To enhance the micromixing in chemical processes, a jet-stirred tank reactor (JSTR) was developed by integrating an annular jet with the width of 0.38 mm and 0.60 mm into a 81 L stirred tank to provide localized energy intensification. The iodide-iodate reaction method was used to evaluate the micromixing performance under different operating and structural conditions. Results showed that the micromixing time determined by the incorporation model decreased with increasing impeller speed and jet width, while the feeding position and jet velocity can significantly influence the micromixing due to jet deflection. The micromixing time in the JSTR ranged from 10 to 30 ms, representing a reduction of up to 3.45 times compared with the situation that only stirred tank was used. Numerical simulations of flow in the JSTR revealed four typical flow patterns illustrating the way by which the jet can affect the micromixing within the reactor. Furthermore, an operating diagram for mapping the micromixing time based on energy dissipation rate analysis was developed. (c) 2025 The Chemical Industry and Engineering Society of China, and Chemical Industry Press Co., Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

Federated graph learning (FGL) enables multiple participants with distributed but connected graph data to collaboratively train a model in a privacy-preserving way. However, the high communication cost hinders the adoption of FGL in many resource-limited or delay-sensitive applications. In this work, we focus on reducing the communication cost incurred by the transmission of neighborhood information in FGL. We propose to search for local proxies that can play a substitute role as the external neighbors and develop a novel federated graph learning framework named Gproxy. Gproxy utilizes representation similarity and class correlation to select local proxies for external neighbors. Additionally, we propose to dynamically adjust the proxy strategy according to the changing representation of nodes during the iterative training process. We also design a proxy cache to accelerate the search process by reusing proxy search outcomes for similar external neighbors. Furthermore, we provide a theoretical analysis and show that using a proxy node has a similar influence on training when it is sufficiently similar to the external one. Extensive evaluations show that Gproxy significantly reduces communication cost while maintaining model performance compared to strong baselines.

Fabrication of ionophore-based polymeric membrane anion-selective electrodes is still a big challenge due to the lack of suitable anion ionophores. Screening of suitable ionophores by computer-aided techniques may be a good option to change this situation. Herein, density functional theory (DFT) calculations which take into account solvation effects are proposed to screen anion ionophores. The binding energy and binding free energy between the ionophores and anions are used as criteria for the selection of ionophores. The optimal ionophore is then synthesized and employed for the fabrication of a potentiometric sensor. As a proof-of-concept experiment, monohydrogen phosphate (HPO42-) is chosen as a model anion; five compounds, including pyrrole derivatives as well as amide-and guanidinium-based compounds, are selected as potential ionophores. After computational screening, the pyrrole derivative (i.e., 5,5 ',5 '',5 ''''-(1,3-phenylenebis(ethane-1,1,1-triyl)) tetrakis(1H-pyrrole-2-carbaldehyde)) shows the highest affinity toward HPO42-.The fabricated potentiometric sensor exhibits good selectivity and excellent sensitivity with a linear response to HPO42-over a concentration range of 10-5 to 10-2 mol/L with a Nernstian slope of-30.8 mV/dec and a detection limit of 1.0 x 10-6 mol/L. We believe that the computer-aided technique for ionophore selection could provide a general and facile way to fabricate ionophorebased electrochemical and optical sensors.

High-viscosity fluid mixing is a critical unit operation in chemical engineering and polymer processing. Its efficiency and uniformity directly determine product quality and production costs. This paper reviews the mixing mechanisms of high-viscosity fluids, which rely on laminar stretching and folding rather than turbulent eddies, and summarizes the corresponding CFD models and constitutive equations of fluid. It also introduces different methods to reduce numerical diffusion and summarizes high-viscosity mixers. Current challenges, such as numerical diffusion in laminar simulations and the high Weissenberg number problem, are discussed. The application of machine learning in CFD simulations is considered a prospective method to enhance the accuracy and efficiency of simulations involving high-viscosity fluid mixing.

Background: Sepsis-associated encephalopathy (SAE) is a severe complication of sepsis, affecting approximately 70% of patients, leading to increased mortality and long-term cognitive impairments among survivors. However, there is a lack of comprehensive studies on the development of SAE, especially related to the cellular communication networks in the brain microenvironment. Methods: We evaluated the impact of myeloid cells on the brain's immune microenvironment through glial cell alterations using bulk and single-cell transcriptomics data from human and mouse models and validated this with correlative experiments. We also developed the DeconvCellLink R package to study neuroinflammation-associated cellular interaction networks. A dynamic brain immune microenvironment map showing temporal alterations in brain cellular network during systemic inflammatory reactions was constructed using time-series data. Results: While brain cellular alterations differed between human and animal models, a highly conserved set of sepsis-associated genes regulating immune microenvironment signalling was identified. The dynamic alterations in cellular interaction networks and cytokines revealed brain immune cells' temporal response to systemic inflammation. We also found that valproic acid could mitigate sepsis-induced neuroinflammation by regulating glial cell balance and modulating the neuroimmune microenvironment. Conclusion: Through dynamic cellular communication networks, the study revealed that, immune dysregulation in the inflamed brain in SAE involves overactivation of innate immunity, with neutrophils playing a crucial role, providing a scientific framework for developing novel therapeutic strategies and offering new insights into the mechanisms underlying sepsis-induced brain dysfunction.

Stirred tanks are widely used in polymerization processes, where the residence time distribution (RTD) significantly affects monomer conversion and polymer quality. In this study, the RTD in the stirred tank with both constant and variable viscosity fluids was investigated numerically. To account for the viscosity evolution during polymerization, a model relating fluid viscosity to the mean age of the fluid was developed. After verifying mesh and time step independence, the effects of impeller speed, fluid space time, and viscosity varying on RTD were examined in both single-tank and two-tank configurations. Compared to the constant-viscosity fluids, the variable-viscosity fluid shows different flow behaviors such as dead zones and short-circuiting. Analysis based on the number of tanks in series showed that increasing impeller speed and extending space time can enhance mixing efficiency, where the improved mixing in the second stage of the two-tank configuration eliminated the concentration fluctuations caused by recirculating flow in the first tank, which may result in a more uniform RTD curves.

Portfolio management is the art and science in fiance that concerns continuous reallocation of funds and assets across financial instruments to meet the desired returns to risk profile. Deep reinforcement learning (RL) has gained increasing interest in portfolio management, where RL agents are trained base on financial data to optimize the asset reallocation process. Though there are prior efforts in trying to combine RL and portfolio management, previous works did not consider practical aspects such as transaction costs or short selling restrictions, limiting their applicability. To address these limitations, we propose a general RL framework for asset management that enables continuous asset weights, short selling and making decisions with relevant features. We compare the performance of three different RL algorithms: Policy Gradient with Actor-Critic (PGAC), Proximal Policy Optimization (PPO), and Evolution Strategies (ES) and demonstrate their advantages in a simulated environment with transaction costs. Our work aims to provide more options for utilizing RL frameworks in real-life asset management scenarios and can benefit further research in financial applications.

Conversion rate (CVR) prediction is essential in recommender systems, facilitating precise matching between recommended items and users' preferences. However, the sample selection bias (SSB) and data sparsity (DS) issues pose challenges to accurate prediction. Existing works have proposed the click-through and conversion rate (CTCVR) prediction task which models samples from exposure to "click and conversion" in entire space and incorporates multi-task learning. This approach has shown efficacy in mitigating these challenges. Nevertheless, it intensifies the false negative sample (FNS) problem. To be more specific, the CTCVR task implicitly treats all the CVR labels of non-click samples as negative, overlooking the possibility that some samples might convert if clicked. This oversight can negatively impact CVR model performance, as empirical analysis has confirmed. To this end, we advocate for discarding the CTCVR task and proposing a Non-click samples Improved Semi-supErvised (NISE) method for conversion rate prediction, where the non-click samples are treated as unlabeled. Our approach aims to predict their probabilities of conversion if clicked, utilizing these predictions as pseudo-labels for further model training. This strategy can help alleviate the FNS problem, and direct modeling of the CVR task across the entire space also mitigates the SSB and DS challenges. Additionally, we conduct multi-task learning by introducing an auxiliary click-through rate prediction task, thereby enhancing embedding layer representations. Our approach is applicable to various multi-task architectures. Comprehensive experiments are conducted on both public and production datasets, demonstrating the superiority of our proposed method in mitigating the FNS challenge and improving the CVR estimation. The implementation code is available at https://github.com/Hjh233/NISE.

Federated Learning (FL) has emerged as a privacy-preserving distributed machine learning paradigm. To motivate data owners to contribute towards FL, research on FL incentive mechanisms is gaining great interest. Existing monetary incentive mechanisms generally share the same FL model with all participants regardless of their contributions. Such an assumption can be unfair towards participants who contributed more and promote undesirable free-riding, especially when the final model is of great utility value to participants. In this paper, we propose a Fairness-Aware Incentive Mechanism for federated learning (FedFAIM) to address such problem. It satisfies two types of fairness notion: 1) aggregation fairness, which determines aggregation results according to data quality; 2) reward fairness, which assigns each participant a unique model with performance reflecting his contribution. Aggregation fairness is achieved through efficient gradient aggregation which examines local gradient quality and aggregates them based on data quality. Reward fairness is achieved through an efficient Shapley value-based contribution assessment method and a novel reward allocation method based on reputation and distribution of local and global gradients. We further prove reward fairness is theoretically guaranteed. Extensive experiments show that FedFAIM provides stronger incentives than similar non-monetary FL incentive mechanisms while achieving a high level of fairness.

The development of sensors for detection of biomarkers exhibits an exciting potential in diagnosis of diseases. Herein, we propose a novel electrochemical sensing strategy for label-free dual-biomarker detection, which is based on the combination of stimulus-responsive molecularly imprinted polymer (MIP)-modified nanopores and a polymeric membrane chronopotentiometric sensor. The ion fluxes galvanostatically imposed on the sensing membrane surface can be blocked by the recognition reaction between the target biomarker in the sample solution and the stimulus-responsive MIP receptor in the nanopores, thus causing a potential change. By using two external stimuli (i.e., pH and temperature), the recognition abilities of the stimulus-responsive MIP receptor can be effectively modulated so that dual-biomarker label-free chronopotentiometric detection can be achieved. Using alpha fetoprotein (AFP) and prostate-specific antigen (PSA) as model biomarkers, the proposed sensor offers detection limits of 0.17 and 0.42 ng/mL for AFP and PSA, respectively.

Trading is a highly competitive task that requires a combination of strategy, knowledge, and psychological fortitude. With the recent success of large language models(LLMs), it is appealing to apply the emerging intelligence of LLM agents in this competitive arena and understanding if they can outperform professional traders. In this survey, we provide a comprehensive review of the current research on using LLMs as agents in financial trading. We summarize the common architecture used in the agent, the data inputs, and the performance of LLM trading agents in backtesting as well as the challenges presented in these research. This survey aims to provide insights into the current state of LLM-based financial trading agents and outline future research directions in this field.

Potentiometric sensors based on the molecularly imprinted polymers (MIPs) as the receptors have been successfully developed for determination of various organic and biological species. However, these MIP receptors may suffer from problems of low selectivity. Especially, it would be difficult to distinguish the target analyte from its structurally similar interferents. In this work, we propose a novel strategy that using specific functional monomer to fabricate MIP with high selectivity towards the target molecule. The density functional theory calculations are used to investigate the interactions between the template and the functional monomer. The binding energy between the template and functional monomer can be used as the criterion for identifying the optimal monomer. As a proof-of-concept experiment, bisphenol A (BPA) is chosen as the template and the MIP is synthesized by the precipitation polymerization method using the specific allyl-β-cyclodextrin (allyl-β-CD) with high affinity towards BPA as the functional monomer. The high-affinity MIP is employed as the receptor for the construction of the potentiometric sensor. The proposed potentiometric sensor based on the MIP using allyl-β-CD as the functional monomer shows an improved response performance in terms of selectivity and sensitivity compared to the conventional potentiometric sensor based on the MIP with the common monomer (i.e., methacrylic acid). This allyl-β-CD MIP-based potentiometric sensor shows a detection limit of 0.29 μM for BPA, which is about one order of magnitude lower than that obtained by the conventional MIP-based potentiometric sensor. We believe that utilizing a functional monomer with specific recognition ability towards target in the fabrication of MIP could provide an appealing way to construct highly selective MIP-based electrochemical and optical sensors.

Computational fluid dynamics (CFD) was used to assess the flow patterns and mixing characteristics of highly viscous fluids in a dynamic mixer. The simulated concentration distribution was validated by the data measured using planar laser-induced fluorescence (PLIF). The refractive indices of the two fluids and the transparent solid material comprising the mixer were matched to minimize the bending of sent and received light. The simulated concentration fields agreed well with the measured concentration fields. The flow pattern and mixing performance were quantitatively evaluated by using the mixing index lambda and the coefficient of variation (COV), respectively. After the mixing performance reached time independence, a reduction of the COV at the outlet was associated with a slight increase of the volume-averaged |lambda|. The results suggested that elongational flow was more effective than shear flow in obtaining a more uniform concentration distribution in highly viscous fluids.

The concentration and velocity fields of two refractive index matched miscible shear-thinning fluids in a lid-driven cavity were investigated by using planar laser-induced fluorescence and particle image velocimetry,as well by computational fluid dynamics. Quantitative analyses show that the results obtained by flow simulations with the species transport model are in good agreement with the experimental results. The effects of different parameters were studied by using the intensity of segregation. For two fluids with the same rheological parameters,the relative amounts of liquids H1/H and the power-law index n dominate the mixing process while the Reynolds number Re plays a marginal role. As for two fluids with density dif-ference,buoyancy has significant influence on the mixing process. The dimensionless group Ar/Re (rede-fined such as to include shear thinning behavior) is proposed for assessing the effect of buoyancy and rheological properties on the mixing of miscible shear-thinning fluids.

In this paper, we investigate the streaming bandits problem, wherein the learner aims to minimize regret by dealing with online arriving arms and sublinear arm memory. We establish the tight worst-case regret lower bound of $\Omega \left( (TB)^{\alpha} K^{1-\alpha}\right), \alpha = 2^{B} / (2^{B+1}-1)$ for any algorithm with a time horizon $T$, number of arms $K$, and number of passes $B$. The result reveals a separation between the stochastic bandits problem in the classical centralized setting and the streaming setting with bounded arm memory. Notably, in comparison to the well-known $\Omega(\sqrt{KT})$ lower bound, an additional double logarithmic factor is unavoidable for any streaming bandits algorithm with sublinear memory permitted. Furthermore, we establish the first instance-dependent lower bound of $\Omega \left(T^{1/(B+1)} \sum_{\Delta_x>0} \frac{\mu^*}{\Delta_x}\right)$ for streaming bandits. These lower bounds are derived through a unique reduction from the regret-minimization setting to the sample complexity analysis for a sequence of $\epsilon$-optimal arms identification tasks, which maybe of independent interest. To complement the lower bound, we also provide a multi-pass algorithm that achieves a regret upper bound of $\tilde{O} \left( (TB)^{\alpha} K^{1 - \alpha}\right)$ using constant arm memory.

Since the 1960s, neonatal clinicians have known that newborns suffering from certain neurological conditions exhibit altered crying patterns such as the high-pitched cry in birth asphyxia. Despite an annual burden of over 1.5 million infant deaths and disabilities, early detection of neonatal brain injuries due to asphyxia remains a challenge, particularly in developing countries where the majority of births are not attended by a trained physician. Here we report on the first inter-continental clinical study to demonstrate that neonatal brain injury can be reliably determined from recorded infant cries using an AI algorithm we call Roseline. Previous and recent work has been limited by the lack of a large, high-quality clinical database of cry recordings, constraining the application of state-of-the-art machine learning. We develop a new training methodology for audio-based pathology detection models and evaluate this system on a large database of newborn cry sounds acquired from geographically diverse settings -- 5 hospitals across 3 continents. Our system extracts interpretable acoustic biomarkers that support clinical decisions and is able to accurately detect neurological injury from newborns' cries with an AUC of 92.5% (88.7% sensitivity at 80% specificity). Cry-based neurological monitoring opens the door for low-cost, easy-to-use, non-invasive and contact-free screening of at-risk babies, especially when integrated into simple devices like smartphones or neonatal ICU monitors. This would provide a reliable tool where there are no alternatives, but also curtail the need to regularly exert newborns to physically-exhausting or radiation-exposing assessments such as brain CT scans. This work sets the stage for embracing the infant cry as a vital sign and indicates the potential of AI-driven sound monitoring for the future of affordable healthcare.

In this paper, we explore self-supervised learning (SSL) for analyzing a first-of-its-kind database of cry recordings containing clinical indications of more than a thousand newborns. Specifically, we target cry-based detection of neurological injury as well as identification of cry triggers such as pain, hunger, and discomfort. Annotating a large database in the medical setting is expensive and timeconsuming, typically requiring the collaboration of several experts over years. Leveraging large amounts of unlabeled audio data to learn useful representations can lower the cost of building robust models and, ultimately, clinical solutions. In this work, we experiment with self-supervised pre-training of a convolutional neural network on large audio datasets. We show that pre-training with SSL contrastive loss (SimCLR) performs significantly better than supervised pre-training for both neuro injury and cry triggers. In addition, we demonstrate further performance gains through SSL-based domain adaptation using unlabeled infant cries. We also show that using such SSL-based pre-training for adaptation to cry sounds decreases the need for labeled data of the overall system.

Vertical federated learning (VFL) enables multiple participants with different data features and the same sample ID space to collaboratively train a model in a privacy-preserving way. However, the high computational and communication overheads hinder the adoption of VFL in many resource-limited or delay-sensitive applications. In this work, we focus on reducing the communication cost and delay incurred by the transmission of intermediate results in VFL model serving. We investigate the inference results, and find that a large portion of test samples can be predicted correctly by the active party alone, thus the corresponding communication for federated inference is dispensable. Based on this insight, we theoretically analyze the "dispensable communication" and propose a novel tunable vertical federated learning framework, named TVFL, to avoid "dispensable communication" in model serving as much as possible. TVFL can smartly switch between independent inference and federated inference based on the features of the input sample. We further reveal that such tunability is highly related to the importance of participants’ features. Our evaluations on seven datasets and three typical VFL models show that TVFL can save 57.6% communication cost and reduce 57.1% prediction latency with little performance degradation.

Continuously operated multistage stirred reactors are widely applied in the chemical industry. Understanding the flow and residence time distribution is vital in the design and optimization of such reactor. In the present work, the fluid flow field and residence time distribution in the multistage stirred reactor equipped with Rushton turbines were investigated numerically. After grid independence and the time sensitivity test and validation with the literature, the effect of several dimensionless terms, including the flow and impeller Reynolds numbers and normalized opening size on the horizontal plate, on the fluid average residence time and residence time distribution was quantitatively evaluated. Thus, the number of equivalent reactors can be varied accordingly by adjusting the flow Reynolds number, the impeller Reynolds number, and the opening between adjacent chambers. Finally, a correlation of the number of equivalent reactors is proposed, which can be used to predict the number of equivalent reactors under various operating conditions.

搅拌反应器气液传质实验是《化学反应工程》的基础实验,主要通过对搅拌反应器内的气液传质特性进行测定,并通过关联式回归得到诸因素对传质系数的定量影响规律.但在传统的实验教学中,学时及实验测试技术的限制使得无法直接获取反应器内气液两相流的流场结构等影响传质特性相关信息.计算流体力学(CFD)方法可以从定性和定量角度揭示流场结构、气体分布以及气液传质等特性,并具有可视化、计算效率高等优点.将 CFD 方法与气液搅拌反应器传质实验教学融合,可加深学生对搅拌反应器内气液分散及传质机理的理解,培养学生解决化学工程问题的高阶能力.