Conservative Networks Research Articles

Exactly conservative physics-informed neural networks and deep operator networks for dynamical systems

We introduce a method for training exactly conservative physics-informed neural networks and physics-informed deep operator networks for dynamical systems, that is, for ordinary differential equations. The method employs a projection-based technique that maps a candidate solution learned by the neural network solver for any given dynamical system possessing at least one first integral onto an invariant manifold. We illustrate that exactly conservative physics-informed neural network solvers and physics-informed deep operator networks for dynamical systems vastly outperform their non-conservative counterparts for several real-world problems from the mathematical sciences.

Conservative deep neural networks for modeling competition of ribosomes with extended length

We develop a network model that combines several ribosome flow models with extended objects (RFMEO) competing for the finite pool of ribosomes. This alleviates the need to systematically coarse-grain the mRNA molecules. The dynamics of the network is described by a system of non-linear ordinary differential equations. It is shown that the network always converges to a steady state for a fixed number of ribosomes. Our analysis shows that increasing any of the transition rates along an RFMEO increases its output rate and either the output rates of the other RFMEOs all increase or all decrease. Simulations also demonstrate a counterintuitive result that increasing the ribosomal footprint may sometimes lead to an increase in the network production rate. Next, we propose a conservative deep neural network (CDNN) framework to approximate the solution of the network. The proposed loss function also incorporates the term satisfying the first integral property of the network. Point-wise comparison of the solutions by CDNN is in good agreement with the Runge–Kutta based numerical solution. Also, the CDNN framework offers a closed-form solution of the RFMEONP as a function of free parameters, thus allowing evaluation of the solution at any parameter value without again simulating the system.

Genocide, surveillance, and babies: “embodied propaganda” and the anti-abortion to conspiracy pipeline

ABSTRACT The judicial decision to overturn the federal right to abortion in the United States has profoundly influenced women across the country. Relying on 2,230 individual stories and over 500 posts on Instagram by anti-abortion activists, we investigate their discourse around abortion produced in a specific cultural moment—that of post-Roe America and a post-truth online society in which disinformation and conspiracy spread rampantly. This combined moment is particularly relevant given that people who can get pregnant are being marginalized and criminalized in American society at the same time that trust in experts and institutions is declining, thus troubling both institutional and feminist knowledge-production. Conceptually we draw on existing studies on embodied, feminist knowledge, intersectionality, populist expertise and conspiracy-believing, as well as gendered disinformation and conspiracy to show how (a) anti-abortion rhetoric devolves not only into disinformation but also conspiracy theories that are common among the alt-right and conservative online networks; (b) that the feedback loop among these communities is strong, and (c) that the anti-abortion movement on Instagram is experimenting with Gen Z rhetoric. We introduce and demonstrate the concept of embodied propaganda to capture the phenomenon of co-opting the experiences of marginalized groups in society to manipulate public opinion.

FastCCLasso: a fast and efficient algorithm for estimating correlation matrix from compositional data.

The composition and structure of microbial communities on the body surface are closely related to human health. The interaction relationship among microbes can help us understand the formation of the microecological environment and the biological mechanism by which microorganisms influence host health. With the help of high-throughput sequencing technologies, microbial abundances in a natural environment can be directly measured without the isolation of microorganisms in culture. Sequencing experiments in microbiome studies can measure the relative abundance of microbes, which is called compositional data. Although there are already many methods for correlation analysis for compositional data, the computation time or accuracy still needs to be improved for current microbiome studies. We develop a fast and efficient algorithm, called fastCCLasso, based on a penalized weighted least squares for inferring the correlation structure of microbes from compositional data in microbiome studies. We perform a large number of numerical experiments and the simulation results show that fastCCLasso outperforms its competitors in edge detection for inferring the correlation network. We also apply fastCCLasso for estimating microbial networks in microbiome studies and fastCCLasso provides a conservative network with comparable false discovery counts that are derived from shuffled data. FastCCLasso is open source and freely available from https://github.com/ShenZhang-Statistics/fastCCLasso under GNU LGPL v3.

Enhancing Fingerprint Forensics: A Comprehensive Study of Gender Classification Based on Advanced Data-Centric AI Approaches and Multi-Database Analysis

Fingerprint analysis has long been a cornerstone in criminal investigations for suspect identification. Beyond this conventional role, recent efforts have aimed to extract additional demographic information from fingerprints, such as gender, age, and nationality. While studies have demonstrated promising accuracy in gender classification based on fingerprints, practical implementation faces challenges, including the often low quality of crime scene fingerprints. This study presents a pioneering comparison of gender classification across diverse datasets, considering variations in fingerprint image quality. We examine the results from four databases, encompassing both public and private sources, employing state-of-the-art Data-Centric AI (DCAI) approaches for enhanced classification. Our findings reveal that a conservative Convolutional Neural Network (CNN)—specifically VGG—proves effective, achieving an accuracy ranging from 70% to 95% based on fingerprint quality. DCAI methods contribute a noteworthy 1–4% improvement. Notably, for partial or low-quality fingerprints, the periphery emerges as a critical determinant of gender classification. This study contributes insights into practical gender classification from fingerprints, emphasizing the significance of the fingerprint periphery. Furthermore, we provide the source code for future research and accessibility in real-world applications.

Conservative network for offline reinforcement learning

Offline reinforcement learning (RL) aims to learn policies from static datasets. The value overestimation of out-of-distribution (OOD) actions makes it difficult to directly apply general RL methods in the offline setting. To overcome this problem, many works focus on estimating the value function conservatively or pessimistically. However, existing methods require additional OOD sampling or uncertainty estimation to underestimate OOD values, making them complex and vulnerable to hyperparameters. Is it possible to design a specific value function that can automatically be conservative on OOD samples? In this study, we reveal the anti-conservation property of the widely used ReLU network under certain conditions and explain the reason theoretically. Based on the analysis of the ReLU network, we propose a novel neural network architecture that pushes down the value of those samples far away from the datasets; we call this kind of new architecture the Conservative Network (ConsNet). Based on ConsNet, a new offline RL algorithm with simple implementation and high performance is proposed. Since we can obtain additional conservation from the ConsNet itself, by integrating the ConsNet into several existing offline RL methods, we find that it can significantly improve the performance or reduce the original algorithm complexity. With its simplicity and superiority, we hope that ConsNet could be a new fundamental network architecture for offline RL.

Conservative Sparse Neural Network Embedded Frequency-Constrained Unit Commitment With Distributed Energy Resources

The increasing penetration of distributed energy resources (DERs) will decrease the rotational inertia of the power system and further degrade the system frequency stability. To address the above issues, this paper leverages the advanced neural network (NN) to learn the frequency dynamics and incorporates NN to facilitate system reliable operation. This paper proposes the conservative sparse neural network (CSNN) embedded frequency-constrained unit commitment (FCUC) with converter-based DERs, including the learning and optimization stages. In the learning stage, it samples the inertia parameters, calculates the corresponding frequency, and characterizes the stability region of the sampled parameters using the convex hulls to ensure stability and avoid extrapolation. For conservativeness, the positive prediction error penalty is added to the loss function to prevent possible frequency requirement violation. For the sparsity, the NN topology pruning is employed to eliminate unnecessary connections for solving acceleration. In the optimization stage, the trained CSNN is transformed into mixed-integer linear constraints using the big-M method and then incorporated to establish the data-enhanced model. The case study verifies 1) the effectiveness of the proposed model in terms of high accuracy, fewer parameters, and significant solving acceleration; 2) the stable system operation against frequency violation under contingency.

One nation, under war: Did the language of Fox News and MSNBC converge during the invasion of Ukraine?

AbstractEvidence suggests that political differences have increased markedly in the United States in recent decades. Differences may also emerge in the way that partisans express themselves through language, and it is possible that language differences vary in times of crisis and war. In the current work we examined over a decade's worth of transcripts from a liberal (MSNBC) and conservative (Fox) news network. More specifically, we examined evidence for two competing perspectives on language differences during the Russian invasion of Ukraine in 2022. One perspective, the depolarization perspective, suggests that we should see decreased differences, or parity in language styles between the two networks leading up to and during the invasion. Another perspective, the polarization perspective, suggests we should see increased differences in language styles between networks leading up to and during the early stages of the invasion. We examined an index of personalizing and formalizing language as well as 77 Linguistic Inquiry and Word Count (LIWC) dictionaries plus noun frequency using smoothed curves and linear discriminant function analyses (LDA) to examine the pattern of results in our data. Our results provide more support for the depolarization perspective, showing that both Fox News and MSNBC became more similar than different leading up to and during the invasion. Implications are discussed.

A conservative network element method for diffusion-advection-reaction problems

We derive a conservative network element method for heterogeneous and anisotropic diffusion problems by modifying the non-conservative version, and extend it to the approximation of an additional advection term. The numerical scheme possesses the flux formulation reminiscent of classical finite volume methods. Its convergence is naturally governed by the network element theory. Numerical results illustrate the good behavior of the method even on distorted point clouds.

Religion and Subjective Social Class in the United States

Abstract Subjective social class identities—lower, working, middle, and upper—are conditioned by culture and social interactions. I argue that conservative Christianity influences subjective class identification because conservative Christian social networks are highly insular, and its culture prioritizes lower- and working-class ideologies. Using nationally representative data, I find that conservative Christians—operationalized with views of the Bible and religious tradition—are relatively likely to identify as lower and working class, and unlikely to identify as middle and upper class; that these associations are partially but not wholly mediated by higher education and family income; and that there are robust associations between religion and subjective class among those with a bachelor’s degree and above-average family incomes, but not among less-educated and lower-income Americans. These results indicate that conservative Christianity promotes a specific class culture, and that this class culture more closely aligns with biblical literalism than with affiliation with evangelical Protestant churches.

Traditional values, family, homeschooling: The role of Russia and the Russian Orthodox Church in transnational moral conservative networks and their efforts at reshaping human rights

Abstract In the last two decades, Russian state actors and the Russian Orthodox Church have come to play an increasingly important role in the undermining of established understandings of international human rights law by reinterpreting its aims and repurposing its institutions, in particular the Universal Declaration of Human Rights, the United Nations Human Rights Council, and the European Court of Human Rights. Russian state and church leaders have appropriated and coopted the language of human rights in order to advance an illiberal and nationalist agenda that undercuts democratic values and targets particular groups and their rights and freedoms—most notably liberal civil society, political opposition, and the LGBTIQ+ community. Written from the angle of a constructivist sociology of human rights, this article brings together three case studies of Russian rights appropriation around the topics of traditional values, family, and homeschooling and draws six lessons on the (mis)appropriation of human rights for illiberal purposes. The analysis of Russia’s rights appropriation sheds light on the background and build-up for current events in Russia’s war against Ukraine.

Selecting Sunak: Conservative MPs’ Nomination Preferences in the (Second) British Conservative Party Leadership Election

Abstract This article utilises an original dataset covering all members of the Parliamentary Conservative Party (PCP) to analyse the basis of support for Rishi Sunak, Boris Johnson and Penny Mordaunt in the October 2022 British Conservative Party leadership election. The significance of our findings is that they form the basis for a new understanding of how conservatism in the UK is changing post-Brexit. Our political variables include constituency majority size, red wall status, constituency support for leave, an MP’s position in the June 2022 vote of confidence in Johnson’s leadership of the Conservative Party, ministerial status and rebellion rates on key legislation. Our ideological variables include EU referendum position and membership of the European Research Group and the Conservative Environment Network, along with whether MPs are classed as socially liberal, socially conservative or pro-levelling up. We find that candidates appealed to different, often rival, sections of the Conservative Party. In line with previous studies of this nature, social liberalism still represents a key dividing line within the party, and the European dimension of intra-party conflict has remained significant in light of the UK’s decision to leave the European Union.

“I'm going to do battle… I'm going to do some good”. Biographical trajectories, moral politics, and public engagement among highly religious young Catholics in Spain and Mexico

AbstractPublic debates and controversies on moral issues have gained visibility in recent decades in both Spain and Mexico. Conservative Catholic groups and networks are increasingly playing a crucial role in raising and framing highly morally charged themes in the public arena, especially on intimacy‐related policies, including sexual and reproductive freedoms. A new generation of young Catholic leaders becoming key political agents has emerged in this scenario. They are fostering new mobilisation repertoires in the public sphere (e.g., digital activism and public performances) and promoting new political narratives merging moral, cultural, and religious elements. These highly religious and conservative young Catholics consider themselves part of a cultural/cognitive minority and feel entitled to become what Howard Becker termed “moral entrepreneurs.” Most have been educated in Catholic schools and belong to Catholic movements. Still, they can also navigate secular politics and strategically use different regimes of justification and action. Within this context, the article aims to examine and compare the intersection of historical trends and the relevance of individual biographies to understand the role of these young Catholics in the public space in both countries.

Random Forest model reveals the interaction between N6-methyladenosine modifications and RNA-binding proteins

RNA-binding proteins (RBPs) have critical roles in N6-methyladenosine (m6A) modification process. We designed a Random Forest (RF) model to systematically analyze the interaction among RBPs and m6A modifications by integrating the binding signals from hundreds of RBPs. Accurate prediction of m6A sites demonstrated significant connections between RBP bindings and m6A modifications. The relative importance of different RBPs from the model provided a quantitative metric to evaluate their interactions with m6A modifications. Redundancy analysis showed that several RBPs may have similar binding patterns with m6A sites. The RF model exhibited fairly high prediction accuracy across cell lines, suggesting a conservative RBP interaction network regulates m6A occupancy. Specific RBPs can engage to the corresponding regional m6A sites and deploy distinct regulatory processes, such as cleavage site selection of the alternative polyadenylation (APA). We also integrated histone modifications into our RF model, which demonstrated H3K36me3 and H3K27me3 as determining features for m6A distribution.

Integrative Complexity, COVID-19, and Political Ideology

This study explores the relationship between political ideology and integrative complexity in terms of COVID-19 news coverage. Briefly, integrative complexity is a psycholinguistic construct that examines cognitive structure. Based on a sample of COVID-19 news articles sourced from liberal, moderate, and conservative news networks, the results show liberal and moderate networks displayed higher complexity scores than conservative networks. The results further confirm a complexity pattern seen in previous research — liberals and moderates tend to demonstrate more complex rhetoric than conservatives. Moreover, the conservative networks’ lower scores may be explained by the predominance of misinformation in their coverage of COVID-19.

Not Policies. People

Not Policies. People

Fully hyperbolic convolutional neural networks

Convolutional neural networks (CNN) have recently seen tremendous success in various computer vision tasks. However, their application to problems with high dimensional input and output, such as high-resolution image and video segmentation or 3D medical imaging, has been limited by various factors. Primarily, in the training stage, it is necessary to store network activations for back-propagation. In these settings, the memory requirements associated with storing activations can exceed what is feasible with current hardware, especially for problems in 3D. Motivated by the propagation of signals over physical networks, that are governed by the hyperbolic Telegraph equation, in this work we introduce a fully conservative hyperbolic network for problems with high-dimensional input and output. We introduce a coarsening operation that allows completely reversible CNNs by using a learnable discrete wavelet transform and its inverse to both coarsen and interpolate the network state and change the number of channels. We show that fully reversible networks are able to achieve results comparable to the state of the art in 4D time-lapse hyper-spectral image segmentation and full 3D video segmentation, with a much lower memory footprint that is a constant independent of the network depth. We also extend the use of such networks to variational auto-encoders, where optimization begins from an exact recovery and we discover the level of compression through optimization.

The Role of Influential Actors in Fostering the Polarized COVID-19 Vaccine Discourse on Twitter: Mixed Methods of Machine Learning and Inductive Coding.

BackgroundSince COVID-19 vaccines became broadly available to the adult population, sharp divergences in uptake have emerged along partisan lines. Researchers have indicated a polarized social media presence contributing to the spread of mis- or disinformation as being responsible for these growing partisan gaps in uptake.ObjectiveThe major aim of this study was to investigate the role of influential actors in the context of the community structures and discourse related to COVID-19 vaccine conversations on Twitter that emerged prior to the vaccine rollout to the general population and discuss implications for vaccine promotion and policy.MethodsWe collected tweets on COVID-19 between July 1, 2020, and July 31, 2020, a time when attitudes toward the vaccines were forming but before the vaccines were widely available to the public. Using network analysis, we identified different naturally emerging Twitter communities based on their internal information sharing. A PageRank algorithm was used to quantitively measure the level of “influentialness” of Twitter accounts and identifying the “influencers,” followed by coding them into different actor categories. Inductive coding was conducted to describe discourses shared in each of the 7 communities.ResultsTwitter vaccine conversations were highly polarized, with different actors occupying separate “clusters.” The antivaccine cluster was the most densely connected group. Among the 100 most influential actors, medical experts were outnumbered both by partisan actors and by activist vaccine skeptics or conspiracy theorists. Scientists and medical actors were largely absent from the conservative network, and antivaccine sentiment was especially salient among actors on the political right. Conversations related to COVID-19 vaccines were highly polarized along partisan lines, with “trust” in vaccines being manipulated to the political advantage of partisan actors.ConclusionsThese findings are informative for designing improved vaccine information communication strategies to be delivered on social media especially by incorporating influential actors. Although polarization and echo chamber effect are not new in political conversations in social media, it was concerning to observe these in health conversations on COVID-19 vaccines during the vaccine development process.

Surrogate modeling for neutron diffusion problems based on conservative physics-informed neural networks with boundary conditions enforcement

Application of physics-informed neural network(PINN) on neutron diffusion equation, which is of great engineering significance for reactor physics field, has not received much attention yet. Meanwhile, the non-smoothness of solution for neutron diffusion equation brings difficulties for PINN’s application. Therefore, we introduce the conservative PINN(cPINN) which develops PINN for each sub-domain and considers additional conservation law along the sub-domains’ interfaces, to solve heterogeneous neutron diffusion problems. Specifically, we develop PINNs on each sub-domain which has same material property and set equality constraints for neutron flux and neutron current on adjacent sub-domains. Furthermore, we propose a neural network constructing method to ensure that PINN/cPINN predictions strictly conform to three types of boundary conditions(BCs) involved in neutron diffusion problems. The results of numerical examples demonstrate that cPINN can solve heterogeneous neutron diffusion problems with non-smooth solutions and the proposed BC-imposed method can help to improve the cPINN performance on complex heterogeneous problems.

Evolutionarily conservative and non-conservative regulatory networks during primate interneuron development revealed by single-cell RNA and ATAC sequencing

The differences in size and function between primate and rodent brains, and the association of disturbed excitatory/inhibitory balance with many neurodevelopmental disorders highlight the importance to study primate ganglionic eminences (GEs) development. Here we used single-cell RNA and ATAC sequencing to characterize the emergence of cell diversity in monkey and human GEs where most striatal and cortical interneurons are generated. We identified regional and temporal diversity among progenitor cells which give rise to a variety of interneurons. These cells are specified within the primate GEs by well conserved gene regulatory networks, similar to those identified in mice. However, we detected, in human, several novel regulatory pathways or factors involved in the specification and migration of interneurons. Importantly, comparison of progenitors between our human and published mouse GE datasets led to the discovery and confirmation of outer radial glial cells in GEs in human cortex. Our findings reveal both evolutionarily conservative and nonconservative regulatory networks in primate GEs, which may contribute to their larger brain sizes and more complex neural networks compared with mouse.