Search

Objective: To investigate the associations of triglyceride-glucose (TyG) index, a reliable surrogate marker for insulin resistance, with the function of various cognitive domains and brain structures among older adults. Design: A population-based cross-sectional study. Setting: Older adults living in the rural communities in China. Participants: About 4,541 rural-dwelling dementia-free participants (age ≥65 years; 56.37% women) undertook examinations in March–September 2018 for MIND-China. Measurements: TyG index was calculated as ln[fasting triglyceride (mg/dL) × fasting glucose (mg/dL)/2]. A neuropsychological test battery was used to assess memory, attention, verbal fluency, and executive function. Volumetric brain measures were assessed on magnetic resonance imaging (MRI) in a subsample (n = 1,019). Data were analyzed with restricted cubic spline and multivariable general linear models. Results: An inverted J-shaped association was observed between TyG index and z-scores of multiple cognitive domains, such that among individuals with TyG index ≥8.57 (median), a higher TyG index was significantly associated with lower z-scores of memory, attention, verbal fluency, executive function, and global cognition (all p < 0.05); among people with TyG index <8.57, a higher TyG index was significantly associated with a higher executive function z-score (p < 0.05), but not with any of the other examined cognitive domains. In the MRI subsample, a higher TyG index was significantly associated with lower volumes of total brain tissue, gray matter, and white matter as well as greater cerebrospinal fluid volume (p < 0.05), but not with white matter hyperintensity volume. Conclusions: Insulin resistance, as indicated by a high TyG index, was associated with poor function in multiple cognitive domains and global brain atrophy.

Rationale: Benzene has been classified as carcinogenic to humans, but there is limited evidence linking benzene exposure to lung cancer. Objectives: We aimed to examine the relationship between occupational benzene exposure and lung cancer. Methods: Subjects from 14 case-control studies across Europe and Canada were pooled. We used a quantitative job-exposure matrix to estimate benzene exposure. Logistic regression models assessed lung cancer risk across different exposure indices. We adjusted for smoking and five main occupational lung carcinogens and stratified analyses by smoking status and lung cancer subtypes. Measurements and Main Results: Analyses included 28,048 subjects (12,329 cases, 15,719 control subjects). Lung cancer odds ratios ranged from 1.12 (95% confidence interval, 1.03-1.22) to 1.32 (95% confidence interval, 1.18-1.48) (Ptrend = 0.002) for groups with the lowest and highest cumulative occupational exposures, respectively, compared with unexposed subjects. We observed an increasing trend of lung cancer with longer duration of exposure (Ptrend < 0.001) and a decreasing trend with longer time since last exposure (Ptrend = 0.02). These effects were seen for all lung cancer subtypes, regardless of smoking status, and were not influenced by specific occupational groups, exposures, or studies. Conclusions: We found consistent and robust associations between different dimensions of occupational benzene exposure and lung cancer after adjusting for smoking and main occupational lung carcinogens. These associations were observed across different subgroups, including nonsmokers. Our findings support the hypothesis that occupational benzene exposure increases the risk of developing lung cancer. Consequently, there is a need to revisit published epidemiological and molecular data on the pulmonary carcinogenicity of benzene.

This paper aims to develop smart hydrogels based on functionalized hyaluronic acid (HA) and PLGA-PEG-PLGA (PLGA,poly-(DL-lactic-co-glycolic acid); PEG,polyethylene glycol) for use as intraocular drug-delivery platforms. Anti-inflammatory agent dexamethasone-phosphate (0.2 %w/v) was the drug selected to load on the hydrogels. Initially, different ratios of HA-aldehyde (HA-CHO) and thiolated-HA (HA-SH) were assayed, selecting as optimal concentrations 2 and 3 % (w/v), respectively. Optimized HA hydrogel formulations presented fast degradation (8 days) and drug release (91.46 ± 3.80 % in 24 h), thus being suitable for short-term intravitreal treatments. Different technology-based strategies were adopted to accelerate PLGA-PEG-PLGA water solubility, e.g. substituting PEG1500 in synthesis for higher molecular weight PEG3000 or adding cryopreserving substances to the buffer dissolution. PEG1500 was chosen to continue optimization and the final PLGA-PEG-PLGA hydrogels (PPP1500) were dissolved in trehalose or mannitol carbonate buffer. These presented more sustained release (71.77 ± 1.59 % and 73.41 ± 0.83 % in 24 h, respectively) and slower degradation (>14 days). In vitro cytotoxicity studies in the retinal-pigmented epithelial cell line (RPE-1) demonstrated good tolerance (viability values > 90 %). PLGA-PEG-PLGA hydrogels are proposed as suitable candidates for long-term intravitreal treatments. Preliminary wound healing studies with PLGA-PEG-PLGA hydrogels suggested faster proliferation at 8 h than controls., ISCIII-FEDER “Una manera de hacer Europa”, ERDF A way of making Europe, UCM Research Group 920415 (InnOftal), ESF Investing in your future, Depto. de Farmacia Galénica y Tecnología Alimentaria, Fac. de Farmacia, TRUE, pub

Dynamically organized chromatin complexes often involve multiplex chromatin interactions and sometimes chromatin-associated RNA1-3. Chromatin complex compositions change during cellular differentiation and ageing, and are expected to be highly heterogeneous among terminally differentiated single cells4-7. Here we introduce the multinucleic acid interaction mapping in single cells (MUSIC) technique for concurrent profiling of multiplex chromatin interactions, gene expression and RNA-chromatin associations within individual nuclei. When applied to 14 human frontal cortex samples from older donors, MUSIC delineated diverse cortical cell types and states. We observed that nuclei exhibiting fewer short-range chromatin interactions were correlated with both an older transcriptomic signature and Alzheimers disease pathology. Furthermore, the cell type exhibiting chromatin contacts between cis expression quantitative trait loci and a promoter tends to be that in which these cis expression quantitative trait loci specifically affect the expression of their target gene. In addition, female cortical cells exhibit highly heterogeneous interactions between XIST non-coding RNA and chromosome X, along with diverse spatial organizations of the X chromosomes. MUSIC presents a potent tool for exploration of chromatin architecture and transcription at cellular resolution in complex tissues.

A continual learning (CL) model is designed to solve the catastrophic forgetting problem, which damages the performance of neural networks by overwriting previous knowledge with new knowledge. The fundamental cause of this problem is that previous data is not available when training new data in the CL setting. The memory-based CL methods leverage a memory buffer to address this problem by storing a limited subset of previous data for replay, and most methods of this type adopt random storage and replay strategies. In the human brain, the hippocampus replays consolidated knowledge from the neocortex in a random manner, e.g., random dreaming. Inspired by this memory mechanism, we propose a memory-based method, which replays more consolidated memory data while maintaining the randomness. Our work highlights that random replaying is important for the CL model, which confirms the effectiveness of random dreaming in the human brain.

Direct saline (seawater) electrolysis is a well-recognized system to generate active chlorine species for the chloride-mediated electrosynthesis, environmental remediation and sterilization over the past few decades. However, the large energy consumption originated from the high cell voltage of traditional direct saline electrolysis system, greatly restricts its practical application. Here, we report an acid-saline hybrid electrolysis system for energy-saving co-electrosynthesis of active chlorine and H2. We demonstrate that this system just requires a low cell voltage of 1.59 V to attain 10 mA cm−2 with a large energy consumption decrease of 27.7 % compared to direct saline electrolysis system (2.20 V). We further demonstrate that such acid-saline hybrid electrolysis system could be extended to realize energy-saving and sustainable seawater electrolysis. The acidified seawater in this system can absolutely avoid the formation of Ca/Mg-based sediments that always form in the seawater electrolysis system. We also prove that this system in the half-flow mode can realize real-time preparation of active chlorine used for sterilization and pea sprout production.

Most existing data center network (DCN) flow scheduling solutions aim to minimize flow completion times (FCT). However, these solutions either require precise flow information (e.g., per-flow size), which is challenging to implement on commodity switches (e.g., pFabric [7]), or no prior flow information at all, which is at the cost of performance (e.g., PIAS [10]). In this work, we present QCLIMB, a new flow scheduling solution designed to minimize FCT by utilizing imprecise flow information. Our key observation is that although obtaining precise flow information can be challenging, it is possible to accurately estimate each flow’s lower and upper bounds with machine learning techniques. QCLIMB has two key parts: i) a novel scheduling algorithm that leverages the lower bounds of different flows to prioritize small flow over large flows from the beginning of transmission, rather than at later stages; and ii) an efficient out-of-order handling mechanism that addresses practical reordering issues resulting from the algorithm. We show that QCLIMB significantly outperforms PIAS (88% lower average FCT of small flows) and is surprisingly close to pFabric (around 9% gap) while not requiring any switch modifications. © 2024 Proceedings of the 21st USENIX Symposium on Networked Systems Design and Implementation, NSDI 2024. All rights reserved.

Though driving automation promises to improve driving safety, drivers are still required to be ready to retake control in conditionally automated vehicles, which are defined by the Society of Automotive Engineers (SAE) as SAE L3 vehicles. Thus, drivers’ states can still affect driving safety in SAE L3 vehicles. Previous research found that a high cognitive load may impair drivers’ takeover performance. Thus, it is still necessary to estimate drivers’ cognitive load in SAE L3 vehicles. However, existing driver cognitive load estimation algorithms mostly focus on vehicles with a lower level of driving automation (e.g., SAE L0), which may not be relevant when estimating driver states in SAE L3 vehicles, given that drivers’ responsibilities are different, and several commonly used measures (e.g., driving performance) are unavailable when drivers are not continuously controlling the vehicle. Further, previous driver cognitive load estimation algorithms rarely considered the temporal information in the input features. Thus, we proposed a deep-learning algorithm to estimate driver cognitive load in SAE L3 vehicles, which integrated multiple physiological features (i.e., electrocardiogram, electrodermal activity, respiration) and considered the temporal correlation of the data using a transformer-encoder-based network. The performance of our algorithm was compared with baseline models on an open data set. Results showed that our algorithm outperformed baseline models and achieved an accuracy of 94.4% using within-subject data partition (proportionally splitting data from the same subject into the training and testing data sets) and an accuracy of 89% using across-subjects data partition (dividing the training and testing data sets based on individual subjects). © The Author(s) 2024.

The traditional methodology utilized in dynamic tracking control synthesis is usually model-based, and therefore, the performance is highly dependent on a precise mathematical model. However, with the growth in system complexity, extremely precise dynamic models for modern robotic and automation systems are very hard to obtain. This challenge has sparked the interest of researchers in moving toward data-driven learning-based concepts, specifically aiming to fully exploit the abundant data available to learn better controls. Along this research direction, continued efforts have been spent on learning a single feedback controller. However, just attempting learning procedures on the feedback controller alone could suffer from severe performance limitations due to its reactive nature, i.e., an error must occur first before any corrective action is taken. In line with this consideration, we propose an integrated two-degree-of-freedom (2-DOF) learning-based tracking control synthesis for high-precision systems, consisting of both feedforward and feedback controllers. Unlike the traditional control design where full knowledge of the dynamics is assumed, we explore the use of actual motion data to iteratively determine the optimal controller parameters using gradient-based optimization. The key step here is to estimate the gradient and Hessian of the cost function without a priori knowledge about the dynamics. Experiments are further conducted based on a prototype flexure-based nanopositioner for spiral scanning applications, demonstrating high-precision nanometer-scale tracking performance that is close to the sensor resolution.

With the development of conditionally automated driving, drivers will be allowed to perform non-driving-related tasks. Under such circumstances, continuous monitoring of driver cognitive load will play an increasingly important role in ensuring that drivers have sufficient mental resources to take over control of the vehicle should the driving automation fail. However, estimation of cognitive load is challenging because of the difficulties in identifying high-level feature representation and accounting for interindividual differences. Physiological measures are believed to be promising candidates for cognitive load estimation in partially automated vehicles. However, current estimation methods are mainly based on the manual feature extraction of time- or frequency-domain indicators from physiological signals, which may not adapt to dynamic driving conditions. With the development of deep learning, the neural network has shown good performance in automatically capturing high-level features from input data. Inspired by this, we adopted a novel approach to classify driver cognitive load based on electrocardiogram (ECG) spectrograms, in which the driver’s ECG signal was collected and transformed into a 2D spectrogram by a short-time Fourier transform. A squeeze-and-excitation network-based deep-learning framework that can capture high-level features and pays more attention to the cognition-related features of the spectrogram was proposed for classification. Experiments on a publicly available dataset demonstrated that our model achieved an accuracy of 96.76% in differentiating two levels of cognitive load for a within-subject evaluation and 71.50% accuracy with an across-subjects evaluation. The results demonstrated the feasibility of detecting drivers’ cognitive load through deep learning using ECG spectrogram alone. © The Author(s) 2024.

This article proposes a Stackelberg game-theoretic optimization approach for adaptive robust controller design of fuzzy underactuated mechanical systems (UMSs). As a commonly encountered challenge, several factors render it difficult to further improve the control performance of these UMSs, such as the coupling effects and nonlinearities resulting from the underactuated structure. Meanwhile, various sources of unexpected time-varying uncertainties can further add to the difficulty in the controller design procedure pipeline. To overcome the above-mentioned issues and further enhance the system performance, a dual-parameter optimization framework is presented here for adaptive robust controller design, leveraging fuzzy set theory and Stackelberg game. In this development, the time-varying uncertainty is characterized via suitable membership functions; and thus the appropriate fuzzy UMS is established. Then, an adaptive robust controller with a leakage-type adaptation law is presented to tackle the uncertainty appropriately. To further improve the system performance considering the trade-off among the transient performance; the steady-state performance; and the control cost; a Stackelberg game-based dual-parameter optimization method is proposed to determine the Stackelberg strategy. It is rigorously proved in the developments here that the proposed controller design and parameters optimization method guarantees the desired deterministic system performance, i.e., uniform boundedness (UB) and uniform ultimate boundedness (UUB). Moreover, a series of numerical simulations based on a flywheel inverted pendulum (FIP) are performed, which illustrate the effectiveness and superiority of the proposed method. © 2024 Elsevier Inc.

The implementation of lift-up buildings has demonstrated their practicality in enhancing urban wind comfort in high-density cities. The pedestrian-level wind environment around a central lift-up building surrounded by low-rise buildings was measured using wind tunnel experiments. The time-averaged velocity field and turbulence statistics, including the turbulent kinetic energy, normal component of the Reynolds stresses, gust factor, and gust wind velocity field, were discussed. Based on the results, the pedestrian-level gust wind flow and comfort for two typical building shapes (Arc, V-shaped) were compared with Rectangular lift-up buildings under different wind directions (0∘, 90∘ and 180∘), the pedestrian-level wind environments (PLWEs) were assessed following two wind criteria for weak and windy wind conditions by wind speed thresholds and maximum allowed exceedance probabilities. The results showed that the erection of the high-rise lift-up building modified the pedestrian-level wind environment around the building groups, owing to the interaction between the downwash flow around the building and the surrounding building blocks. A region of high mean and gust wind velocity was observed at the lateral and rear side of the central lift-up building due to the lateral entrained wind and Venturi effect between elevated columns, indicating that lift-up buildings can enhance PLWEs even in the presence of surrounding buildings. The Arc-shaped lift-up building tends to generate a larger “Intolerable” area compared with V- and Rectangular-shaped buildings due to its curved surface. The wind impacts the convex portion of the building, leading to airflow divergence and facilitating downstream ventilation. © 2024 Elsevier Ltd

Accurate IP geolocation is indispensable for location-aware applications. While recent advances based on router-centric IP graphs are considered cutting-edge, one challenge remain: the prevalence of sparse IP graphs (14.24% with fewer than 10 nodes, 9.73% isolated) limits graph learning. To mitigate this issue, we designate the target host as the central node and aggregate multiple last-hop routers to construct the target-centric IP graph, instead of relying solely on the router with the smallest last-hop latency as in previous works. Experiments on three real-world datasets show that our method significantly improves the geolocation accuracy compared to existing baselines. Copyright © 2024, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.

Accurate IP geolocation is crucial for location-aware applications. While recent advances in router-centric IP graph methods have garnered attention, they face two persistent challenges: (1) the sparsity problem of IP graphs in rural areas and (2) the limited explainability of current IP geolocation systems. To tackle these issues, we present ExGeo, a novel and explainable graph-based approach for IP geolocation. Specifically, we introduce a target-centric IP graph, reducing sparsity and enhancing contextual information utilization. Additionally, we endow the model with explainability through a variational graph information bottleneck strategy. Experiments on three real-world datasets demonstrate significant accuracy and explainability improvements. Source code is released at https://github.com/ICDM-UESTC/ExGeo. © 2024 IEEE.

Observational studies revealed changes in Immunoglobulin G (IgG) N-glycosylation during the aging process. However, it lacks causal insights and remains unclear in which direction causal relationships exist. The two-sample bidirectional Mendelian randomization (MR) design was adopted to explore causal associations between IgG N-glycans and the senescence-associated secretory phenotype (SASP). Inverse variance weighted (IVW) and Wald ratio methods were used as the main analyses, supplemented by sensitivity analyses. Forward MR analyses revealed causal associations between the glycan peak (GP) and SASP, including GP6 (odds ratio [OR] = 0.428, 95% confidence interval [CI] = 0.189–0.969) and GP17 (OR = 0.709, 95%CI = 0.504–0.995) with growth/differentiation factor 15 (GDF15), GP19 with an advanced glycosylation end-product-specific receptor (RAGE) (OR = 2.142, 95% CI = 1.384–3.316), and GP15 with matrix metalloproteinase 2 (MMP2) (OR = 1.136, 95% CI =1.008–1.282). The reverse MR indicated that genetic liability to RAGE was associated with increased levels of GP17 (OR = 1.125, 95% CI = 1.003–1.261) and GP24 (OR = 1.222, 95% CI = 1.046–1.428), while pulmonary and activation-regulated chemokines (PARC) exhibited causal associations with GP10 (OR = 1.269, 95% CI = 1.048–1.537) and GP15 (OR = 1.297, 95% CI = 1.072–1.570). The findings provided suggested evidence on the bidirectional causality between IgG N-glycans and SASP, which might reveal potential regulatory mechanisms.

Background: Neurexins, essential synaptic proteins, are linked to neurodevelopmental and neuropsychiatric disorders like autism spectrum disorder (ASD) and schizophrenia. Objective: Through this systematic review, we aimed to shed light on the relationship between neurexin dysfunction and its implications in neurodevelopmental and neuropsychiatric manifestations. Both animal and human-induced pluripotent stem cell (hiPSC) models served as our primary investigative platforms. Methods: Utilizing the PRISMA 2020 guidelines, our search strategy involved scouring articles from the PubMed and Google Scholar databases covering a span of two decades (2003–2023). Of the initial collection, 27 rigorously evaluated studies formed the essence of our review. Results: Our review suggested the significant ties between neurexin anomalies and neurodevelopmental and neuropsychiatric outcomes, most notably ASD. Rodent-based investigations delineated pronounced ASD-associated behaviors, and hiPSC models derived from ASD-diagnosed patients revealed the disruptions in calcium dynamics and synaptic activities. Additionally, our review underlined the integral role of specific neurexin variants, primarily NRXN1, in the pathology of schizophrenia. It was also evident from our observation that neurexin malfunctions were implicated in a broader array of these disorders, including ADHD, intellectual challenges, and seizure disorders. Conclusion: This review accentuates the cardinal role neurexins play in the pathological process of neurodevelopmental and neuropsychiatric disorders. The findings underscore a critical need for standardized methodologies in developing animal and hiPSC models for future studies, aiming to minimize heterogeneity. Moreover, we highlight the need to expand research into less studied neurexin variants (i.e., NRXN2 and NRXN3), broadening the scope of our understanding in this field. Our observation also projects hiPSC models as potent tools for bridging research gaps, p

Background: Globally, there is increasing interest in the use of real-world data (RWD) and real-world evidence (RWE) to inform health technology assessment (HTA) and reimbursement decision-making. Using current practices and case studies shared by eleven health systems in Asia, a non-binding guidance that seeks to align practices for generating and using RWD/RWE for decision-making in Asia was developed by the REAL World Data In ASia for HEalth Technology Assessment in Reimbursement (REALISE) Working Group, addressing a current gap and needs among HTA users and generators. Methods: The guidance document was developed over two face-to-face workshops, in addition to an online survey, a face-to-face interview and pragmatic search of literature. The specific focus was on what, where and how to collect RWD/RWE. Results: All 11 REALISE member jurisdictions participated in the online survey and the first in-person workshop, 10 participated in the second in-person workshop, and 8 participated in the in-depth face-to-face interviews. The guidance document was iteratively reviewed by all working group members and the International Advisory Panel. There was substantial variation in: (a) sources and types of RWD being used in HTA, and (b) the relative importance and prioritization of RWE being used for policy-making. A list of national-level databases and other sources of RWD available in each country was compiled. A list of useful guidance on data collection, quality assurance and study design were also compiled. Conclusion: The REALISE guidance document serves to align the collection of better quality RWD and generation of reliable RWE to ultimately inform HTA in Asia.

For various typical cases and situations where the formulation results in an optimal control problem, the linear quadratic regulator (LQR) approach and its variants continue to be highly attractive. In certain scenarios, it can happen that some prescribed structural constraints on the gain matrix would arise. Consequently then, the algebraic Riccati equation (ARE) is no longer applicable in a straightforward way to obtain the optimal solution. This work presents a rather effective alternative optimization approach based on gradient projection. The utilized gradient is obtained through a data-driven methodology, and then projected onto applicable constrained hyperplanes. Essentially, this projection gradient determines a direction of progression and computation for the gain matrix update with a decreasing functional cost; and then the gain matrix is further refined in an iterative framework. With this formulation, a data-driven optimization algorithm is summarized for controller synthesis with structural constraints. This data-driven approach has the key advantage that it avoids the necessity of precise modeling which is always required in the classical model-based counterpart; and thus the approach can additionally accommodate various model uncertainties. Illustrative examples are also provided in the work to validate the theoretical results. IEEE

To support in vivo and in vitro studies of intravascular triglyceride metabolism in mice, we created rat monoclonal antibodies (mAbs) against mouse LPL. Two mAbs, mAbs 23A1 and 31A5, were used to develop a sandwich ELISA for mouse LPL. The detection of mouse LPL by the ELISA was linear in concentrations ranging from 0.31 ng/ml to 20 ng/ml. The sensitivity of the ELISA made it possible to quantify LPL in serum and in both pre-heparin and post-heparin plasma samples (including in grossly lipemic samples). LPL mass and activity levels in the post-heparin plasma were lower in Gpihbp1-/- mice than in wild-type mice. In both groups of mice, LPL mass and activity levels were positively correlated. Our mAb-based sandwich ELISA for mouse LPL will be useful for any investigator who uses mouse models to study LPL-mediated intravascular lipolysis.

Background: Immunoglobulin G (IgG) N-glycosylation is considered a potential biomarker for aging and various pathological conditions. However, whether these changes in IgG N-glycosylation are a consequence or a contributor to the aging process remains unclear. This study aims to investigate the causality between IgG N-glycosylation and aging using Mendelian randomization (MR) analysis. Methods: We utilized genetic variants associated with IgG N-glycosylation traits, the frailty index (FI), and leukocyte telomere length (LTL) from a previous genome-wide association study (GWAS) on individuals of European ancestry. Two-sample and multivariable MR analyses were conducted, employing the inverse-variance weighted (IVW) method. Sensitivity analyses were performed to assess potential confounding factors. Results: Using the IVW method, we found suggestive evidence of a causal association between GP14 and FI ( 0.026, 95% CI 0.003 to 0.050, p = 0.027) and LTL ( −0.020, 95% CI −0.037 to −0.002, p = 0.029) in the two-sample MR analysis. In the multivariable MR analysis, suggestive evidence was found for GP23 and FI ( −0.119, 95% CI −0.219 to −0.019, p = 0.019) and GP2 and LTL ( 0.140, 95% CI 0.020 to 0.260, p = 0.023). Conclusions: In conclusion, our results supported a potentially causal effect of lower GP23 levels on an advanced aging state. Additional verification is required to further substantiate the causal relationship between glycosylation and aging.

The particle acceleration of blazar jets is crucial to high-energy astrophysics, yet the acceleration mechanism division in blazar subclasses and the underlying nature of these mechanisms remain elusive. In this work, we utilized the synchrotron spectral information (synchrotron peak frequency, $\log \nu_{\rm sy}$, and corresponding curvature, $b_{\rm sy}$) of 2705 blazars from the literature and studied the subject of particle acceleration in blazar jets by analysing the correlation between $\log \nu_{\rm sy}$ and $1/b_{\rm sy}$. Our results suggested that the entire sample follows an energy-dependent probability acceleration (EDPA). Specifically, the low inverse Compton peak sources (LCPs) follow the mechanism that fluctuations of fractional gain acceleration (FFGA), while the high inverse Compton peak sources (HCPs) follow an acceleration mechanism of EDPA. Our results indicated that the separation between LCPs and HCPs results from the electron peak Lorentz factor ($\gamma_{\rm p}$), and the differentiation should originate from different acceleration mechanisms. Moreover, our study revealed a transition in the acceleration mechanism from FFGA to EDPA around $\log \nu_{\rm sy} \sim 15$ through a detailed analysis of binned-$\log \nu_{\rm sy}$. The mechanism of FFGA dominates the particle acceleration in LCP jets because of stronger jets and the EDPA dominates the particle energy gain in the HCPs due to a more efficient acceleration process., Comment: Accepted to ApJ

Transcription factor MAFB regulates various homeostatic functions of macrophages. This study explores the role of MAFB in brown adipose tissue (BAT) thermogenesis using macrophage-specific Mafb-deficient (Mafbf/f::LysM-Cre) mice. We find that Mafb deficiency in macrophages reduces thermogenesis, energy expenditure, and sympathetic neuron (SN) density in BAT under cold conditions. This phenotype features a proinflammatory environment that is characterized by macrophage/granulocyte accumulation, increases in interleukin-6 (IL-6) production, and IL-6 trans-signaling, which lead to decreases in nerve growth factor (NGF) expression and reduction in SN density in BAT. We confirm MAFB regulation of IL-6 expression using luciferase readout driven by IL-6 promoter in RAW-264.7 macrophage cell lines. Immunohistochemistry shows clustered organization of NGF-producing cells in BAT, which are primarily TRPV1+ vascular smooth muscle cells, as additionally shown using single-cell RNA sequencing and RT-qPCR of the stromal vascular fraction. Treating Mafbf/f::LysM-Cre mice with anti-IL-6 receptor antibody rescues SN density, body temperature, and energy expenditure.

The genus Bifidobacterium widely exists in human gut and has been increasingly used as the adjuvant probiotics for the prevention and treatment of diseases. However, the functional differences of Bifidobacterium genomes from different regions of the world remain unclear. We here describe an extensive study on the genomic characteristics and function annotations of 1512 genomes (clustered to 849 non-redundant genomes) of Bifidobacterium cultured from human gut. The distribution of some carbohydrate-active enzymes varied among different Bifidobacterium species and continents. More than 36% of the genomes of B. pseudocatenulatum harbored biosynthetic gene clusters of lanthipeptide-class-iv. 99.76% of the cultivated genomes of Bifidobacterium harbored genes of bile salt hydrolase. Most genomes of B. adolescentis, and all genomes of B. dentium harbored genes involved in gamma-aminobutyric acid synthesis. B. longum subsp. infantis were characterized harboring most genes related to human milk oligosaccharide utilization. Significant differences between the distribution of antibiotic resistance genes among different species and continents revealed the importance to use antibiotics precisely in the clinical treatment. Phages infecting Bifidobacterium and horizontal gene transfers occurring in genomes of Bifidobacterium were dependent on species and region sources, and might help Bifidobacterium adapt to the environment. In addition, the distribution of Bifidobacterium in human gut was found varied from different regions of the world. This study represents a comprehensive view of characteristics and functions of genomes of cultivated Bifidobacterium from human gut, and enables clinical advances in the future., The genus Bifidobacterium widely exists in human gut and has been increasingly used as the adjuvant probiotics for the prevention and treatment of diseases. However, the functional differences of Bifidobacterium genomes from different regions of the world remain unclear. We here describe an extensive study on the genomic characteristics and function annotations of 1512 genomes (clustered to 849 non-redundant genomes) of Bifidobacterium cultured from human gut. The distribution of some carbohydrate-active enzymes varied among different Bifidobacterium species and continents. More than 36% of the genomes of B. pseudocatenulatum harbored biosynthetic gene clusters of lanthipeptide-class-iv. 99.76% of the cultivated genomes of Bifidobacterium harbored genes of bile salt hydrolase. Most genomes of B. adolescentis, and all genomes of B. dentium harbored genes involved in gamma-aminobutyric acid synthesis. B. longum subsp. infantis were characterized harboring most genes related to human milk oligosaccharide utilization. Significant differences between the distribution of antibiotic resistance genes among different species and continents revealed the importance to use antibiotics precisely in the clinical treatment. Phages infecting Bifidobacterium and horizontal gene transfers occurring in genomes of Bifidobacterium were dependent on species and region sources, and might help Bifidobacterium adapt to the environment. In addition, the distribution of Bifidobacterium in human gut was found varied from different regions of the world. This study represents a comprehensive view of characteristics and functions of genomes of cultivated Bifidobacterium from human gut, and enables clinical advances in the future.

The Illumina HiSeq platform has been a commonly used option for bacterial genome sequencing. Now the BGI DNA nanoball (DNB) nanoarrays platform may provide an alternative platform for sequencing of bacterial genomes. To explore the impact of sequencing platforms on bacterial genome assembly, quality assessment, sequence alignment, functional annotation, mutation detection, and metagenome mapping, we compared genome assemblies based on sequencing of cultured bacterial species using the HiSeq 2000 and BGISEQ-500 platforms. In addition, simulated reads were used to evaluate the impact of insert size on genome assembly. Genome assemblies based on BGISEQ-500 sequencing exhibited higher completeness and fewer N bases in high GC genomes, whereas HiSeq 2000 assemblies exhibited higher N50. The majority of assembly assessment parameters, sequences of 16S rRNA genes and genomes, numbers of single nucleotide variants (SNV), and mapping to metagenome data did not differ significantly between platforms. More insertions were detected in HiSeq 2000 genome assemblies, whereas more deletions were detected in BGISEQ-500 genome assemblies. Insert size had no significant impact on genome assembly. Taken together, our results suggest that DNBSEQ platforms would be a valid substitute for HiSeq 2000 for bacterial genome sequencing., The Illumina HiSeq platform has been a commonly used option for bacterial genome sequencing. Now the BGI DNA nanoball (DNB) nanoarrays platform may provide an alternative platform for sequencing of bacterial genomes. To explore the impact of sequencing platforms on bacterial genome assembly, quality assessment, sequence alignment, functional annotation, mutation detection, and metagenome mapping, we compared genome assemblies based on sequencing of cultured bacterial species using the HiSeq 2000 and BGISEQ-500 platforms. In addition, simulated reads were used to evaluate the impact of insert size on genome assembly. Genome assemblies based on BGISEQ-500 sequencing exhibited higher completeness and fewer N bases in high GC genomes, whereas HiSeq 2000 assemblies exhibited higher N50. The majority of assembly assessment parameters, sequences of 16S rRNA genes and genomes, numbers of single nucleotide variants (SNV), and mapping to metagenome data did not differ significantly between platforms. More insertions were detected in HiSeq 2000 genome assemblies, whereas more deletions were detected in BGISEQ-500 genome assemblies. Insert size had no significant impact on genome assembly. Taken together, our results suggest that DNBSEQ platforms would be a valid substitute for HiSeq 2000 for bacterial genome sequencing.

Fenge Li,1,2,&ast; Yupeng Wang,1,&ast; Jin Yan,1,&ast; Huancheng Wu,3 Xueming Du,1 Weihong Feng,1 Xiaoqing Zhang,4 Yongming Xue,4 Huaqing Wang,5,6 Wenxin Liu7 1Department of Oncology, Tianjin Beichen Hospital Tianjin People’s Republic of China; 2Core Laboratory, Tianjin Beichen Hospital, Tianjin, People’s Republic of China; 3Department of Neurosurgery, Tianjin Beichen Hospital, Tianjin, People’s Republic of China; 4Department of Basic Research, Suzhou Lanma Biotechnology Co, Suzhou, People’s Republic of China; 5Department of Oncology, Tianjin Union Medical Center, Tianjin, People’s Republic of China; 6Department of Translational Medicine, Tianjin Union Medical Center of Nankai University, Tianjin, People’s Republic of China; 7Department of Gynecological Oncology, Tianjin Medical Cancer Institute and Hospital, National Clinical Research Center for Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin, People’s Republic of China&ast;These authors contributed equally to this workCorrespondence: Wenxin Liu, Department of Gynecological Oncology, Tianjin Medical Cancer Institute and Hospital, Huanhuxi Road, Tiyuanbei, Hexi District, Tianjin, People’s Republic of China, Tel +8618622221101, Email wenxin1973@163.com Huaqing Wang, Department of Translational Medicine, Tianjin Union Medical Center of Nankai University, Jieyuan Road No. 190, Nankai District, Tianjin, People’s Republic of China, Tel +8618622221223, Email huaqingw@163.comIntroduction: Tumor-infiltrating lymphocytes (TILs) therapy is one of the most promising adoptive T cell therapies, which has shown great clinical efficacy against several solid malignancies. Nevertheless, clinical response to TILs mono-therapy in Asian patients with recurrent cervical cancer has not been well reported.Case Presentation: Here, we report two patients who were diagnosed with metastatic cervical cancer and tumor progression following multiple conventional treatments. In particular, one of

Aims To investigate the physiological nyctohemeral intraocular pressure (IOP) rhythms of normal Chinese adults using a novel contact lens sensor system (CLS) that can output IOP in millimetres of mercury (mm Hg) continuously. Methods Fifty-nine eyes of 59 normal Chinese adults completed 24-hour IOP monitoring using the novel CLS. A descriptive analysis was conducted on the 24-hour IOP mean, peak and acrophase, trough and bathyphase, fluctuation, and mean amplitude of intraocular pressure excursion (MAPE). The continuous data were analysed at several periods (diurnal period, 08:00-20:00 hours; nocturnal period, 22:00-06:00 hours; sleep time, 0:00-06:00 hours), and compared between right and left eyes, males and females, and different age ranges (<30, and >= 30), respectively. Results Normal adults had a lower peak, higher trough, smaller fluctuation and smaller MAPE (p<0.05 for all comparisons) but non-significantly different mean (p=0.695) in the nocturnal period or sleep time compared with the diurnal period. The 24-hour IOP peak and trough showed the frequency of occurrence ranging from 1.69% to 15.25% at an interval of 2 hours. No IOP parameter showed significant difference between right and left eyes (p>0.1 for all comparisons). The male group had larger 24-hour and diurnal IOP fluctuation and MAPE (p<0.05 for all comparisons). Subjects aged 30 or over had higher 24-hour and diurnal mean, higher peak, and larger MAPE (p<0.05 for all comparisons). Conclusion Continuous 24-hour IOP output from the CLS in normal Chinese was stable with a comparable mean level between day and night, as well as scattered acrophase and bathyphase. The 24-hour IOP mean increased with age, and IOP variations were positively correlated to age and male sex.

This work presents a post-data analysis procedure, namely, proper orthogonal decomposition (POD)–dynamic mode decomposition (DMD)–discrete Fourier transform analysis, for evaluating the dominant features of the flow fields from both energetic and dynamic perspectives. The large-eddy simulation (LES) was first employed to reproduce the flow field surrounding three types of building layouts. Subsequently, both POD and DMD were conducted according to LES simulation results. The extracted modes were classified into three types based on the POD and DMD: Type-1 mode: energetically and dynamically significant mode, Type-2 mode: energetically significant and dynamically insignificant mode, and Type-3 mode: energetically insignificant and dynamically significant mode. The findings indicate that Type-1 mode governs the primary velocity field and the predominant vortex patterns observed at the rear of the building arrays, as the reduction of inter-building widths leads to a shorter flow separation region. Type-2 mode is characterized by the presence of small-scale vortices and the high turbulent kinetic energy region, which periodically triggers pollutant increase in the vicinity of structures. Type-3 mode demonstrates a minimal energetic influence on the flow field; nevertheless, it significantly contributes to the consistent build-up of pollutants within the far-wake region. The present study also investigates the predominant coherent structures of flow fields concerning various building layouts and highlights the influence of passage widths on the efficiency of pollutant removal. This comprehensive analysis enables a systematic exploration of flow patterns within various building layouts, offering potential solutions for pollutant dispersion challenges in metropolitan areas.

Humans learn social skills through both imitation and social interaction. This social learning process is largely understudied by existing research on building language agents. Motivated by this gap, we propose an interactive learning method, SOTOPIA-$\pi$, improving the social intelligence of language agents. This method leverages behavior cloning and self-reinforcement training on filtered social interaction data according to large language model (LLM) ratings. We show that our training method allows a 7B LLM to reach the social goal completion ability of an expert model (GPT-4-based agent), while improving the safety of language agents and maintaining general QA ability on the MMLU benchmark. We also find that this training paradigm uncovers some difficulties in LLM-based evaluation of social intelligence: LLM-based evaluators overestimate the abilities of the language agents trained specifically for social interaction.

In the pursuit of transferring a source model to a target domain without access to the source training data, Source-Free Domain Adaptation (SFDA) has been extensively explored across various scenarios, including closed-set, open-set, partial-set, and generalized settings. Existing methods, focusing on specific scenarios, not only address only a subset of challenges but also necessitate prior knowledge of the target domain, significantly limiting their practical utility and deployability. In light of these considerations, we introduce a more practical yet challenging problem, termed unified SFDA, which comprehensively incorporates all specific scenarios in a unified manner. To tackle this unified SFDA problem, we propose a novel approach called Latent Causal Factors Discovery (LCFD). In contrast to previous alternatives that emphasize learning the statistical description of reality, we formulate LCFD from a causality perspective. The objective is to uncover the causal relationships between latent variables and model decisions, enhancing the reliability and robustness of the learned model against domain shifts. To integrate extensive world knowledge, we leverage a pre-trained vision-language model such as CLIP. This aids in the formation and discovery of latent causal factors in the absence of supervision in the variation of distribution and semantics, coupled with a newly designed information bottleneck with theoretical guarantees. Extensive experiments demonstrate that LCFD can achieve new state-of-the-art results in distinct SFDA settings, as well as source-free out-of-distribution generalization.Our code and data are available at https://github.com/tntek/source-free-domain-adaptation.

This work investigates a game-theoretic optimization approach toward robust control of uncertain dynamical systems with state and input constraints. The uncertainty involved is possibly rapidly time-varying but bounded within a prescribed fuzzy set. For this, the associated fuzzy dynamical system is appropriately established and constructed based on fuzzy set theory. To cope with the bounded state and input constraints, a novel state-and-input diffeomorphism technique is proposed, where a transformed system is formulated such that the prescribed inequality constraints are innovatively merged into the stabilization and trajectory tracking problems. Furthermore, a diffeomorphism-based robust control strategy is developed to ensure the uniform boundedness and uniform ultimate boundedness of the transformed system. Under this proposed control architecture, the constraint satisfaction of the original system is, thus, always analytically ensured based on the rigorous properties of the diffeomorphism technique. The resulting control parameter optimization problem then has to take into account the multiple considerations (and compromise) amongst the factors of the steady-state performance; the finite convergence time, and the control effort. For this, a two-player Nash game is formulated and solved in an effective manner. The Nash equilibrium is obtained and the existence of the solution is also proved theoretically. With this methodology, and with the resulting attainment of the desired Nash equilibrium, the attendant outcome of superior system performance is achieved. Finally, numerical simulations on a steer-by-wire system demonstrate the effectiveness of the proposed approach.

Anecdotal evidence of cannabis use by professional programmers abounds. Recent studies have found that some professionals regularly use cannabis while programming even for work-related tasks. However, accounts of the impacts of cannabis on programming vary widely and are often contradictory. For example, some programmers claim that it impairs their ability to generate correct solutions while others claim it enhances creativity and focus. There remains a need for an empirical understanding of the true impacts of cannabis on programming. This paper presents the first controlled observational study of the effects of cannabis on programming ability. Based on a within-subjects design with over 70 participants, we find that at ecologically valid dosages, cannabis significantly impairs programming performance. Programs implemented while high contain more bugs and take longer to write (p < 0.05), a small to medium effect (0.22 <= d <= 0.44). We also did not find any evidence that high programmers generate more divergent solutions. However, programmers can accurately assess differences in their programming performance (r = 0.59), even when under the influence of cannabis. We hope that this research will facilitate evidence-based policies and help developers make informed decisions regarding cannabis use while programming., Comment: To appear in the proceedings of the International Conference of Software Engineering (ICSE), 2024

When completing knowledge-intensive tasks, humans sometimes need not just an answer but also a corresponding reference passage for auxiliary reading. Previous methods required obtaining pre-segmented article chunks through additional retrieval models. This paper explores leveraging the parameterized knowledge stored during the pre-training phase of large language models (LLMs) to independently recall reference passage from any starting position. We propose a two-stage framework that simulates the scenario of humans recalling easily forgotten references. Initially, the LLM is prompted to recall document title identifiers to obtain a coarse-grained document set. Then, based on the acquired coarse-grained document set, it recalls fine-grained passage. In the two-stage recall process, we use constrained decoding to ensure that content outside of the stored documents is not generated. To increase speed, we only recall a short prefix in the second stage, then locate its position to retrieve a complete passage. Experiments on KILT knowledge-sensitive tasks have verified that LLMs can independently recall reference passage location in various task forms, and the obtained reference significantly assist downstream tasks.

Accuracy and efficiency remain challenges for multi-party computation (MPC) frameworks. Spin is a GPU-accelerated MPC framework that supports multiple computation parties and a dishonest majority adversarial setup. We propose optimized protocols for non-linear functions that are critical for machine learning, as well as several novel optimizations specific to attention that is the fundamental unit of Transformer models, allowing Spin to perform non-trivial CNNs training and Transformer inference without sacrificing security. At the backend level, Spin leverages GPU, CPU, and RDMA-enabled smart network cards for acceleration. Comprehensive evaluations demonstrate that Spin can be up to $2\times$ faster than the state-of-the-art for deep neural network training. For inference on a Transformer model with 18.9 million parameters, our attention-specific optimizations enable Spin to achieve better efficiency, less communication, and better accuracy.

During the evolution of large models, performance evaluation is necessarily performed to assess their capabilities and ensure safety before practical application. However, current model evaluations mainly rely on specific tasks and datasets, lacking a united framework for assessing the multidimensional intelligence of large models. In this perspective, we advocate for a comprehensive framework of cognitive science-inspired artificial general intelligence (AGI) tests, aimed at fulfilling the testing needs of large models with enhanced capabilities. The cognitive science-inspired AGI tests encompass the full spectrum of intelligence facets, including crystallized intelligence, fluid intelligence, social intelligence, and embodied intelligence. To assess the multidimensional intelligence of large models, the AGI tests consist of a battery of well-designed cognitive tests adopted from human intelligence tests, and then naturally encapsulates into an immersive virtual community. We propose increasing the complexity of AGI testing tasks commensurate with advancements in large models and emphasizing the necessity for the interpretation of test results to avoid false negatives and false positives. We believe that cognitive science-inspired AGI tests will effectively guide the targeted improvement of large models in specific dimensions of intelligence and accelerate the integration of large models into human society.

The development and training of deep learning models have become increasingly costly and complex. Consequently, software engineers are adopting pre-trained models (PTMs) for their downstream applications. The dynamics of the PTM supply chain remain largely unexplored, signaling a clear need for structured datasets that document not only the metadata but also the subsequent applications of these models. Without such data, the MSR community cannot comprehensively understand the impact of PTM adoption and reuse. This paper presents the PeaTMOSS dataset, which comprises metadata for 281,638 PTMs and detailed snapshots for all PTMs with over 50 monthly downloads (14,296 PTMs), along with 28,575 open-source software repositories from GitHub that utilize these models. Additionally, the dataset includes 44,337 mappings from 15,129 downstream GitHub repositories to the 2,530 PTMs they use. To enhance the dataset's comprehensiveness, we developed prompts for a large language model to automatically extract model metadata, including the model's training datasets, parameters, and evaluation metrics. Our analysis of this dataset provides the first summary statistics for the PTM supply chain, showing the trend of PTM development and common shortcomings of PTM package documentation. Our example application reveals inconsistencies in software licenses across PTMs and their dependent projects. PeaTMOSS lays the foundation for future research, offering rich opportunities to investigate the PTM supply chain. We outline mining opportunities on PTMs, their downstream usage, and cross-cutting questions., Comment: Accepted at MSR'24

Objective: Subcutaneous Immunotherapy (SCIT) is the long-lasting causal treatment of allergic rhinitis. How to enhance the adherence of patients to maximize the benefit of allergen immunotherapy (AIT) plays a crucial role in the management of AIT. This study aims to leverage novel machine learning models to precisely predict the risk of non-adherence of patients and related systematic symptom scores, to provide a novel approach in the management of long-term AIT. Methods: The research develops and analyzes two models, Sequential Latent Actor-Critic (SLAC) and Long Short-Term Memory (LSTM), evaluating them based on scoring and adherence prediction capabilities. Results: Excluding the biased samples at the first time step, the predictive adherence accuracy of the SLAC models is from $60\,\%$ to $72\%$, and for LSTM models, it is $66\,\%$ to $84\,\%$, varying according to the time steps. The range of Root Mean Square Error (RMSE) for SLAC models is between $0.93$ and $2.22$, while for LSTM models it is between $1.09$ and $1.77$. Notably, these RMSEs are significantly lower than the random prediction error of $4.55$. Conclusion: We creatively apply sequential models in the long-term management of SCIT with promising accuracy in the prediction of SCIT nonadherence in Allergic Rhinitis (AR) patients. While LSTM outperforms SLAC in adherence prediction, SLAC excels in score prediction for patients undergoing SCIT for AR. The state-action-based SLAC adds flexibility, presenting a novel and effective approach for managing long-term AIT.

As the deployment of deep learning models continues to expand across industries, the threat of malicious incursions aimed at gaining access to these deployed models is on the rise. Should an attacker gain access to a deployed model, whether through server breaches, insider attacks, or model inversion techniques, they can then construct white-box adversarial attacks to manipulate the model's classification outcomes, thereby posing significant risks to organizations that rely on these models for critical tasks. Model owners need mechanisms to protect themselves against such losses without the necessity of acquiring fresh training data - a process that typically demands substantial investments in time and capital. In this paper, we explore the feasibility of generating multiple versions of a model that possess different attack properties, without acquiring new training data or changing model architecture. The model owner can deploy one version at a time and replace a leaked version immediately with a new version. The newly deployed model version can resist adversarial attacks generated leveraging white-box access to one or all previously leaked versions. We show theoretically that this can be accomplished by incorporating parameterized hidden distributions into the model training data, forcing the model to learn task-irrelevant features uniquely defined by the chosen data. Additionally, optimal choices of hidden distributions can produce a sequence of model versions capable of resisting compound transferability attacks over time. Leveraging our analytical insights, we design and implement a practical model versioning method for DNN classifiers, which leads to significant robustness improvements over existing methods. We believe our work presents a promising direction for safeguarding DNN services beyond their initial deployment., Comment: Published in IEEE SaTML 2024

Brain tumors are the most common solid tumors and the leading cause of cancer-related death among children. Tumor segmentation is essential in surgical and treatment planning, and response assessment and monitoring. However, manual segmentation is time-consuming and has high inter-operator variability, underscoring the need for more efficient methods. We compared two deep learning-based 3D segmentation models, DeepMedic and nnU-Net, after training with pediatric-specific multi-institutional brain tumor data using based on multi-parametric MRI scans.Multi-parametric preoperative MRI scans of 339 pediatric patients (n=293 internal and n=46 external cohorts) with a variety of tumor subtypes, were preprocessed and manually segmented into four tumor subregions, i.e., enhancing tumor (ET), non-enhancing tumor (NET), cystic components (CC), and peritumoral edema (ED). After training, performance of the two models on internal and external test sets was evaluated using Dice scores, sensitivity, and Hausdorff distance with reference to ground truth manual segmentations. Dice score for nnU-Net internal test sets was (mean +/- SD (median)) 0.9+/-0.07 (0.94) for WT, 0.77+/-0.29 for ET, 0.66+/-0.32 for NET, 0.71+/-0.33 for CC, and 0.71+/-0.40 for ED, respectively. For DeepMedic the Dice scores were 0.82+/-0.16 for WT, 0.66+/-0.32 for ET, 0.48+/-0.27, for NET, 0.48+/-0.36 for CC, and 0.19+/-0.33 for ED, respectively. Dice scores were significantly higher for nnU-Net (p<=0.01). External validation of the trained nnU-Net model on the multi-institutional BraTS-PEDs 2023 dataset revealed high generalization capability in segmentation of whole tumor and tumor core with Dice scores of 0.87+/-0.13 (0.91) and 0.83+/-0.18 (0.89), respectively. Pediatric-specific data trained nnU-Net model is superior to DeepMedic for whole tumor and subregion segmentation of pediatric brain tumors.

Diffusion Magnetic Resonance Imaging (dMRI) plays a crucial role in the noninvasive investigation of tissue microstructural properties and structural connectivity in the \textit{in vivo} human brain. However, to effectively capture the intricate characteristics of water diffusion at various directions and scales, it is important to employ comprehensive q-space sampling. Unfortunately, this requirement leads to long scan times, limiting the clinical applicability of dMRI. To address this challenge, we propose SSOR, a Simultaneous q-Space sampling Optimization and Reconstruction framework. We jointly optimize a subset of q-space samples using a continuous representation of spherical harmonic functions and a reconstruction network. Additionally, we integrate the unique properties of diffusion magnetic resonance imaging (dMRI) in both the q-space and image domains by applying $l1$-norm and total-variation regularization. The experiments conducted on HCP data demonstrate that SSOR has promising strengths both quantitatively and qualitatively and exhibits robustness to noise.

Large Language Models (LLMs) have exhibited remarkable capabilities in understanding and interacting with natural language across various sectors. However, their effectiveness is limited in specialized areas requiring high accuracy, such as plant science, due to a lack of specific expertise in these fields. This paper introduces PLLaMa, an open-source language model that evolved from LLaMa-2. It's enhanced with a comprehensive database, comprising more than 1.5 million scholarly articles in plant science. This development significantly enriches PLLaMa with extensive knowledge and proficiency in plant and agricultural sciences. Our initial tests, involving specific datasets related to plants and agriculture, show that PLLaMa substantially improves its understanding of plant science-related topics. Moreover, we have formed an international panel of professionals, including plant scientists, agricultural engineers, and plant breeders. This team plays a crucial role in verifying the accuracy of PLLaMa's responses to various academic inquiries, ensuring its effective and reliable application in the field. To support further research and development, we have made the model's checkpoints and source codes accessible to the scientific community. These resources are available for download at \url{https://github.com/Xianjun-Yang/PLLaMa}., Comment: Work in progress

Pre-trained code models have recently achieved substantial improvements in many code intelligence tasks. These models are first pre-trained on large-scale unlabeled datasets in a task-agnostic manner using self-supervised learning, and then fine-tuned on labeled datasets in downstream tasks. However, the labeled datasets are usually limited in size (i.e., human intensive efforts), which may hinder the performance of pre-trained code models in specific tasks. To mitigate this, one possible solution is to leverage the large-scale unlabeled data in the tuning stage by pseudo-labeling. However, directly employing the pseudo-labeled data can bring a large amount of noise, i.e., incorrect labels, leading to suboptimal performance. How to effectively leverage the noisy pseudo-labeled data is a challenging yet under-explored problem.In this paper, we propose a novel approach named HINT to improve pre-trained code models with large-scale unlabeled datasets by better utilizing the pseudo-labeled data. HINT includes two main modules: HybrId pseudo-labeled data selection and Noise-tolerant Training. In the hybrid pseudo-data selection module, considering the robustness issue, apart from directly measuring the quality of pseudo labels through training loss, we further propose to employ a retrieval-based method to filter low-quality pseudo-labeled data. The noise-tolerant training module aims to further mitigate the influence of errors in pseudo labels by training the model with a noise-tolerant loss function and by regularizing the consistency of model predictions.The experimental results show that HINT can better leverage those unlabeled data in a task-specific way and provide complementary benefits for pre-trained models, e.g., improving the best baseline model by 15.33%, 16.50%, and 8.98% on code summarization, defect detection, and assertion generation, respectively., Comment: Accepted by ICSE 2024

Deep learning has shown great potential in accelerating diffusion tensor imaging (DTI). Nevertheless, existing methods tend to suffer from Rician noise and detail loss in reconstructing the DTI-derived parametric maps especially when sparsely sampled q-space data are used. This paper proposes a novel method, AID-DTI (Accelerating hIgh fiDelity Diffusion Tensor Imaging), to facilitate fast and accurate DTI with only six measurements. AID-DTI is equipped with a newly designed Singular Value Decomposition (SVD)-based regularizer, which can effectively capture fine details while suppressing noise during network training. Experimental results on Human Connectome Project (HCP) data consistently demonstrate that the proposed method estimates DTI parameter maps with fine-grained details and outperforms three state-of-the-art methods both quantitatively and qualitatively.

Managing privacy to reach privacy goals is challenging, as evidenced by the privacy attitude-behavior gap. Mitigating this discrepancy requires solutions that account for both system opaqueness and users’ hesitations in testing diferent privacy settings due to fears of unintended data exposure.We introduce an empathy-based approach that allows users to experience how privacy attributes may alter system outcomes in a risk-free sandbox environment from the perspective of artifcially generated personas. To generate realistic personas, we introduce a novel pipeline that augments the outputs of large language models (e.g., GPT-4) using few-shot learning, contextualization, and chain of thoughts. Our empirical studies demonstrated the adequate quality of generated personas and highlighted the changes in privacy-related applications (e.g., online advertising) caused by diferent personas. Furthermore, users demonstrated cognitive and emotional empathy towards the personas when interacting with our sandbox. We ofered design implications for downstream applications in improving user privacy literacy.

We introduce a low-resource safety enhancement method for aligning large language models (LLMs) without the need for supervised fine-tuning (SFT) or reinforcement learning from human feedback (RLHF). Our main idea is to exploit knowledge distillation to extract the alignment information from existing well-aligned LLMs and integrate it into unaligned LLMs in a plug-and-play fashion. Methodology, we employ delta debugging to identify the critical components of knowledge necessary for effective distillation. On the harmful question dataset, our method significantly enhances the average defense success rate by approximately 14.41%, reaching as high as 51.39%, in 17 unaligned pre-trained LLMs, without compromising performance.

Source-free Domain Adaptation (SFDA) aims to adapt a pre-trained source model to an unlabeled target domain with no access to the source data. Inspired by the success of pre-trained large vision-language (ViL) models in many other applications, the latest SFDA methods have also validated the benefit of ViL models by leveraging their predictions as pseudo supervision. However, we observe that ViL's predictions could be noisy and inaccurate at an unknown rate, potentially introducing additional negative effects during adaption. To address this thus-far ignored challenge, in this paper, we introduce a novel Proxy Denoising (ProDe) approach. Specifically, we leverage the ViL model as a proxy to facilitate the adaptation process towards the latent domain-invariant space. Critically, we design a proxy denoising mechanism for correcting ViL's predictions. This is grounded on a novel proxy confidence theory by modeling elegantly the domain adaption effect of the proxy's divergence against the domain-invariant space. To capitalize the corrected proxy, we further derive a mutual knowledge distilling regularization. Extensive experiments show that our ProDe significantly outperforms the current state-of-the-art alternatives under both conventional closed-set setting and the more challenging open-set, partial-set and generalized SFDA settings. The code will release soon.