University of Illinois, Urbana-Champaign

This cautionary article urges general education preschool teachers to evaluate their responses to challenging student behavior. The manuscript delivers practical advice to improve teachers' responses to challenging behavior by refreshing their understanding of the various consequences that influence behavior recurrence. It highlights the need to avoid punitive measures that could harm teacher–child relationships and instead offer non-generic and function-specific consequences that support the child's needs and promote positive behavior development. By providing a review of the different types of consequences and real-life examples to illustrate their use, the article encourages teachers to consider the consequences they give, especially when a child's behavior changes in an unexpected direction.

This chapter explores the rise of competitive gamer SonicFox by interrogating how their intersecting identities as Black, gay, non-binary, and a furry produces new queer of colour potentialities within a historically straight, cisgender, white space. The chapter specifically highlights how, through hypermediation, SonicFox produces an embodied archive that demands simultaneous acknowledgement of, rather than essentialising, their identities across online spaces. Additionally, the chapter examines how SonicFox, by donning their fursona, playing with Asian-coded characters, and collaborating with other competitive gamers across the Asian diaspora, produces a complex and, at times, appropriative relationship with Asian culture. Contending that SonicFox knowingly deploys Asian iconography to usurp and reimagine racialised presumptions of non-white competitive gamers, not dissimilar to Asian and Asian American athletes, the chapter argues that SonicFox’s multiple sites of embodiment provide modes for destabilising the white colonialist and heteropatriarchal logics of eSports more broadly.

Dendritic cells (DCs), the main type of antigen-presenting cells in the body, act as key mediators of adaptive immunity by sampling antigens from diseased cells for the subsequent priming of antigen-specific T and B cells. While DCs can secrete a diverse array of cytokines that profoundly shape the immune milieu, exogenous cytokines are often needed to maintain the survival, proliferation, and differentiation of DCs, T cells, and B cells. However, conventional cytokine therapies for cancer treatment are limited by their low therapeutic benefit and severe side effects. The overexpression of cytokines in DCs, followed by paracrine release or membrane display, has emerged as a viable approach for controlling the exposure of cytokines to interacting DCs and T/B cells. This approach can potentially reduce the necessary dose of cytokines and associated side effects to achieve comparable or enhanced antitumor efficacy. Various strategies have been developed to enable the overexpression or chemical conjugation of cytokines on DCs for the subsequent modulation of DC–T/B-cell interactions. This review provides a brief overview of strategies that enable the overexpression of cytokines in or on DCs via genetic engineering or chemical modification methods and discusses the promise of cytokine-overexpressing DCs for the development of new-generation cancer immunotherapy.

A new method is described to characterize the transformation kinetics of reconstructive systems at high temperatures using synchrotron X‐ray radiation, a quadrupole lamp furnace, and a detector capable of collecting a diffraction pattern of sufficient range in a minimal amount of time. Using this method, the activation energy for a reconstructive phase transformation in Dy2TiO5 was determined to be 149 kJ/mol. This value is more reasonable when compared with the uncertain values determined using conventional methods such as differential scanning calorimetry. In addition, the new method also provides higher quality data with improved time resolution, leading to better determination of fitted kinetic parameters.

Intimate partner violence (IPV) is most prevalent in young adults, yet scarce evidence is available regarding South Korean young adults’ experience of IPV and culturally tailored IPV prevention programs. To address this gap, this study aimed to holistically assess IPV victimization and perpetration rates and the related risk and protective factors among Korean young adults. Using online survey data from 600 Korean young adults using simple random sampling, this study found that the lifetime prevalence of both IPV victimization and perpetration was about 30%. Both IPV victimization and perpetration had affected over 20% in the past 12 months. Independent variables in multiple logistic regression models explained 18% and 23% of variances in lifetime IPV victimization and perpetration, respectively. Korean young adults who reported more depressive symptoms were more likely to report IPV victimization. Those who reported more alcohol consumption, traditional attitudes about gender roles, being more tolerant of IPV, and poorer physical health status were also more likely to commit IPV. However, those who had experienced family neglect were less likely to report IPV perpetration. The findings of this study highlighted that childhood adverse experiences minimally explained IPV and that alcohol consumption, mental health, and attitudinal variables should be targets of IPV prevention among Korean young adults.

In Federated Learning (FL), clients may have weak devices that cannot train the full model or even hold it in their memory space. To implement large-scale FL applications, thus, it is crucial to develop a distributed learning method that enables the participation of such weak clients. We propose EmbracingFL , a general FL framework that allows all available clients to join the distributed training regardless of their system resource capacity. The framework is built upon a novel form of partial model training method in which each client trains as many consecutive output-side layers as its system resources allow. Our study demonstrates that EmbracingFL encourages each layer to have similar data representations across clients, improving FL efficiency. The proposed partial model training method guarantees convergence to a neighbor of stationary points for non-convex and smooth problems. We evaluate the efficacy of EmbracingFL under a variety of settings with a mixed number of strong, moderate ( $\sim 40\%$ memory), and weak ( $\sim 15\%$ memory) clients, datasets (CIFAR-10, FEMNIST, and IMDB), and models (ResNet20, CNN, and LSTM). Our empirical study shows that EmbracingFL consistently achieves high accuracy as like all clients are strong, outperforming the state-of-the-art width reduction methods (i.e. HeteroFL and FjORD).

This article proposes a low-side-extended-phase-shift-under-resonant modulation (LS-EPSURM) scheme for the dual-bridge-series-resonant converter (DBSRC), which is characterized by the introduction of an inner-phase-shift angle on the relatively low-voltage side while operating at the under-resonant mode. Compared with the extended-phase-shift-over-resonant modulation (EPSORM) scheme, the proposed LS-EPSURM scheme can effectively extend the soft-switching range of the circuit, while the required frequency range can be reduced instead. Moreover, this article investigates and identifies the optimal selection of control variables to reduce the backflow current and inductor root-mean-square current while achieving full-range zero voltage switching operation, thereby further improving the transmission efficiency of the circuit in the whole load range. The developed approach is validated on a 10.2 kW/L 2kW DBSRC prototype with output voltage varying from 250 to 500 V, the switching frequency range is 325–450 kHz, and the peak efficiency is 97.5%.

Printed circuit board (PCB) motors present a compelling alternative to commercial electric motors due to their cost-effective mass production. Furthermore, integrated motor drives have gained popularity thanks to their high power density and ease of installation. However, PCB motors face challenges due to their large effective air gap and fewer turns, resulting in low phase inductance. This article addresses these challenges by pairing the PCB motor with wide-bandgap devices and increasing the switching frequency to eliminate current ripple while reducing the size of filtering components, ultimately achieving a compact and efficient motor drive. The study explores various filtering options and decides upon for an L filter with line inductors on the motor phases. The losses of the system are analytically evaluated to a great extend leading to a selection of the switching frequency which is 1 MHz. Theoretical loss calculations are validated for both the electrical machine and inverter by means of calorimetric and electrical measurements. Finally, the integrated system is tested in motoring mode, operating at 5000 rpm and delivering 0.36 Nm of torque, resulting in an 82% efficiency. The article concludes that increasing the switching frequency is a feasible solution for low inductance motor drives.

In the era where the pace of Moore’s law is decelerating, specialized hardware emerges as the new frontier to sustain exponential growth in computing performance. This paper presents a novel approach to chip design, harnessing the power of Generative Artificial Intelligence (AI) to expedite the design and verification of a Vector Processor System on Chip (SoC). With the advent of open-source ASIC toolchains like OpenROAD and Process Design Kits (PDKs) such as SKY130, the landscape of chip fabrication has been democratized, allowing even small entities and hobbyists to venture into custom chip production. Despite this accessibility, the complexity of design and verification stands as a formidable challenge. Our work introduces a generative AI methodology that streamlines the creation and validation processes of SoCs. By leveraging AI-generated hardware, we aim to minimize the iterative cycles of design and testing, reduce errors, and shorten time-to-market for custom chip solutions. We discuss the implications of this approach for the industry and provide a case study on the rapid development of a Vector Processor SoC to illustrate the practical benefits and potential of AI-assisted chip design.

Bond wire lift-off is one of the main failure mechanisms for Silicon-Carbide MOSFETs. This occurs when the metallurgical weld between the bond-wire and the device source breaks creating an open circuit. The higher current on the remaining wires subjects them to additional thermal stress creating a positive feedback loop that can quickly lead to device failure if left unchecked. By detecting lift-off events during power converter operation damaged devices can be de-rated and equipment operators can be alerted to replace the power module before a failure occurs. This paper presents an online, in-situ monitoring circuit that can detect the incremental change in onstate inductance (Lds) caused by the redistribution of bond-wire current after a lift-off occurs. Unlike on-state resistance (Rds), the inductance does not depend on temperature, gate-oxide health or gate-bias level. Additionally, since the inductive impedance is frequency dependent, any bias related to converter operation can be suppressed by increasing the measurement frequency well above the switching frequency. Experimental results show that the proposed circuit can accurately measure inductive changes down to 10 pH, providing sufficient sensitivity to detect single wire lift-off events in test devices

Resumo Introdução: O exercício físico está sendo incorporado ao tratamento de pacientes em hemodiálise, porém pouco se sabe sobre as principais características dessas intervenções. Objetivo: Descrever os protocolos de exercício físico prescritos para pacientes em hemodiálise no Brasil. Métodos: Uma revisão de escopo foi conduzida de acordo com as diretrizes JBI e Prisma-ScR. Foram realizadas pesquisas na Medline, Embase e em outras três bases de dados até maio de 2024. Outras fontes (sites, livros e diretrizes) também foram pesquisadas. Foram incluídas evidências de pacientes em hemodiálise, descrevendo protocolos de exercício físico em todos os ambientes e desenhos no Brasil. Resultados: Encontradas 45 evidências, resultando em 54 protocolos de exercício físico de 16 estados brasileiros. O exercício de força (33,3%), seguido do exercício aeróbico (22,2%), foi o mais prescrito para ser realizado durante a diálise (85,2%). Os profissionais mais prevalentes na supervisão dos programas foram fisioterapeutas e profissionais de educação física (37,0% e 18,5%, respectivamente). Todos os protocolos adotaram os princípios de treinamento tipo e frequência, enquanto a progressão foi adotada em apenas 53,7%. A frequência mais prescrita foi três vezes por semana (88,9%). A intensidade do exercício foi determinada predominantemente por métodos subjetivos (33,3%). Conclusão: Os exercícios aeróbicos e de força durante a diálise foram as modalidades mais prescritas no Brasil, com a maioria dos programas sendo adequadamente supervisionada por profissionais qualificados. No entanto, os protocolos existentes não adotaram a progressão sistemática no decorrer da intervenção, o que seria adequado para proporcionar melhores respostas e adaptações fisiológicas.

Large language models (LLMs), such as GPT4 and LLaMA, are creating significant advancements in natural language processing, due to their strong text encoding/decoding ability and newly found emergent capability ( e.g. , reasoning). While LLMs are mainly designed to process pure texts, there are many real-world scenarios where text data is associated with rich structure information in the form of graphs ( e.g. , academic networks, and e-commerce networks) or scenarios where graph data is paired with rich textual information ( e.g. , molecules with descriptions). Besides, although LLMs have shown their pure text-based reasoning ability, it is underexplored whether such ability can be generalized to graphs ( i.e. , graph-based reasoning). In this paper, we provide a systematic review of scenarios and techniques related to large language models on graphs. We first summarize potential scenarios of adopting LLMs on graphs into three categories, namely pure graphs, text-attributed graphs, and text-paired graphs. We then discuss detailed techniques for utilizing LLMs on graphs, including LLM as Predictor, LLM as Encoder, and LLM as Aligner, and compare the advantages and disadvantages of different schools of models. Furthermore, we discuss the real-world applications of such methods and summarize open-source codes and benchmark datasets. Finally, we conclude with potential future research directions in this fast-growing field. The related source can be found at https://github.com/PeterGriffinJin/Awesome-Language-Model-on-Graphs .

This paper considers the problem of sequentially detecting a change in the joint distribution of multiple data sources under a sampling constraint. Specifically, the channels or sources generate observations that are independent over time, but not necessarily across channels. The joint distribution of an unknown subset of sources changes at an unknown time instant. Moreover, there is a hard constraint that only a fixed number of sources can be sampled at each time instant, but the sources can be selected dynamically based on the already collected data. The goal is to sequentially observe the sources according to the constraint, and stop sampling as quickly as possible after the change while controlling the false alarm rate below a user-specified level. Thus, a policy for this problem consists of a joint sampling and change-detection rule. A non-randomized policy is studied, and an upper bound is established on its worst-case conditional expected detection delay with respect to both the change point and the observations from the affected sources before the change. In certain cases, this rule achieves first-order asymptotic optimality as the false alarm rate tends to zero, simultaneously under every possible post-change distribution and among all schemes that satisfy the same sampling and false alarm constraints. These general results are subsequently applied to the problems of (i) detecting a change in the marginal distributions of (not necessarily independent) information sources, and (ii) detecting a change in the covariance structure of Gaussian information sources.

Identifying the dynamic properties of manipulated objects is essential for safe and accurate robot control. Most methods rely on low-noise force-torque sensors, long exciting signals, and solving nonlinear optimization problems, making the estimation process slow. In this work, we propose a fast, online learning-based inertial parameter estimation framework that enhances model-based control. We aim to quickly and accurately estimate the parameters of an unknown object using only the robot's proprioception through end-to-end learning, which is applicable for real-time system. To effectively capture features in robot proprioception solely affected by object dynamics and address the challenge of obtaining ground truth inertial parameters in the real world, we developed a high-fidelity simulation that uses more accurate robot dynamics through real-to-sim adaptation. Since our adaptation focuses solely on the robot, taskrelevant data (e.g., holding an object) is not required from the real world, simplifying the data collection process. Moreover, we address both parametric and non-parametric modeling errors independently using Robot System Identification and Gaussian Processes. We validate our estimator to assess how quickly and accurately it can estimate physically feasible parameters of an manipulated object given a specific trajectory obtained from a wheeled humanoid robot. Our estimator achieves faster estimation speeds (around 0.1 seconds) while maintaining accuracy comparable to other methods. Additionally, our estimator further highlight its benefits in improving the performance of model based control by compensating object's dynamics and re initializing new equilibrium point of wheeled humanoid

Centralized learning requires data to be aggregated at a central server, which poses significant challenges in terms of data privacy and bandwidth consumption. Federated learning presents a compelling alternative, however, vanilla federated learning methods deployed in robotics aim to learn a single global model across robots that works ideally for all. But in practice one model may not be well suited for robots deployed in various environments. This paper proposes Federated-EmbedCluster (Fed-EC), a clustering-based federated learning framework that is deployed with vision based autonomous robot navigation in diverse outdoor environments. The framework addresses the key federated learning challenge of deteriorating model performance of a single global model due to the presence of non-IID data across real-world robots. Extensive real-world experiments validate that Fed-EC reduces the communication size by 23x for each robot while matching the performance of centralized learning for goal-oriented navigation and outperforms local learning. Fed-EC can transfer previously learnt models to new robots that join the cluster.

Artificial Bee Colony (ABC) algorithms represent a category of optimization techniques inspired by the foraging behavior of honeybees. This chapter explores the fundamentals of ABC algorithms, shedding light on their underlying principles and mechanics. The ABC algorithm mimics the collaborative foraging process of bees to iteratively search for optimal solutions in complex problem spaces. The chapter delves into the key components of the algorithm, such as the employed bees, onlooker bees, and scout bees, elucidating their roles in the optimization process. Furthermore, the study extends its focus to variants of the ABC algorithm, which have evolved to address specific challenges or cater to diverse optimization scenarios. These variants may incorporate adaptive strategies, hybrid approaches, or modifications to enhance convergence speed and solution quality. By examining these variants, we aim to provide a comprehensive understanding of how the ABC algorithm family adapts to different problem domains. In summary, this chapter serves as a primer on the core concepts of ABC algorithms and offers insights into the various adaptations and enhancements that have been introduced to optimize their performance across a variety of real-world applications.

Exam retakes provide students with an additional opportunity to demonstrate mastery of learning objectives. However, this preparation might adversely affect performance on subsequent exams. This study suggests that students who choose to prepare for and take an exam retake not only improve their original exam score but show a larger improvement on subsequent exam performance than those students who did not take an exam retake.