: text overlap with arXiv:2310.02988

NeuroPrompts: An Adaptive Framework to Optimize Prompts for Text-to-Image Generation

Authors: Shachar Rosenman, Vasudev Lal, Phillip Howard

Abstract: Despite impressive recent advances in text-to-image diffusion models, obtaining high-quality images often requires prompt engineering by humans who have developed expertise in using them. In this work, we present NeuroPrompts, an adaptive framework that automatically enhances a user's prompt to improve the quality of generations produced by text-to-image models. Our framework utilizes constrained… ▽ More Despite impressive recent advances in text-to-image diffusion models, obtaining high-quality images often requires prompt engineering by humans who have developed expertise in using them. In this work, we present NeuroPrompts, an adaptive framework that automatically enhances a user's prompt to improve the quality of generations produced by text-to-image models. Our framework utilizes constrained text decoding with a pre-trained language model that has been adapted to generate prompts similar to those produced by human prompt engineers. This approach enables higher-quality text-to-image generations and provides user control over stylistic features via constraint set specification. We demonstrate the utility of our framework by creating an interactive application for prompt enhancement and image generation using Stable Diffusion. Additionally, we conduct experiments utilizing a large dataset of human-engineered prompts for text-to-image generation and show that our approach automatically produces enhanced prompts that result in superior image quality. We make our code and a screencast video demo of NeuroPrompts publicly available. △ Less

Submitted 5 April, 2024; v1 submitted 20 November, 2023; originally announced November 2023.

Comments: Accepted to EACL 2024 System Demonstration Track

Semi-Structured Chain-of-Thought: Integrating Multiple Sources of Knowledge for Improved Language Model Reasoning

Authors: Xin Su, Tiep Le, Steven Bethard, Phillip Howard

Abstract: An important open question in the use of large language models for knowledge-intensive tasks is how to effectively integrate knowledge from three sources: the model's parametric memory, external structured knowledge, and external unstructured knowledge. Most existing prompting methods either rely on one or two of these sources, or require repeatedly invoking large language models to generate simil… ▽ More An important open question in the use of large language models for knowledge-intensive tasks is how to effectively integrate knowledge from three sources: the model's parametric memory, external structured knowledge, and external unstructured knowledge. Most existing prompting methods either rely on one or two of these sources, or require repeatedly invoking large language models to generate similar or identical content. In this work, we overcome these limitations by introducing a novel semi-structured prompting approach that seamlessly integrates the model's parametric memory with unstructured knowledge from text documents and structured knowledge from knowledge graphs. Experimental results on open-domain multi-hop question answering datasets demonstrate that our prompting method significantly surpasses existing techniques, even exceeding those that require fine-tuning. △ Less

Submitted 1 April, 2024; v1 submitted 14 November, 2023; originally announced November 2023.

Comments: NAACL 2024 main conference

Fusing Temporal Graphs into Transformers for Time-Sensitive Question Answering

Authors: Xin Su, Phillip Howard, Nagib Hakim, Steven Bethard

Abstract: Answering time-sensitive questions from long documents requires temporal reasoning over the times in questions and documents. An important open question is whether large language models can perform such reasoning solely using a provided text document, or whether they can benefit from additional temporal information extracted using other systems. We address this research question by applying existi… ▽ More Answering time-sensitive questions from long documents requires temporal reasoning over the times in questions and documents. An important open question is whether large language models can perform such reasoning solely using a provided text document, or whether they can benefit from additional temporal information extracted using other systems. We address this research question by applying existing temporal information extraction systems to construct temporal graphs of events, times, and temporal relations in questions and documents. We then investigate different approaches for fusing these graphs into Transformer models. Experimental results show that our proposed approach for fusing temporal graphs into input text substantially enhances the temporal reasoning capabilities of Transformer models with or without fine-tuning. Additionally, our proposed method outperforms various graph convolution-based approaches and establishes a new state-of-the-art performance on SituatedQA and three splits of TimeQA. △ Less

Submitted 30 October, 2023; originally announced October 2023.

Comments: EMNLP 2023 Findings

Probing Intersectional Biases in Vision-Language Models with Counterfactual Examples

Authors: Phillip Howard, Avinash Madasu, Tiep Le, Gustavo Lujan Moreno, Vasudev Lal

Abstract: While vision-language models (VLMs) have achieved remarkable performance improvements recently, there is growing evidence that these models also posses harmful biases with respect to social attributes such as gender and race. Prior studies have primarily focused on probing such bias attributes individually while ignoring biases associated with intersections between social attributes. This could be… ▽ More While vision-language models (VLMs) have achieved remarkable performance improvements recently, there is growing evidence that these models also posses harmful biases with respect to social attributes such as gender and race. Prior studies have primarily focused on probing such bias attributes individually while ignoring biases associated with intersections between social attributes. This could be due to the difficulty of collecting an exhaustive set of image-text pairs for various combinations of social attributes from existing datasets. To address this challenge, we employ text-to-image diffusion models to produce counterfactual examples for probing intserctional social biases at scale. Our approach utilizes Stable Diffusion with cross attention control to produce sets of counterfactual image-text pairs that are highly similar in their depiction of a subject (e.g., a given occupation) while differing only in their depiction of intersectional social attributes (e.g., race & gender). We conduct extensive experiments using our generated dataset which reveal the intersectional social biases present in state-of-the-art VLMs. △ Less

Submitted 4 October, 2023; originally announced October 2023.

COCO-Counterfactuals: Automatically Constructed Counterfactual Examples for Image-Text Pairs

Authors: Tiep Le, Vasudev Lal, Phillip Howard

Abstract: Counterfactual examples have proven to be valuable in the field of natural language processing (NLP) for both evaluating and improving the robustness of language models to spurious correlations in datasets. Despite their demonstrated utility for NLP, multimodal counterfactual examples have been relatively unexplored due to the difficulty of creating paired image-text data with minimal counterfactu… ▽ More Counterfactual examples have proven to be valuable in the field of natural language processing (NLP) for both evaluating and improving the robustness of language models to spurious correlations in datasets. Despite their demonstrated utility for NLP, multimodal counterfactual examples have been relatively unexplored due to the difficulty of creating paired image-text data with minimal counterfactual changes. To address this challenge, we introduce a scalable framework for automatic generation of counterfactual examples using text-to-image diffusion models. We use our framework to create COCO-Counterfactuals, a multimodal counterfactual dataset of paired image and text captions based on the MS-COCO dataset. We validate the quality of COCO-Counterfactuals through human evaluations and show that existing multimodal models are challenged by our counterfactual image-text pairs. Additionally, we demonstrate the usefulness of COCO-Counterfactuals for improving out-of-domain generalization of multimodal vision-language models via training data augmentation. △ Less

Submitted 31 October, 2023; v1 submitted 22 September, 2023; originally announced September 2023.

Comments: Accepted to NeurIPS 2023 Datasets and Benchmarks Track

NeuroComparatives: Neuro-Symbolic Distillation of Comparative Knowledge

Authors: Phillip Howard, Junlin Wang, Vasudev Lal, Gadi Singer, Ye** Choi, Swabha Swayamdipta

Abstract: Comparative knowledge (e.g., steel is stronger and heavier than styrofoam) is an essential component of our world knowledge, yet understudied in prior literature. In this paper, we harvest the dramatic improvements in knowledge capabilities of language models into a large-scale comparative knowledge base. While the ease of acquisition of such comparative knowledge is much higher from extreme-scale… ▽ More Comparative knowledge (e.g., steel is stronger and heavier than styrofoam) is an essential component of our world knowledge, yet understudied in prior literature. In this paper, we harvest the dramatic improvements in knowledge capabilities of language models into a large-scale comparative knowledge base. While the ease of acquisition of such comparative knowledge is much higher from extreme-scale models like GPT-4, compared to their considerably smaller and weaker counterparts such as GPT-2, not even the most powerful models are exempt from making errors. We thus ask: to what extent are models at different scales able to generate valid and diverse comparative knowledge? We introduce NeuroComparatives, a novel framework for comparative knowledge distillation overgenerated from language models such as GPT-variants and LLaMA, followed by stringent filtering of the generated knowledge. Our framework acquires comparative knowledge between everyday objects, producing a corpus of up to 8.8M comparisons over 1.74M entity pairs - 10X larger and 30% more diverse than existing resources. Moreover, human evaluations show that NeuroComparatives outperform existing resources in terms of validity (up to 32% absolute improvement). Our acquired NeuroComparatives leads to performance improvements on five downstream tasks. We find that neuro-symbolic manipulation of smaller models offers complementary benefits to the currently dominant practice of prompting extreme-scale language models for knowledge distillation. △ Less

Submitted 5 April, 2024; v1 submitted 8 May, 2023; originally announced May 2023.

Comments: Accepted to NAACL 2024 Findings

Thrill-K Architecture: Towards a Solution to the Problem of Knowledge Based Understanding

Authors: Gadi Singer, Joscha Bach, Tetiana Grinberg, Nagib Hakim, Phillip Howard, Vasudev Lal, Zev Rivlin

Abstract: While end-to-end learning systems are rapidly gaining capabilities and popularity, the increasing computational demands for deploying such systems, along with a lack of flexibility, adaptability, explainability, reasoning and verification capabilities, require new types of architectures. Here we introduce a classification of hybrid systems which, based on an analysis of human knowledge and intelli… ▽ More While end-to-end learning systems are rapidly gaining capabilities and popularity, the increasing computational demands for deploying such systems, along with a lack of flexibility, adaptability, explainability, reasoning and verification capabilities, require new types of architectures. Here we introduce a classification of hybrid systems which, based on an analysis of human knowledge and intelligence, combines neural learning with various types of knowledge and knowledge sources. We present the Thrill-K architecture as a prototypical solution for integrating instantaneous knowledge, standby knowledge and external knowledge sources in a framework capable of inference, learning and intelligent control. △ Less

Submitted 28 February, 2023; originally announced March 2023.

Comments: Artificial General Intelligence: 15th International Conference, AGI 2022, Seattle, WA, USA, August 2022, Proceedings

Journal ref: Springer Lecture Notes in Computer Science, vol 13539, 2023

Oscillation Theory and Instability of Nonlinear Waves

Authors: Peter Howard

Abstract: In recent work, Baird et al. have introduced a generalized Maslov index which allows oscillation techniques that have previously been restricted to eigenvalue problems with underlying Hamiltonian structure to be extended to the non-Hamiltonian setting [T. J. Baird, P. Cornwell, G. Cox, C. Jones, and R. Marangell, Generalized Maslov indices for non-Hamiltonian systems, SIAM J. Math. Anal. 54 (2022)… ▽ More In recent work, Baird et al. have introduced a generalized Maslov index which allows oscillation techniques that have previously been restricted to eigenvalue problems with underlying Hamiltonian structure to be extended to the non-Hamiltonian setting [T. J. Baird, P. Cornwell, G. Cox, C. Jones, and R. Marangell, Generalized Maslov indices for non-Hamiltonian systems, SIAM J. Math. Anal. 54 (2022) 1623-1668]. We show that this approach can be implemented in the analysis of spectral instability for nonlinear waves, taking as our setting a class of equations previously investigated by Pego and Weinstein via the Evans function [R. L. Pego and M. I. Weinstein, Eigenvalues, and instabilities of solitary waves, Phil. Trans. R. Soc. Lond. A 340 (1992) 47-94]. △ Less

Submitted 17 January, 2023; originally announced January 2023.

MSC Class: 34C10

A Herschel study of G214.5-1.8: a young, cold and quiescent giant molecular filament on the shell of a HI superbubble

Authors: S. D. Clarke, A. Sanchez-Monge, G. M. Williams, A. D. P. Howard, S. Walch, N. Schneider

Abstract: We present an analysis of the outer Galaxy giant molecular filament (GMF) G214.5-1.8 (G214.5) using Herschel data. We find that G214.5 has a mass of $\sim$ 16,000 M$_{\odot}$, yet hosts only 15 potentially protostellar 70 $μ$m sources, making it highly quiescent compared to equally massive clouds such as Serpens and Mon R2. We show that G214.5 has a unique morphology, consisting of a narrow `Main… ▽ More We present an analysis of the outer Galaxy giant molecular filament (GMF) G214.5-1.8 (G214.5) using Herschel data. We find that G214.5 has a mass of $\sim$ 16,000 M$_{\odot}$, yet hosts only 15 potentially protostellar 70 $μ$m sources, making it highly quiescent compared to equally massive clouds such as Serpens and Mon R2. We show that G214.5 has a unique morphology, consisting of a narrow `Main filament' running north-south and a perpendicular `Head' structure running east-west. We identify 33 distinct massive clumps from the column density maps, 8 of which are protostellar. However, the star formation activity is not evenly spread across G214.5 but rather predominantly located in the Main filament. Studying the Main filament in a manner similar to previous works, we find that G214.5 is most like a 'Bone' candidate GMF, highly elongated and massive, but it is colder and narrower than any such GMF. It also differs significantly due to its low fraction of high column density gas. Studying the radial profile, we discover that G214.5 is highly asymmetric and resembles filaments which are known to be compressed externally. Considering its environment, we find that G214.5 is co-incident, spatially and kinematically, with a HI superbubble. We discuss how a potential interaction between G214.5 and the superbubble may explain G214.5's morphology, asymmetry and, paucity of dense gas and star formation activity, highlighting the intersection of a bubble-driven interstellar medium paradigm with that of a filament paradigm for star formation. △ Less

Submitted 4 November, 2022; originally announced November 2022.

Comments: Accepted in MNRAS. 26 pages, 24 figures

NeuroCounterfactuals: Beyond Minimal-Edit Counterfactuals for Richer Data Augmentation

Authors: Phillip Howard, Gadi Singer, Vasudev Lal, Ye** Choi, Swabha Swayamdipta

Abstract: While counterfactual data augmentation offers a promising step towards robust generalization in natural language processing, producing a set of counterfactuals that offer valuable inductive bias for models remains a challenge. Most existing approaches for producing counterfactuals, manual or automated, rely on small perturbations via minimal edits, resulting in simplistic changes. We introduce Neu… ▽ More While counterfactual data augmentation offers a promising step towards robust generalization in natural language processing, producing a set of counterfactuals that offer valuable inductive bias for models remains a challenge. Most existing approaches for producing counterfactuals, manual or automated, rely on small perturbations via minimal edits, resulting in simplistic changes. We introduce NeuroCounterfactuals, designed as loose counterfactuals, allowing for larger edits which result in naturalistic generations containing linguistic diversity, while still bearing similarity to the original document. Our novel generative approach bridges the benefits of constrained decoding, with those of language model adaptation for sentiment steering. Training data augmentation with our generations results in both in-domain and out-of-domain improvements for sentiment classification, outperforming even manually curated counterfactuals, under select settings. We further present detailed analyses to show the advantages of NeuroCounterfactuals over approaches involving simple, minimal edits. △ Less

Submitted 22 October, 2022; originally announced October 2022.

Comments: Findings of EMNLP 2022

Cross-Domain Aspect Extraction using Transformers Augmented with Knowledge Graphs

Authors: Phillip Howard, Arden Ma, Vasudev Lal, Ana Paula Simoes, Daniel Korat, Oren Pereg, Moshe Wasserblat, Gadi Singer

Abstract: The extraction of aspect terms is a critical step in fine-grained sentiment analysis of text. Existing approaches for this task have yielded impressive results when the training and testing data are from the same domain. However, these methods show a drastic decrease in performance when applied to cross-domain settings where the domain of the testing data differs from that of the training data. To… ▽ More The extraction of aspect terms is a critical step in fine-grained sentiment analysis of text. Existing approaches for this task have yielded impressive results when the training and testing data are from the same domain. However, these methods show a drastic decrease in performance when applied to cross-domain settings where the domain of the testing data differs from that of the training data. To address this lack of extensibility and robustness, we propose a novel approach for automatically constructing domain-specific knowledge graphs that contain information relevant to the identification of aspect terms. We introduce a methodology for injecting information from these knowledge graphs into Transformer models, including two alternative mechanisms for knowledge insertion: via query enrichment and via manipulation of attention patterns. We demonstrate state-of-the-art performance on benchmark datasets for cross-domain aspect term extraction using our approach and investigate how the amount of external knowledge available to the Transformer impacts model performance. △ Less

Submitted 18 October, 2022; originally announced October 2022.

ACM Class: I.2.7

Journal ref: Proceedings of the 31st ACM International Conference on Information & Knowledge Management (CIKM 2022). Association for Computing Machinery, New York, NY, USA, 780-790

Multiscale modeling of solute diffusion in triblock copolymer membranes

Authors: Anthony J. Cooper, Michael P. Howard, Sanket Kadulkar, David Zhao, Kris T. Delaney, Venkat Ganesan, Thomas M. Truskett, Glenn H. Fredrickson

Abstract: We develop a multiscale simulation model for diffusion of solutes through porous triblock copolymer membranes. The approach combines two techniques: self-consistent field theory (SCFT) to predict the structure of the self-assembled, solvated membrane and on-lattice kinetic Monte Carlo (kMC) simulations to model diffusion of solutes. Solvation is simulated in SCFT by constraining the glassy membran… ▽ More We develop a multiscale simulation model for diffusion of solutes through porous triblock copolymer membranes. The approach combines two techniques: self-consistent field theory (SCFT) to predict the structure of the self-assembled, solvated membrane and on-lattice kinetic Monte Carlo (kMC) simulations to model diffusion of solutes. Solvation is simulated in SCFT by constraining the glassy membrane matrix while relaxing the brush-like membrane pore coating against the solvent. The kMC simulations capture the resulting solute spatial distribution and concentration-dependent local diffusivity in the polymer-coated pores; we parameterize the latter using particle-based simulations. We apply our approach to simulate solute diffusion through nonequilibrium morphologies of a model triblock copolymer, and we correlate diffusivity with structural descriptors of the morphologies. We also compare the model's predictions to alternative approaches based on simple lattice random walks and find our multiscale model to be more robust and systematic to parameterize. Our multiscale modeling approach is general and can be readily extended in the future to other chemistries, morphologies, and models for the local solute diffusivity and interactions with the membrane. △ Less

Submitted 20 September, 2022; originally announced September 2022.

Journal ref: J. Chem. Phys. 158, 024905 (2023)

Dynamic density functional theory for drying colloidal suspensions: Comparison of hard-sphere free-energy functionals

Authors: Mayukh Kundu, Michael P. Howard

Abstract: Dynamic density functional theory (DDFT) is a promising approach for predicting the structural evolution of a drying suspension containing one or more types of colloidal particles. The assumed free-energy functional is a key component of DDFT that dictates the thermodynamics of the model and, in turn, the density flux due to a concentration gradient. In this work, we compare several commonly used… ▽ More Dynamic density functional theory (DDFT) is a promising approach for predicting the structural evolution of a drying suspension containing one or more types of colloidal particles. The assumed free-energy functional is a key component of DDFT that dictates the thermodynamics of the model and, in turn, the density flux due to a concentration gradient. In this work, we compare several commonly used free-energy functionals for drying hard-sphere suspensions including local-density approximations based on the ideal-gas, virial, and Boublík-Mansoori-Carnahan-Starling-Leland (BMCSL) equations of state as well as a weighted-density approximation based on fundamental measure theory (FMT). To determine the accuracy of each functional, we model one- and two-component hard-sphere suspensions in a drying film with varied initial heights and compositions, and we compare the DDFT-predicted volume-fraction profiles to particle-based Brownian dynamics (BD) simulations. FMT accurately predicts the structure of the one-component suspensions even at high concentrations and when significant density gradients develop, but the virial and BMCSL equations of state provide reasonable approximations for smaller concentrations at a reduced computational cost. In the two-component suspensions, FMT and BMCSL are similar to each other but modestly overpredict the extent of stratification by size compared to BD simulations. This work provides helpful guidance for selecting thermodynamic models for soft materials in nonequilibrium processes such as solvent drying, solvent freezing, and sedimentation. △ Less

Submitted 2 August, 2022; originally announced August 2022.

Renormalized oscillation theory for regular linear non-Hamiltonian systems

Authors: Peter Howard

Abstract: In recent work, Baird et al. have generalized the definition of the Maslov index to paths of Grassmannian subspaces that are not necessarily contained in the Lagrangian Grassmannian [T. J. Baird, P. Cornwell, G. Cox, C. Jones, and R. Marangell, {\it Generalized Maslov indices for non-Hamiltonian systems}, SIAM J. Math. Anal. {\bf 54} (2022) 1623-1668]. Such an extension opens up the possibility of… ▽ More In recent work, Baird et al. have generalized the definition of the Maslov index to paths of Grassmannian subspaces that are not necessarily contained in the Lagrangian Grassmannian [T. J. Baird, P. Cornwell, G. Cox, C. Jones, and R. Marangell, {\it Generalized Maslov indices for non-Hamiltonian systems}, SIAM J. Math. Anal. {\bf 54} (2022) 1623-1668]. Such an extension opens up the possibility of applications to non-Hamiltonian systems of ODE, and Baird and his collaborators have taken advantage of this observation to establish oscillation-type results for obtaining lower bounds on eigenvalue counts in this generalized setting. In the current analysis, the author shows that renormalized oscillation theory, appropriately defined in this generalized setting, can be applied in a natural way, and that it has the advantage, as in the traditional setting of linear Hamiltonian systems, of ensuring monotonicity of crossing points as the independent variable increases for a wide range of system/boundary-condition combinations. This seems to mark the first effort to extend the renormalized oscillation approach to the non-Hamiltonian setting. △ Less

Submitted 11 May, 2022; originally announced May 2022.

MSC Class: 34C10

TempoQR: Temporal Question Reasoning over Knowledge Graphs

Authors: Costas Mavromatis, Prasanna Lakkur Subramanyam, Vassilis N. Ioannidis, Soji Adeshina, Phillip R. Howard, Tetiana Grinberg, Nagib Hakim, George Karypis

Abstract: Knowledge Graph Question Answering (KGQA) involves retrieving facts from a Knowledge Graph (KG) using natural language queries. A KG is a curated set of facts consisting of entities linked by relations. Certain facts include also temporal information forming a Temporal KG (TKG). Although many natural questions involve explicit or implicit time constraints, question answering (QA) over TKGs has bee… ▽ More Knowledge Graph Question Answering (KGQA) involves retrieving facts from a Knowledge Graph (KG) using natural language queries. A KG is a curated set of facts consisting of entities linked by relations. Certain facts include also temporal information forming a Temporal KG (TKG). Although many natural questions involve explicit or implicit time constraints, question answering (QA) over TKGs has been a relatively unexplored area. Existing solutions are mainly designed for simple temporal questions that can be answered directly by a single TKG fact. This paper puts forth a comprehensive embedding-based framework for answering complex questions over TKGs. Our method termed temporal question reasoning (TempoQR) exploits TKG embeddings to ground the question to the specific entities and time scope it refers to. It does so by augmenting the question embeddings with context, entity and time-aware information by employing three specialized modules. The first computes a textual representation of a given question, the second combines it with the entity embeddings for entities involved in the question, and the third generates question-specific time embeddings. Finally, a transformer-based encoder learns to fuse the generated temporal information with the question representation, which is used for answer predictions. Extensive experiments show that TempoQR improves accuracy by 25--45 percentage points on complex temporal questions over state-of-the-art approaches and it generalizes better to unseen question types. △ Less

Submitted 10 December, 2021; originally announced December 2021.

Comments: AAAI 2022

Diffusion and sedimentation in colloidal suspensions using multiparticle collision dynamics with a discrete particle model

Authors: Yashraj M. Wani, Penelope Grace Kovakas, Arash Nikoubashman, Michael P. Howard

Abstract: We study self-diffusion and sedimentation in colloidal suspensions of nearly-hard spheres using the multiparticle collision dynamics simulation method for the solvent with a discrete mesh model for the colloidal particles (MD+MPCD). We cover colloid volume fractions from 0.01 to 0.40 and compare the MD+MPCD simulations to Brownian dynamics simulations with free-draining hydrodynamics (BD) as well… ▽ More We study self-diffusion and sedimentation in colloidal suspensions of nearly-hard spheres using the multiparticle collision dynamics simulation method for the solvent with a discrete mesh model for the colloidal particles (MD+MPCD). We cover colloid volume fractions from 0.01 to 0.40 and compare the MD+MPCD simulations to Brownian dynamics simulations with free-draining hydrodynamics (BD) as well as pairwise far-field hydrodynamics described using the Rotne--Prager--Yamakawa mobility tensor (BD+RPY). The dynamics in MD+MPCD suggest that the colloidal particles are only partially coupled to the solvent at short times. However, the long-time self-diffusion coefficient in MD+MPCD is comparable to that in BD and BD+RPY, and the sedimentation coefficient in MD+MPCD is in good agreement with that in BD+RPY, suggesting that MD+MPCD gives a reasonable description of the hydrodynamic interactions in colloidal suspensions. The discrete-particle MD+MPCD approach is convenient and readily extended to more complex shapes, and we determine the long-time self-diffusion coefficient in suspensions of nearly-hard cubes to demonstrate its generality. △ Less

Submitted 12 October, 2021; originally announced October 2021.

Comments: 13 pages, 8 figures

A PPMAP analysis of the filamentary structures in Ophiuchus L1688 and L1689

Authors: A. D. P. Howard, A. P. Whitworth, M. J. Griffin, K. A. Marsh, M. W. L. Smith

Abstract: We use the PPMAP (Point Process MAP**) algorithm to re-analyse the \textit{Herschel} and SCUBA-2 observations of the L1688 and L1689 sub-regions of the Ophiuchus molecular cloud. PPMAP delivers maps with high resolution (here $14''$, corresponding to $\sim 0.01\,{\rm pc}$ at $\sim 140\,{\rm pc}$), by using the observations at their native resolutions. PPMAP also delivers more accurate dust optic… ▽ More We use the PPMAP (Point Process MAP**) algorithm to re-analyse the \textit{Herschel} and SCUBA-2 observations of the L1688 and L1689 sub-regions of the Ophiuchus molecular cloud. PPMAP delivers maps with high resolution (here $14''$, corresponding to $\sim 0.01\,{\rm pc}$ at $\sim 140\,{\rm pc}$), by using the observations at their native resolutions. PPMAP also delivers more accurate dust optical depths, by distinguishing dust of different types and at different temperatures. The filaments and prestellar cores almost all lie in regions with $N_{\rm H_2}\gtrsim 7\times 10^{21}\,{\rm cm}^{-2}$ (corresponding to $A_{_{\rm V}}\gtrsim 7$). The dust temperature, $T$, tends to be correlated with the dust opacity index, $β$, with low $T$ and low $β$ tend concentrated in the interiors of filaments. The one exception to this tendency is a section of filament in L1688 that falls -- in projection -- between the two B stars, S1 and HD147889; here $T$ and $β$ are relatively high, and there is compelling evidence that feedback from these two stars has heated and compressed the filament. Filament {\sc fwhm}s are typically in the range $0.10\,{\rm pc}$ to $0.15\,{\rm pc}$. Most filaments have line densities in the range $25\,{\rm M_{_\odot}\,pc^{-1}}$ to $65\,{\rm M_{_\odot}\,pc^{-1}}$. If their only support is thermal gas pressure, and the gas is at the canonical temperature of $10\,{\rm K}$, the filaments are highly supercritical. However, there is some evidence from ammonia observations that the gas is significantly warmer than this, and we cannot rule out the possibility of additional support from turbulence and/or magnetic fields. On the basis of their spatial distribution, we argue that most of the starless cores are likely to disperse (rather than evolving to become prestellar). △ Less

Submitted 8 April, 2021; originally announced April 2021.

Comments: 22 pages, 18 figures, Accepted for publication by MNRAS March 2021

Effects of linker flexibility on phase behavior and structure of linked colloidal gels

Authors: Michael P. Howard, Zachary M. Sherman, Adithya N Sreenivasan, Stephanie A. Valenzuela, Eric V. Anslyn, Delia J. Milliron, Thomas M. Truskett

Abstract: Colloidal nanocrystal gels can be assembled using a difunctional "linker" molecule to mediate bonding between nanocrystals. The conditions for gelation and the structure of the gel are controlled macroscopically by the linker concentration and microscopically by the linker's molecular characteristics. Here, we demonstrate using a toy model for a colloid-linker mixture that linker flexibility plays… ▽ More Colloidal nanocrystal gels can be assembled using a difunctional "linker" molecule to mediate bonding between nanocrystals. The conditions for gelation and the structure of the gel are controlled macroscopically by the linker concentration and microscopically by the linker's molecular characteristics. Here, we demonstrate using a toy model for a colloid-linker mixture that linker flexibility plays a key role in determining both phase behavior and structure of the mixture. We fix the linker length and systematically vary its bending stiffness to span the flexible, semiflexible, and rigid regimes. At fixed linker concentration, flexible-linker and rigid-linker mixtures phase separate at low colloid volume fractions in agreement with predictions of first-order thermodynamic perturbation theory, but the semiflexible-linker mixtures do not. We correlate and attribute this qualitatively different behavior to undesirable "loop" linking motifs that are predicted to be more prevalent for linkers with end-to-end distances commensurate with the locations of chemical bonding sites on the colloids. Linker flexibility also influences the spacing between linked colloids, suggesting strategies to design gels with desired phase behavior, structure, and by extension, structure-dependent properties. △ Less

Submitted 24 November, 2020; originally announced November 2020.

Comments: 11 pages, 8 figures, supplementary material

Journal ref: J. Chem. Phys. 154, 074901 (2021)

The Maslov index and spectral counts for linear Hamiltonian systems on $\mathbb{R}$

Authors: Peter Howard

Abstract: Working with a general class of linear Hamiltonian systems specified on $\mathbb{R}$, we develop a framework for relating the Maslov index to the number of eigenvalues the systems have on intervals of the form $[λ_1, λ_2)$ and $(-\infty, λ_2)$. We verify that our framework can be implemented for Sturm-Liouville systems, fourth-order potential systems, and a family of systems nonlinear in the spect… ▽ More Working with a general class of linear Hamiltonian systems specified on $\mathbb{R}$, we develop a framework for relating the Maslov index to the number of eigenvalues the systems have on intervals of the form $[λ_1, λ_2)$ and $(-\infty, λ_2)$. We verify that our framework can be implemented for Sturm-Liouville systems, fourth-order potential systems, and a family of systems nonlinear in the spectral parameter. The analysis is primarily motivated by applications to the analysis of spectral stability for nonlinear waves, and aspects of such analyses are emphasized. △ Less

Submitted 14 October, 2020; originally announced November 2020.

Comments: 44 pages, 2 figures

MSC Class: 34C10

Predicting Engagement with the Internet Research Agency's Facebook and Instagram Campaigns around the 2016 U.S. Presidential Election

Authors: Dimitra Liotsiou, Bharath Ganesh, Philip N. Howard

Abstract: The Russian Internet Research Agency's (IRA) online interference campaign in the 2016 U.S. presidential election represents a turning point in the trajectory of democratic elections in the digital age. What can we learn about how the IRA engages U.S. audiences, ahead of the 2020 U.S. presidential election? We provide the first in-depth analysis of the relationships between IRA content characterist… ▽ More The Russian Internet Research Agency's (IRA) online interference campaign in the 2016 U.S. presidential election represents a turning point in the trajectory of democratic elections in the digital age. What can we learn about how the IRA engages U.S. audiences, ahead of the 2020 U.S. presidential election? We provide the first in-depth analysis of the relationships between IRA content characteristics and user engagement on Facebook and Instagram around the 2016 election. We find that content targeting right-wing and non-Black marginalised groups had the strongest positive association with engagement on both Facebook and Instagram, in contrast to findings from the IRA campaign on Twitter and to some previous commentary in the media. Higher engagement was associated with posting later in the 2015-2017 period and using less text on both platforms, using negative wording and not including links on Facebook, and using fewer hashtags on Instagram. The sub-audiences and sub-issues associated with most engagement differed across the platforms. △ Less

Submitted 28 October, 2020; originally announced October 2020.

Wertheim's thermodynamic perturbation theory with double-bond association and its application to colloid-linker mixtures

Authors: Michael P. Howard, Zachary M. Sherman, Delia J. Milliron, Thomas M. Truskett

Abstract: We extend Wertheim's thermodynamic perturbation theory to derive the association free energy of a multicomponent mixture for which double bonds can form between any two pairs of the molecules' arbitrary number of bonding sites. This generalization reduces in limiting cases to prior theories that restrict double bonding to at most one pair of sites per molecule. We apply the new theory to an associ… ▽ More We extend Wertheim's thermodynamic perturbation theory to derive the association free energy of a multicomponent mixture for which double bonds can form between any two pairs of the molecules' arbitrary number of bonding sites. This generalization reduces in limiting cases to prior theories that restrict double bonding to at most one pair of sites per molecule. We apply the new theory to an associating mixture of colloidal particles ("colloids") and flexible chain molecules ("linkers"). The linkers have two functional end groups, each of which may bond to one of several sites on the colloids. Due to their flexibility, a significant fraction of linkers can "loop" with both ends bonding to sites on the same colloid instead of bridging sites on different colloids. We use the theory to show that the fraction of linkers in loops depends sensitively on the linker end-to-end distance relative to the colloid bonding-site distance, which suggests strategies for mitigating the loop formation that may otherwise hinder linker-mediated colloidal assembly. △ Less

Submitted 10 December, 2020; v1 submitted 15 October, 2020; originally announced October 2020.

Comments: 19 pages, 9 figures

Journal ref: J. Chem. Phys. 154, 024905 (2021)

Renormalized Oscillation Theory for Singular Linear Hamiltonian Systems

Authors: Peter Howard, Alim Sukhtayev

Abstract: Working with a general class of linear Hamiltonian systems with at least one singular boundary condition, we show that renormalized oscillation results can be obtained in a natural way through consideration of the Maslov index associated with appropriately chosen paths of Lagrangian subspaces of $\mathbb{C}^{2n}$. This extends previous work by the authors for regular linear Hamiltonian systems. Working with a general class of linear Hamiltonian systems with at least one singular boundary condition, we show that renormalized oscillation results can be obtained in a natural way through consideration of the Maslov index associated with appropriately chosen paths of Lagrangian subspaces of $\mathbb{C}^{2n}$. This extends previous work by the authors for regular linear Hamiltonian systems. △ Less

Submitted 22 September, 2020; originally announced September 2020.

Comments: 55 pages, 2 figures

MSC Class: 34C10

A Galactic Dust Devil: far-infrared observations of the Tornado Supernova Remnant candidate

Authors: Hannah Chawner, Alex D. P. Howard, Haley L. Gomez, Mikako Matsuura, Felix Priestley, Mike J. Barlow, Ilse De Looze, Andreas Papageorgiou, Ken Marsh, Matt W. L. Smith, Alberto Noriega-Crespo, Jeonghee Rho, Loretta Dunne

Abstract: We present complicated dust structures within multiple regions of the candidate supernova remnant (SNR) the `Tornado' (G357.7-0.1) using observations with Spitzer and Herschel. We use Point Process Map**, PPMAP, to investigate the distribution of dust in the Tornado at a resolution of 8", compared to the native telescope beams of 5-36". We find complex dust structures at multiple temperatures wi… ▽ More We present complicated dust structures within multiple regions of the candidate supernova remnant (SNR) the `Tornado' (G357.7-0.1) using observations with Spitzer and Herschel. We use Point Process Map**, PPMAP, to investigate the distribution of dust in the Tornado at a resolution of 8", compared to the native telescope beams of 5-36". We find complex dust structures at multiple temperatures within both the head and the tail of the Tornado, ranging from 15 to 60K. Cool dust in the head forms a shell, with some overlap with the radio emission, which envelopes warm dust at the X-ray peak. Akin to the terrestrial sandy whirlwinds known as `Dust Devils', we find a large mass of dust contained within the Tornado. We derive a total dust mass for the Tornado head of 16.7 solar masses, assuming a dust absorption coefficient of kappa_300 =0.56m^2 kg^1, which can be explained by interstellar material swept up by a SNR expanding in a dense region. The X-ray, infra-red, and radio emission from the Tornado head indicate that this is a SNR. The origin of the tail is more unclear, although we propose that there is an X-ray binary embedded in the SNR, the outflow from which drives into the SNR shell. This interaction forms the helical tail structure in a similar manner to that of the SNR W50 and microquasar SS433. △ Less

Submitted 22 September, 2020; v1 submitted 17 September, 2020; originally announced September 2020.

Comments: 16 pages, 10 figures + 3 appendix figures. Accepted to be published in MNRAS

Stratification of polymer mixtures in drying droplets: hydrodynamics and diffusion

Authors: Michael P. Howard, Arash Nikoubashman

Abstract: We study the evaporation-induced stratification of a mixture of short and long polymer chains in a drying droplet using molecular simulations. We systematically investigate the effects of hydrodynamic interactions (HI) on this process by comparing hybrid simulations accounting for HI between polymers through the multiparticle collision dynamics technique with free-draining Langevin dynamics simula… ▽ More We study the evaporation-induced stratification of a mixture of short and long polymer chains in a drying droplet using molecular simulations. We systematically investigate the effects of hydrodynamic interactions (HI) on this process by comparing hybrid simulations accounting for HI between polymers through the multiparticle collision dynamics technique with free-draining Langevin dynamics simulations neglecting the same. We find that the dried supraparticle morphologies are homogeneous when HI are included but are stratified in core--shell structures (with the short polymers forming the shell) when HI are neglected. The simulation methodology unambiguously attributes this difference to the treatment of the solvent in the two models. We rationalize the presence (or absence) of stratification by measuring phenomenological multicomponent diffusion coefficients for the polymer mixtures. The diffusion coefficients show the importance of not only solvent backflow but also HI between polymers in controlling the dried supraparticle morphology. △ Less

Submitted 16 July, 2020; v1 submitted 19 May, 2020; originally announced May 2020.

Comments: 13 pages, 9 figures

Inverse methods for design of soft materials

Authors: Zachary M. Sherman, Michael P. Howard, Beth A. Lindquist, Ryan B. Jadrich, Thomas M. Truskett

Abstract: Functional soft materials, comprising colloidal and molecular building blocks that self-organize into complex structures as a result of their tunable interactions, enable a wide array of technological applications. Inverse methods provide systematic means for navigating their inherently high-dimensional design spaces to create materials with targeted properties. While multiple physically motivated… ▽ More Functional soft materials, comprising colloidal and molecular building blocks that self-organize into complex structures as a result of their tunable interactions, enable a wide array of technological applications. Inverse methods provide systematic means for navigating their inherently high-dimensional design spaces to create materials with targeted properties. While multiple physically motivated inverse strategies have been successfully implemented in silico, their translation to guiding experimental materials discovery has thus far been limited to a handful of proof-of-concept studies. In this Perspective, we discuss recent advances in inverse methods for design of soft materials that address two challenges: (1) methodological limitations that prevent such approaches from satisfying design constraints and (2) computational challenges that limit the size and complexity of systems that can be addressed. Strategies that leverage machine learning have proven particularly effective, including methods to discover order parameters that characterize complex structural motifs and schemes to efficiently compute macroscopic properties from the underlying structure. We also highlight promising opportunities to improve the experimental realizability of materials designed computationally, including discovery of materials with functionality at multiple thermodynamic states, design of externally directed assembly protocols that are simple to implement in experiments, and strategies to improve the accuracy and computational efficiency of experimentally relevant models. △ Less

Submitted 31 March, 2020; originally announced April 2020.

Journal ref: J. Chem. Phys. 152, 140902 (2020)

Universal Gelation of Metal Oxide Nanocrystals via Depletion Attractions

Authors: Camila A. Saez Cabezas, Zachary M. Sherman, Michael P. Howard, Manuel N. Dominguez, Shin Hum Cho, Gary K. Ong, Allison Green, Thomas M. Truskett, Delia J. Milliron

Abstract: Nanocrystal gelation provides a powerful framework to translate nanoscale properties into bulk materials and to engineer emergent properties through the assembled microstructure. However, many established gelation strategies rely on chemical reactions and specific interactions, e.g., stabilizing ligands or ions on the surface of the nanocrystals, and are therefore not easily transferrable. Here, w… ▽ More Nanocrystal gelation provides a powerful framework to translate nanoscale properties into bulk materials and to engineer emergent properties through the assembled microstructure. However, many established gelation strategies rely on chemical reactions and specific interactions, e.g., stabilizing ligands or ions on the surface of the nanocrystals, and are therefore not easily transferrable. Here, we report a general gelation strategy via non-specific and purely entropic depletion attractions applied to three types of metal oxide nanocrystals. The gelation thresholds of two compositionally distinct spherical nanocrystals agree quantitatively, demonstrating the adaptability of the approach for different chemistries. Consistent with theoretical phase behavior predictions, nanocrystal cubes form gels at a lower polymer concentration than nanocrystal spheres, allowing shape to serve as a handle to control gelation. These results suggest that the fundamental underpinnings of depletion-driven assembly, traditionally associated with larger colloidal particles, are also applicable at the nanoscale. △ Less

Submitted 25 March, 2020; originally announced March 2020.

Automated Segmentation of Left Ventricle in 2D echocardiography using deep learning

Authors: Neda Azarmehr, Xujiong Ye, Faraz Janan, James P Howard, Darrel P Francis, Massoud Zolgharni

Abstract: Following the successful application of the U-Net to medical images, there have been different encoder-decoder models proposed as an improvement to the original U-Net for segmenting echocardiographic images. This study aims to examine the performance of the state-of-the-art proposed models as well as the original U-Net model by applying them to segment the endocardium of the Left Ventricle in 2D a… ▽ More Following the successful application of the U-Net to medical images, there have been different encoder-decoder models proposed as an improvement to the original U-Net for segmenting echocardiographic images. This study aims to examine the performance of the state-of-the-art proposed models as well as the original U-Net model by applying them to segment the endocardium of the Left Ventricle in 2D automatically. The prediction outputs of the models are used to evaluate the performance of the models by comparing the automated results against the expert annotations (gold standard). Our results reveal that the U-Net model outperforms other models by achieving an average Dice coefficient of 0.92$ \pm 0.05$, and Hausdorff distance of 3.97$ \pm 0.82$. △ Less

Submitted 17 March, 2020; originally announced March 2020.

Comments: 4 pages, 1 figure, Extended Abstract MIDL conference

Report number: MIDL/2019/ExtendedAbstract/Sye8klvmcN; MyUni-UID

Journal ref: MIDL/2019/ExtendedAbstract/Sye8klvmcN; MyUni-UID

Junk News & Information Sharing During the 2019 UK General Election

Authors: Nahema Marchal, Bence Kollanyi, Lisa-Maria Neudert, Hubert Au, Philip N. Howard

Abstract: Today, an estimated 75% of the British public access information about politics and public life online, and 40% do so via social media. With this context in mind, we investigate information sharing patterns over social media in the lead-up to the 2019 UK General Elections, and ask: (1) What type of political news and information were social media users sharing on Twitter ahead of the vote? (2) How… ▽ More Today, an estimated 75% of the British public access information about politics and public life online, and 40% do so via social media. With this context in mind, we investigate information sharing patterns over social media in the lead-up to the 2019 UK General Elections, and ask: (1) What type of political news and information were social media users sharing on Twitter ahead of the vote? (2) How much of it is extremist, sensationalist, or conspiratorial junk news? (3) How much public engagement did these sites get on Facebook in the weeks leading and (4) What are the most common narratives and themes relayed by junk news outlets △ Less

Submitted 27 February, 2020; originally announced February 2020.

L1495 Revisited: A PPMAP View of a Star-Forming Filament

Authors: A. D. P. Howard, A. P. Whitworth, K. A. Marsh, S. D. Clarke, M. J. Griffin, M. W. L. Smith, O. D. Lomax

Abstract: We have analysed the Herschel and SCUBA-2 dust continuum observations of the main filament in the Taurus L1495 star forming region, using the Bayesian fitting procedure PPMAP. (i) If we construct an average profile along the whole length of the filament, it has fwhm $\simeq 0.087\pm 0.003\,{\rm pc};\;$, but the closeness to previous estimates is coincidental. (ii) If we analyse small local section… ▽ More We have analysed the Herschel and SCUBA-2 dust continuum observations of the main filament in the Taurus L1495 star forming region, using the Bayesian fitting procedure PPMAP. (i) If we construct an average profile along the whole length of the filament, it has fwhm $\simeq 0.087\pm 0.003\,{\rm pc};\;$, but the closeness to previous estimates is coincidental. (ii) If we analyse small local sections of the filament, the column-density profile approximates well to the form predicted for hydrostatic equilibrium of an isothermal cylinder. (iii) The ability of PPMAP to distinguish dust emitting at different temperatures, and thereby to discriminate between the warm outer layers of the filament and the cold inner layers near the spine, leads to a significant reduction in the surface-density, $\varSigma$, and hence in the line-density, $μ$. If we adopt the canonical value for the critical line-density at a gas-kinetic temperature of $10\,{\rm K}$, $μ_{_{\rm CRIT}}\simeq 16\,{\rm M_{_\odot}\,pc^{-1}}$, the filament is on average trans-critical, with ${\barμ}\sim μ_{_{\rm CRIT}};\;$ local sections where $μ>μ_{_{\rm CRIT}}$ tend to lie close to pre-stellar cores. (iv) The ability of PPMAP to distinguish different types of dust, i.e. dust characterised by different values of the emissivity index, $β$, reveals that the dust in the filament has a lower emissivity index, $β\leq1.5$, than the dust outside the filament, $β\geq 1.7$, implying that the physical conditions in the filament have effected a change in the properties of the dust. △ Less

Submitted 6 August, 2019; originally announced August 2019.

Comments: 16 pages, 9 figures, Accepted for publication by MNRAS August 2019

Stability of force-driven shear flows in nonequilibrium molecular simulations with periodic boundaries

Authors: Michael P. Howard, Antonia Statt, Howard A. Stone, Thomas M. Truskett

Abstract: We analyze the hydrodynamic stability of force-driven parallel shear flows in nonequilibrium molecular simulations with three-dimensional periodic boundary conditions. We show that flows simulated in this way can be linearly unstable, and we derive an expression for the critical Reynolds number as a function of the geometric aspect ratio of the simulation domain. Approximate periodic extensions of… ▽ More We analyze the hydrodynamic stability of force-driven parallel shear flows in nonequilibrium molecular simulations with three-dimensional periodic boundary conditions. We show that flows simulated in this way can be linearly unstable, and we derive an expression for the critical Reynolds number as a function of the geometric aspect ratio of the simulation domain. Approximate periodic extensions of Couette and Poiseuille flows are unstable at Reynolds numbers two orders of magnitude smaller than their aperiodic equivalents because the periodic boundaries impose fundamentally different constraints on the flow. This instability has important implications for simulating shear rheology and for designing nonequilibrium simulation methods that are compatible with periodic boundary conditions. △ Less

Submitted 13 May, 2020; v1 submitted 16 July, 2019; originally announced July 2019.

Comments: 10 pages, 6 figures

Structure and phase behavior of polymer-linked colloidal gels

Authors: Michael P. Howard, Ryan B. Jadrich, Beth A. Lindquist, Fardin Khabaz, Roger T. Bonnecaze, Delia J. Milliron, Thomas M. Truskett

Abstract: Low-density "equilibrium" gels that consist of a percolated, kinetically arrested network of colloidal particles and are resilient to aging can be fabricated by restricting the number of effective bonds that form between the colloids. Valence-restricted patchy particles have long served as one archetypal example of such materials, but equilibrium gels can also be realized through a synthetically s… ▽ More Low-density "equilibrium" gels that consist of a percolated, kinetically arrested network of colloidal particles and are resilient to aging can be fabricated by restricting the number of effective bonds that form between the colloids. Valence-restricted patchy particles have long served as one archetypal example of such materials, but equilibrium gels can also be realized through a synthetically simpler and scalable strategy that introduces a secondary linker, such as a small ditopic molecule, to mediate the bonds between the colloids. Here, we consider the case where the ditopic linker molecules are low-molecular-weight polymers and demonstrate using a model colloid-polymer mixture how macroscopic properties such as the phase behavior as well as the microstructure of the gel can be designed through the polymer molecular weight and concentration. The low-density window for equilibrium gel formation is favorably expanded using longer linkers, while necessarily increasing the spacing between all colloids. However, we show that blends of linkers with different sizes enable wider variation in microstructure for a given target phase behavior. Our computational study suggests a robust and tunable strategy for the experimental realization of equilibrium colloidal gels. △ Less

Submitted 10 July, 2019; originally announced July 2019.

Comments: 11 pages, 8 figures

Journal ref: Journal of Chemical Physics 151, 124901 (2019)

The Role of Pressure in Inverse Design for Assembly

Authors: Beth A. Lindquist, Ryan B. Jadrich, Michael P. Howard, Thomas M. Truskett

Abstract: Isotropic pairwise interactions that promote the self assembly of complex particle morphologies have been discovered by inverse design strategies derived from the molecular coarse-graining literature. While such approaches provide an avenue to reproduce structural correlations, thermodynamic quantities such as the pressure have typically not been considered in self-assembly applications. In this w… ▽ More Isotropic pairwise interactions that promote the self assembly of complex particle morphologies have been discovered by inverse design strategies derived from the molecular coarse-graining literature. While such approaches provide an avenue to reproduce structural correlations, thermodynamic quantities such as the pressure have typically not been considered in self-assembly applications. In this work, we demonstrate that relative entropy optimization can be used to discover potentials that self-assemble into targeted cluster morphologies with a prescribed pressure when the iterative simulations are performed in the isothermal-isobaric ensemble. By tuning the pressure in the optimization, we generate a family of simple pair potentials that all self-assemble the same structure. Selecting an appropriate simulation ensemble to control the thermodynamic properties of interest is a general design strategy that could also be used to discover interaction potentials that self-assemble structures having, for example, a specified chemical potential. △ Less

Submitted 4 June, 2019; originally announced June 2019.

Comments: 29 pages, 8 figures

Journal ref: J. Chem. Phys. 151, 104104 (2019)

Crack formation and self-closing in shrinkable, granular packings

Authors: H. Jeremy Cho, Nancy B. Lu, Michael P. Howard, Rebekah A. Adams, Sujit S. Datta

Abstract: Many clays, soils, biological tissues, foods, and coatings are shrinkable, granular materials: they are composed of packed, hydrated grains that shrink when dried. In many cases, these packings crack during drying, critically hindering applications. However, while cracking has been widely studied for bulk gels and packings of non-shrinkable grains, little is known about how packings of shrinkable… ▽ More Many clays, soils, biological tissues, foods, and coatings are shrinkable, granular materials: they are composed of packed, hydrated grains that shrink when dried. In many cases, these packings crack during drying, critically hindering applications. However, while cracking has been widely studied for bulk gels and packings of non-shrinkable grains, little is known about how packings of shrinkable grains crack. Here, we elucidate how grain shrinkage alters cracking during drying. Using experiments with model shrinkable hydrogel beads, we show that differential shrinkage can dramatically alter crack evolution during drying---in some cases, even causing cracks to spontaneously "self-close". In other cases, packings shrink without cracking or crack irreversibly. We developed both granular and continuum models to quantify the interplay between grain shrinkage, poromechanics, packing size, drying rate, capillarity, and substrate friction on cracking. Guided by the theory, we also found that cracking can be completely altered by varying the spatial profile of drying. Our work elucidates the rich physics underlying cracking in shrinkable, granular packings, and yields new strategies for controlling crack evolution. △ Less

Submitted 15 May, 2019; originally announced May 2019.

Comments: In press

Journal ref: Soft Matter 15, 4689-4702 (2019)

The Maslov and Morse Indices for Sturm-Liouville Systems on the Half-Line

Authors: Peter Howard, Alim Sukhtayev

Abstract: We show that for Sturm-Liouville Systems on the half-line $[0,\infty)$, the Morse index can be expressed in terms of the Maslov index and an additional term associated with the boundary conditions at $x = 0$. Relations are given both for the case in which the target Lagrangian subspace is associated with the space of $L^2 ((0,\infty), \mathbb{C}^{n})$ solutions to the Sturm-Liouville System, and t… ▽ More We show that for Sturm-Liouville Systems on the half-line $[0,\infty)$, the Morse index can be expressed in terms of the Maslov index and an additional term associated with the boundary conditions at $x = 0$. Relations are given both for the case in which the target Lagrangian subspace is associated with the space of $L^2 ((0,\infty), \mathbb{C}^{n})$ solutions to the Sturm-Liouville System, and the case when the target Lagrangian subspace is associated with the space of solutions satisfying the boundary conditions at $x = 0$. In the former case, a formula of Hörmander's is used to show that the target space can be replaced with the Dirichlet space, along with additional explicit terms. We illustrate our theory by applying it to an eigenvalue problem that arises when the nonlinear Schrödinger equation on a star graph is linearized about a half-soliton solution. △ Less

Submitted 18 March, 2019; originally announced March 2019.

Comments: 34 pages

MSC Class: 34B24

Modeling hydrodynamic interactions in soft materials with multiparticle collision dynamics

Authors: Michael P. Howard, Arash Nikoubashman, Jeremy C. Palmer

Abstract: Multiparticle collision dynamics (MPCD) is a flexible and robust mesoscale computational technique for simulating solvent-mediated hydrodynamic interactions in soft materials. Here, we provide a critical overview of the MPCD method and summarize its current strengths and limitations. The capabilities of the method are highlighted by reviewing its recent applications to simulate diverse phenomena,… ▽ More Multiparticle collision dynamics (MPCD) is a flexible and robust mesoscale computational technique for simulating solvent-mediated hydrodynamic interactions in soft materials. Here, we provide a critical overview of the MPCD method and summarize its current strengths and limitations. The capabilities of the method are highlighted by reviewing its recent applications to simulate diverse phenomena, ranging from the flow of complex fluids and thermo-osmotic transport to bacterial swimming and active particle self-assembly. We also discuss outstanding challenges and emerging methodological developments that are expected to greatly expand the applicability of MPCD to other systems of technological importance. △ Less

Submitted 27 February, 2019; originally announced February 2019.

Comments: 3 figures

Quantized bounding volume hierarchies for neighbor search in molecular simulations on graphics processing units

Authors: Michael P. Howard, Antonia Statt, Felix Madutsa, Thomas M. Truskett, Athanassios Z. Panagiotopoulos

Abstract: We present an algorithm for neighbor search in molecular simulations on graphics processing units (GPUs) based on bounding volume hierarchies (BVHs). The BVH is compressed into a low-precision, quantized representation to increase the BVH traversal speed compared to a previous implementation. We find that neighbor search using the quantized BVH is roughly two to four times faster than current stat… ▽ More We present an algorithm for neighbor search in molecular simulations on graphics processing units (GPUs) based on bounding volume hierarchies (BVHs). The BVH is compressed into a low-precision, quantized representation to increase the BVH traversal speed compared to a previous implementation. We find that neighbor search using the quantized BVH is roughly two to four times faster than current state-of-the-art methods using uniform grids (cell lists) for a suite of benchmarks for common molecular simulation models. Based on the benchmark results, we recommend using the BVH instead of a single cell list for neighbor list generation in molecular simulations on GPUs. △ Less

Submitted 25 March, 2019; v1 submitted 23 January, 2019; originally announced January 2019.

Comments: 28 pages, 8 figures

Journal ref: Computational Materials Science 164, 139-146 (2019)

The Junk News Aggregator: Examining junk news posted on Facebook, starting with the 2018 US Midterm Elections

Authors: Dimitra Liotsiou, Bence Kollanyi, Philip N. Howard

Abstract: In recent years, the phenomenon of online misinformation and junk news circulating on social media has come to constitute an important and widespread problem affecting public life online across the globe, particularly around important political events such as elections. At the same time, there have been calls for more transparency around misinformation on social media platforms, as many of the mos… ▽ More In recent years, the phenomenon of online misinformation and junk news circulating on social media has come to constitute an important and widespread problem affecting public life online across the globe, particularly around important political events such as elections. At the same time, there have been calls for more transparency around misinformation on social media platforms, as many of the most popular social media platforms function as "walled gardens," where it is impossible for researchers and the public to readily examine the scale and nature of misinformation activity as it unfolds on the platforms. In order to help address this, we present the Junk News Aggregator, a publicly available interactive web tool, which allows anyone to examine, in near real-time, all of the public content posted to Facebook by important junk news sources in the US. It allows the public to gain access to and examine the latest articles posted on Facebook (the most popular social media platform in the US and one where content is not readily accessible at scale from the open Web), as well as organise them by time, news publisher, and keywords of interest, and sort them based on all eight engagement metrics available on Facebook. Therefore, the Aggregator allows the public to gain insights on the volume, content, key themes, and types and volumes of engagement received by content posted by junk news publishers, in near real-time, hence opening up and offering transparency in these activities as they unfold, at scale across the top most popular junk news publishers. In this way, the Aggregator can help increase transparency around the nature, volume, and engagement with junk news on social media, and serve as a media literacy tool for the public. △ Less

Submitted 17 April, 2019; v1 submitted 23 January, 2019; originally announced January 2019.

Cross-stream migration of a Brownian droplet in a polymer solution under Poiseuille flow

Authors: Michael P. Howard, Thomas M. Truskett, Arash Nikoubashman

Abstract: The migration of a Brownian fluid droplet in a parallel-plate microchannel was investigated using dissipative particle dynamics computer simulations. In a Newtonian solvent, the droplet migrated toward the channel walls due to inertial effects at the studied flow conditions, in agreement with theoretical predictions and recent simulations. However, the droplet focused onto the channel centerline w… ▽ More The migration of a Brownian fluid droplet in a parallel-plate microchannel was investigated using dissipative particle dynamics computer simulations. In a Newtonian solvent, the droplet migrated toward the channel walls due to inertial effects at the studied flow conditions, in agreement with theoretical predictions and recent simulations. However, the droplet focused onto the channel centerline when polymer chains were added to the solvent. Focusing was typically enhanced for longer polymers and higher polymer concentrations with a nontrivial flow-rate dependence due to droplet and polymer deformability. Brownian motion caused the droplet position to fluctuate with a distribution that primarily depended on the balance between inertial lift forces pushing the droplet outward and elastic forces from the polymers driving it inward. The droplet shape was controlled by the local shear rate, and so its average shape depended on the droplet distribution. △ Less

Submitted 17 December, 2018; originally announced December 2018.

Comments: 10 pages, 9 figures

Journal ref: Soft Matter, 2019,15, 3168-3178 (2019)

Unexpected secondary flows in reverse nonequilibrium shear flow simulations

Authors: Antonia Statt, Michael P. Howard, Athanassios Z. Panagiotopoulos

Abstract: We simulated two particle-based fluid models, namely multiparticle collision dynamics and dissipative particle dynamics, under shear using reverse nonequilibrium simulations (RNES). In cubic periodic simulation boxes, the expected shear flow profile for a Newtonian fluid developed, consistent with the fluid viscosities. However, unexpected secondary flows along the shear gradient formed when the s… ▽ More We simulated two particle-based fluid models, namely multiparticle collision dynamics and dissipative particle dynamics, under shear using reverse nonequilibrium simulations (RNES). In cubic periodic simulation boxes, the expected shear flow profile for a Newtonian fluid developed, consistent with the fluid viscosities. However, unexpected secondary flows along the shear gradient formed when the simulation box was elongated in the flow direction. The standard shear flow profile was obtained when the simulation box was longer in the shear-gradient dimension than the flow dimension, while the secondary flows were always present when the flow dimension was at least 25% larger than the shear-gradient dimension. The secondary flows satisfy the boundary conditions imposed by the RNES and have a lower rate of viscous dissipation in the fluid than the corresponding unidirectional flows. This work highlights a previously unappreciated limitation of RNES for generating shear flow in simulation boxes that are elongated in the flow dimension, an important consideration when applying RNES to complex fluids like polymer solutions. △ Less

Submitted 9 November, 2018; originally announced November 2018.

Renormalized oscillation theory for linear Hamiltonian systems on [0,1] via the Maslov index

Authors: Peter Howard, Alim Sukhtayev

Abstract: Working with a general class of linear Hamiltonian systems on $[0, 1]$, we show that renormalized oscillation results can be obtained in a natural way through consideration of the Maslov index associated with appropriately chosen paths of Lagrangian subspaces of $\mathbb{C}^{2n}$. We verify that our applicability class includes Dirac and Sturm-Liouville systems, as well as a system arising from di… ▽ More Working with a general class of linear Hamiltonian systems on $[0, 1]$, we show that renormalized oscillation results can be obtained in a natural way through consideration of the Maslov index associated with appropriately chosen paths of Lagrangian subspaces of $\mathbb{C}^{2n}$. We verify that our applicability class includes Dirac and Sturm-Liouville systems, as well as a system arising from differential-algebraic equations for which the spectral parameter appears nonlinearly. △ Less

Submitted 10 December, 2021; v1 submitted 24 August, 2018; originally announced August 2018.

Studying Politically Vulnerable Communities Online: Ethical Dilemmas, Questions, and Solutions

Authors: Robert Gorwa, Philip N. Howard

Abstract: This short article introduces the concept of political vulnerability for social media researchers. How are traditional notions of harm challenged by research subjects in politically vulnerable communities? Through a selection of case studies, we explore some of the trade-offs, challenges, and questions raised by research that seeks be robust and transparent while also preserving anonymity and priv… ▽ More This short article introduces the concept of political vulnerability for social media researchers. How are traditional notions of harm challenged by research subjects in politically vulnerable communities? Through a selection of case studies, we explore some of the trade-offs, challenges, and questions raised by research that seeks be robust and transparent while also preserving anonymity and privacy, especially in high-stakes, politically fraught contexts. △ Less

Submitted 3 June, 2018; originally announced June 2018.

Comments: 2018 ICWSM Workshop on Exploring Ethical Trade-offs in Social Media Research, June 25, Stanford, CA, USA

Influence of hydrodynamic interactions on stratification in drying mixtures

Authors: Antonia Statt, Michael P. Howard, Athanassios Z. Panagiotopoulos

Abstract: Nonequilibrium molecular dynamics simulations are used to investigate the influence of hydrodynamic interactions on vertical segregation (stratification) in drying mixtures of long and short polymer chains. In agreement with previous computer simulations and theoretical modeling, the short polymers stratify on top of the long polymers at the top of the drying film when hydrodynamic interactions be… ▽ More Nonequilibrium molecular dynamics simulations are used to investigate the influence of hydrodynamic interactions on vertical segregation (stratification) in drying mixtures of long and short polymer chains. In agreement with previous computer simulations and theoretical modeling, the short polymers stratify on top of the long polymers at the top of the drying film when hydrodynamic interactions between polymers are neglected. However, no stratification occurs at the same drying conditions when hydrodynamic interactions are incorporated through an explicit solvent model. Our analysis demonstrates that models lacking hydrodynamic interactions do not faithfully represent stratification in drying mixtures, in agreement with recent analysis of an idealized model for diffusiophoresis, and must be incorporated into such models in future. △ Less

Submitted 30 March, 2018; originally announced April 2018.

Polarization, Partisanship and Junk News Consumption over Social Media in the US

Authors: Vidya Narayanan, Vlad Barash, John Kelly, Bence Kollanyi, Lisa-Maria Neudert, Philip N. Howard

Abstract: What kinds of social media users read junk news? We examine the distribution of the most significant sources of junk news in the three months before President Donald Trump first State of the Union Address. Drawing on a list of sources that consistently publish political news and information that is extremist, sensationalist, conspiratorial, masked commentary, fake news and other forms of junk news… ▽ More What kinds of social media users read junk news? We examine the distribution of the most significant sources of junk news in the three months before President Donald Trump first State of the Union Address. Drawing on a list of sources that consistently publish political news and information that is extremist, sensationalist, conspiratorial, masked commentary, fake news and other forms of junk news, we find that the distribution of such content is unevenly spread across the ideological spectrum. We demonstrate that (1) on Twitter, a network of Trump supporters shares the widest range of known junk news sources and circulates more junk news than all the other groups put together; (2) on Facebook, extreme hard right pages, distinct from Republican pages, share the widest range of known junk news sources and circulate more junk news than all the other audiences put together; (3) on average, the audiences for junk news on Twitter share a wider range of known junk news sources than audiences on Facebook public pages. △ Less

Submitted 4 March, 2018; originally announced March 2018.

Comments: arXiv admin note: text overlap with arXiv:1802.03572

Report number: Data Memo 2018.1