Search

The rapid advancement of Large Language Models (LLMs) has led to extensive discourse regarding their potential to boost the return of quantitative stock trading strategies. This discourse primarily revolves around harnessing the remarkable comprehension capabilities of LLMs to extract sentiment factors which facilitate informed and high-frequency investment portfolio adjustments. To ensure successful implementations of these LLMs into the analysis of Chinese financial texts and the subsequent trading strategy development within the Chinese stock market, we provide a rigorous and encompassing benchmark as well as a standardized back-testing framework aiming at objectively assessing the efficacy of various types of LLMs in the specialized domain of sentiment factor extraction from Chinese news text data. To illustrate how our benchmark works, we reference three distinctive models: 1) the generative LLM (ChatGPT), 2) the Chinese language-specific pre-trained LLM (Erlangshen-RoBERTa), and 3) the financial domain-specific fine-tuned LLM classifier(Chinese FinBERT). We apply them directly to the task of sentiment factor extraction from large volumes of Chinese news summary texts. We then proceed to building quantitative trading strategies and running back-tests under realistic trading scenarios based on the derived sentiment factors and evaluate their performances with our benchmark. By constructing such a comparative analysis, we invoke the question of what constitutes the most important element for improving a LLM's performance on extracting sentiment factors. And by ensuring that the LLMs are evaluated on the same benchmark, following the same standardized experimental procedures that are designed with sufficient expertise in quantitative trading, we make the first stride toward answering such a question.

The development of smart homes has advanced the concept of user authentication to not only protecting user privacy but also facilitating personalized services to users. Along this direction, we propose to integrate user authentication with human-computer interactions between users and smart household appliances through widely-deployed WiFi infrastructures, which is non-intrusive and device-free. In this paper, we propose $FingerPass$ F i n g e r P a s s which leverages channel state information (CSI) of surrounding WiFi signals to continuously authenticate users through finger gestures in smart homes. $FingerPass$ F i n g e r P a s s separates the user authentication process into two stages, login and interaction, to achieve high authentication accuracy and low response latency simultaneously. In the login stage, we develop a deep learning-based approach to extract behavioral characteristics of finger gestures for highly accurate user identification. For the interaction stage, to provide continuous authentication in real time for satisfactory user experience, we design a verification mechanism with lightweight classifiers to continuously authenticate the user’s identity during each interaction of finger gestures. Experiments in real environments show that $FingerPass$ F i n g e r P a s s can achieve the authentication accuracies of 90.6 percent under in-domain scenarios and 87.6 percent under cross-domain scenarios, as well as $186.6\;ms$ 186 . 6 m s response time during interactions.

Clathrin-mediated endocytosis (CME) is a normal biological process where cellular contents are transported into the cells. However, this process is often hijacked by different viruses to enter host cells and cause infections. Recently, two proteins that regulate CME – AAK1 and GAK – have been proposed as potential therapeutic targets for designing broad-spectrum antiviral drugs. In this work, we curated two compound datasets containing 83 AAK1 inhibitors and 196 GAK inhibitors each. Subsequently, machine learning methods, namely Random Forest, Elastic Net and Sequential Minimal Optimization, were used to construct Quantitative Structure Activity Relationship (QSAR) models to predict small molecule inhibitors of AAK1 and GAK. To ensure predictivity, these models were evaluated by using Leave-One-Out (LOO) cross validation and with an external test set. In all cases, our QSAR models achieved a q2LOO in range of 0.64 to 0.84 (Root Mean Squared Error; RMSE = 0.41 to 0.52) and a q2ext in range of 0.57 to 0.92 (RMSE = 0.36 to 0.61). Besides, our QSAR models were evaluated by using additional QSAR performance metrics and y-randomization test. Finally, by using a concensus scoring approach, nine chemical compounds from the Drugbank compound library were predicted as AAK1/GAK dual-target inhibitors. The electrostatic potential maps for the nine compounds were generated and compared against two known dual-target inhibitors, sunitinib and baricitinib. Our work provides the rationale to validate these nine compounds experimentally against the protein targets AAK1 and GAK.

Scaling down the resolution of input images can greatly reduce the computational overhead of convolutional neural networks (CNNs), which is promising for edge AI. However, as an image usually contains much spatial redundancy, e.g., background pixels, directly shrinking the whole image will lose important features of the foreground object and lead to severe accuracy degradation. In this paper, we propose a dynamic image cropping framework to reduce the spatial redundancy by accurately cropping the foreground object from images. To achieve the instance-aware fine cropping, we introduce a lightweight foreground predictor to efficiently localize and crop the foreground of an image. The finely cropped images can be correctly recognized even at a small resolution. Meanwhile, computational redundancy also exists in CNN architectures. To pursue higher execution efficiency on resource-constrained embedded devices, we also propose a compound shrinking strategy to coordinately compress the three dimensions (depth, width, resolution) of CNNs. Eventually, we seamlessly combine the proposed dynamic image cropping and compound shrinking into a unified compression framework, Smart Scissor, which is expected to significantly reduce the computational overhead of CNNs while still maintaining high accuracy. Experiments on ImageNet-1K demonstrate that our method reduces the computational cost of ResNet50 by 41.5% while improving the top-1 accuracy by 0.3%. Moreover, compared to HRank, the state-of-theart CNN compression framework, our method achieves 4.1% higher top-1 accuracy at the same computational cost. The codes and data are available at https://github.com/ntuliuteam/smart-scissor. Nanyang Technological University National Research Foundation (NRF) Submitted/Accepted version This study is partially supported under the RIE2020 Industry Alignment Fund – Industry Collaboration Projects (IAF-ICP) Funding Initiative, as well as cash and in-kind contribution from the industry partner, HP Inc., through the HP-NTU Digital Manufacturing Corporate Lab (I1801E0028). This work is also partially supported by Nanyang Technological University, Singapore, under its NAP (M4082282).

Federated Learning (FL) empowers multiple clients to collaboratively learn a model, enlarging the training data of each client for high accuracy while protecting data privacy. However, when deploying FL in real-time edge systems, the heterogeneity of devices among systems has a severe impact on the performance of the inferred model. Existing optimizations on FL focus on improving the training efficiency but fail to speed up inference, especially when there is a latency constraint. In this work, we propose Collate, a novel training framework that collaboratively learns heterogeneous models to meet the latency constraints of multiple edge systems simultaneously. We design a dynamic zeroizing-recovering method to adjust each local model architecture for high accuracy under its latency constraint. A proto-corrected federated aggregation scheme is also introduced to aggregate all heterogeneous local models, satisfying the latency constraint of different systems with only one training process and maintaining high accuracy. Extensive experiments indicate that, compared to state-of-the-art methods and under a latency constraint, our extended models can improve the accuracy by 1.96% on average, and our shrunk models can also obtain a 3.09% accuracy improvement on average, with almost no extra training overhead. The related codes and data will be available at https://github.com/ntuliuteam/Collate. This study is supported under the RIE2020 Industry Alignment Fund – Industry Collaboration Projects (IAF-ICP) Funding Initiative, as well as cash and in-kind contribution from the industry partner, HP Inc., through the HP-NTU Digital Manufacturing Corporate Lab (I1801E0028).

중국 전자상거래 시장은 지난 1997년부터 2002년까지는 중국의 낮은 정보화 기술 수준과 온라인 비즈니스에 대한 신뢰가 부족해 어려움이 있었다. 그러나 2003년 이후 중국 정부의 적극적인 정책 추진을 통해 빠른 속도로 급성장하고 있다. 특히, 일대일로 정책이 추진되고 물류산업이 발전하면서 전자상거래 시장도 지속적으로 성장하고 있다. 또한 중국 경제의 지속적인 성장과 중산층의 빠른 증가는 중국 전자상거래 발전을 촉진시켰다. 본 논문은 중국 국경무역 전자상거래의 발전 현황과 영향 요인을 도출하고 활성화 방안을 제시하기 전자상거래 현황과 발전에 관한 최신 데이터 및 관련 사례를 분석하였다. 또한 중국 전자 상거래 모델과 산업 사슬, 공급 사슬에 대한 분석을 통해 국경무역 전자상거래 발전에 미치는 영향 요인을 제시하였다. 이를 통해 국경무역 전자상거래 발전을 위해 시급히 해결해야 할 현안 등의 탐색적 분석을 통해 중국 국경무역 전자상거래 활성화를 위한 대책과 방안을 제시하였다. 이를 통해 국경무역 전자상거래에 관련된 이해당사자들과 연구자들에게 의미 있는 기초자료를 제공할 것으로 보인다.

Model scaling is an effective way to improve the accuracy of deep neural networks (DNNs) by increasing the model capacity. However, existing approaches seldom consider the underlying hardware, causing inefficient utilization of hardware resources and consequently high inference latency. In this paper, we propose HACScale, a hardware-aware model scaling strategy to fully exploit hardware resources for higher accuracy. In HACScale, different dimensions of DNNs are jointly scaled with consideration of their contributions to hardware utilization and accuracy. To improve the efficiency of width scaling, we introduce importance-aware width scaling in HACScale, which computes the importance of each layer to the accuracy and scales each layer accordingly to optimize the trade-off between accuracy and model parameters. Experiments show that HACScale improves the hardware utilization by 1.92× on ImageNet, as a result, it achieves 2.41% accuracy improvement with a negligible latency increase of 0.6%. On CIFAR-10, HACScale improves the accuracy by 2.23% with only 6.5% latency growth. Nanyang Technological University National Research Foundation (NRF) Submitted/Accepted version This study is supported under the RIE2020 Industry Alignment Fund – Industry Collaboration Projects (IAF-ICP) Funding Initiative, as well as cash and in-kind contribution from the industry partner, HP Inc., through the HP-NTU Digital Manufacturing Corporate Lab. This work is also partially supported by NTU NAP M4082282 and SUG M4082087, Singapore.

Recently, many works have demonstrated that Symmetric Non-negative Matrix Factorization~(SymNMF) enjoys a great superiority for various clustering tasks. Although the state-of-the-art algorithms for SymNMF perform well on synthetic data, they cannot consistently obtain satisfactory results with desirable properties and may fail on real-world tasks like clustering. Considering the flexibility and strong representation ability of the neural network, in this paper, we propose a neural network called SymNMF-Net for the Symmetric NMF problem to overcome the shortcomings of traditional optimization algorithms. Each block of SymNMF-Net is a differentiable architecture with an inversion layer, a linear layer and ReLU, which are inspired by a traditional update scheme for SymNMF. We show that the inference of each block corresponds to a single iteration of the optimization. Furthermore, we analyze the constraints of the inversion layer to ensure the output stability of the network to a certain extent. Empirical results on real-world datasets demonstrate the superiority of our SymNMF-Net and confirm the sufficiency of our theoretical analysis.

Early prognosis of abnormal vasculopathy is essential for effective clinical management of severe dengue patients. An exaggerated interferon (IFN) response and release of vasoactive factors from endothelial cells cause vasculopathy. This study shows that dengue 2 (DENV2) infection of human umbilical vein endothelial cells (HUVEC) results in differentially regulated miRNAs important for endothelial function. miR-573 was significantly down-regulated in DENV2-infected HUVEC due to decreased Peroxisome Proliferator Activator Receptor Gamma (PPARγ) activity. Restoring miR-573 expression decreased endothelial permeability by suppressing the expression of vasoactive angiopoietin 2 (ANGPT2). We also found that miR-573 suppressed the proinflammatory IFN response through direct downregulation of toll like receptor 2 (TLR2) expression. Our study provides a novel insight into miR-573 mediated regulation of endothelial function during DENV2 infection which can be further translated into a potential therapeutic and prognostic agent for severe dengue patients.

With the increasing integration of humans and the cyber world, user authentication becomes critical to support various emerging application scenarios requiring security guarantees. Existing works utilize Channel State Information (CSI) of WiFi signals to capture single human activities for non-intrusive and device-free user authentication, but multi-user authentication remains a challenging task. In this paper, we present a multi-user authentication system, MultiAuth, which can authenticate multiple users with a single commodity WiFi device. The key idea is to profile multipath components of WiFi signals induced by multiple users, and construct individual CSI from the multipath components to solely characterize each user for user authentication. Specifically, we propose a MUltipath Time-of-Arrival measurement algorithm (MUTA) to profile multipath components of WiFi signals in high resolution. Then, after aggregating and separating the multipath components related to users, MultiAuth constructs individual CSI based on the multipath components to solely characterize each user. To identify users, MultiAuth further extracts user behavior profiles based on the individual CSI of each user through time-frequency analysis, and leverages a dual-task neural network for robust user authentication. Extensive experiments involving 3 simultaneously present users demonstrate that MultiAuth is accurate and reliable for multi-user authentication with 87.6% average accuracy and 8.8% average false accept rate.

Additional file 1.

While successful in many fields, deep neural networks (DNNs) still suffer from some open problems such as bad local minima and unsatisfactory generalization performance. In this work, we propose a novel architecture called Maximum-and-Concatenation Networks (MCN) to try eliminating bad local minima and improving generalization ability as well. Remarkably, we prove that MCN has a very nice property; that is, \emph{every local minimum of an $(l+1)$-layer MCN can be better than, at least as good as, the global minima of the network consisting of its first $l$ layers}. In other words, by increasing the network depth, MCN can autonomously improve its local minima's goodness, what is more, \emph{it is easy to plug MCN into an existing deep model to make it also have this property}. Finally, under mild conditions, we show that MCN can approximate certain continuous functions arbitrarily well with \emph{high efficiency}; that is, the covering number of MCN is much smaller than most existing DNNs such as deep ReLU. Based on this, we further provide a tight generalization bound to guarantee the inference ability of MCN when dealing with testing samples., Comment: Accepted by ICML2020

Additional file 1 Supplementary Table 1. Univariate association between baseline/intraoperative variables and postoperative complications.

Additional file 1: Supplemental Table 1. Complications during hospitalization in propensity score matched patients.

The development of smart homes has advanced the concept of user authentication to not only protecting user privacy but also facilitating personalized services to users. Along this direction, we propose to integrate user authentication with human-computer interactions between users and smart household appliances through widely-deployed WiFi infrastructures, which is non-intrusive and device-free. In this paper, we propose FingerPass which leverages channel state information (CSI) of surrounding WiFi signals to continuously authenticate users through finger gestures in smart homes. We investigate CSI of WiFi signals in depth and find CSI phase can be used to capture and distinguish the unique behavioral characteristics from different users. FingerPass separates the user authentication process into two stages, login and interaction, to achieve high authentication accuracy and low response latency simultaneously. In the login stage, we develop a deep learning-based approach to extract behavioral characteristics of finger gestures for highly accurate user identification. For the interaction stage, to provide continuous authentication in real time for satisfactory user experience, we design a verification mechanism with lightweight classifiers to continuously authenticate the user's identity during each interaction of finger gestures. Experiments in real environments show that FingerPass can achieve 91.4% authentication accuracy, and 186.6ms response time during interactions.

On August 31, 2016, the People’s Bank of China (PBC), along with six other government agencies, issued the Guidelines for Establishing the Green Financial System, introducing a series of measures to support and encourage green investment and financing. On September 2, the China Banking Regulatory Commission (CBRC) proposed a framework for building a green credit system and comprehensive evaluation of green credit, which lays out the development direction for green lending. Learning from international and domestic research and practical experiences, this paper tries to develop a green credit system for marine environmental protection. Based on the current green credit theory and policy trend, this thesis sets the operating mechanism for green credit, which can enforce contact and cooperation between marine environmental protection departments and financial regulators. All of these can promote enterprises and individual polluters to pay attention to marine environment, strengthen pollution control and maintain ecosystem healthy.

Strigolactones (SLs) are a novel class of plant hormones with enormous potential for the prevention and treatment of inflammation. To further investigate the anti-inflammatory activities of SLs, a representative SL, GR24, and the reductive products of its D-ring were synthesized and their anti-inflammatory activities were fully evaluated on both in vitro and in vivo models. Among these compounds, the two most active optical isomers (2a and 6a) demonstrated strong inhibitory activity on the release of inflammatory cytokines, including nitric oxide (NO), tumor necrosis factor-alpha (TNF-α), and interleukin-6 (IL-6) by blocking the nuclear factor kappa B (NF-κB) and mitogen-activated protein kinase (MAPK) signaling pathways; they also greatly inhibited the migration of neutrophils and macrophages in fluorescent protein labeled zebrafish larvae. These results identified the promising anti-inflammatory effects of SLs, and suggested that both the absolute configuration of SL and the α,β-unsaturated D-ring structure are essential for the observed anti-inflammatory activity.

A method for the synthesis of 2,2′-bipyridin-4-one from 4-amino-4-(2-pyridyl)-but-3-en-2-onevia its diphenylboron, chelates was proposed.

Several small-molecule CDK inhibitors have been identified, but none have been approved for clinical use in the past few years. A new series of 4-[(3-hydroxybenzylamino)-methylene]-4H-isoquinoline-1,3-diones were reported as highly potent and selective CDK4 inhibitors. In order to find more potent CDK4 inhibitors, the interactions between these novel isoquinoline-1,3-diones and cyclin-dependent kinase 4 was explored via in silico methodologies such as 3D-QSAR and docking on eighty-one compounds displaying potent selective activities against cyclin-dependent kinase 4. Internal and external cross-validation techniques were investigated as well as region focusing, bootstraping and leave-group-out. A training set of 66 compounds gave the satisfactory CoMFA model (q 2 = 0.695, r 2 = 0.947) and CoMSIA model (q 2 = 0.641, r 2 = 0.933). The remaining 15 compounds as a test set also gave good external predictive abilities with r 2 pred values of 0.875 and 0.769 for CoMFA and CoMSIA, respectively. The 3D-QSAR models generated here predicted that all five parameters are important for activity toward CDK4. Surflex-dock results, coincident with CoMFA/CoMSIA contour maps, gave the path for binding mode exploration between the inhibitors and CDK4 protein. Based on the QSAR and docking models, twenty new potent molecules have been designed and predicted better than the most active compound 12 in the literatures. The QSAR, docking and interactions analysis expand the structure-activity relationships of constrained isoquinoline-1,3-diones and contribute towards the development of more active CDK4 subtype-selective inhibitors.

Using the quasi-classical trajectory method, the product rotational polarization of the ion-molecule reaction He+D2+ has been calculated at different collision energies on the PALMIERI potential energy surface [Palmieri et al. Mol. Phys. 98 (2000) 1835]. The distribution angle between κ and j', P(θr), the distribution of the dihedral angle P(r), and the angular distribution of product rotational vectors in the form of polar plots in θr and r are calculated. In addition, four polarization-dependent differential cross sections are also presented in the center-of-mass frame, respectively. The results indicate that the rotational polarization of the product HeD+ presents different characters for different collision energies. These discrepancies may be ascribed to the different collision energies and constructions of the potential energy surface.

The stereodynamics in the chemical reaction He+H+2 and its isotopic variants at the collision energy of 145 kJ/mol have been studied by using the quasi-classical trajectory （QCT） method on AQUILANTI surface. The correlated k-j′， k-k′-j′ angular distributions, and the product rotational alignment are discussed in detail. The calculations indicate that the stereodynamics of the reactions of He+HH+→HeH++H, He+HD+→HeH++D and He+HT+→HeH++T are sensitive to the mass factor and the repulsive energy of diatomic molecule.

1. A practical method for the synthesis of (-)-4-methyl-3-heptanol, a component of the aggreational pheromone of theScolytus bark beetles, is described, from the accessible chiral hydrocarbon (+)-3,7-dimethyl-1,6-octadiene. The overall yield of (-)-4-methyl-3-heptanol is 17.6% over the nine stages of the synthesis. 2. The ratio of (3S, 4S)-to (3R,4S)-dia'stereoisomers in the (-)-4-methyl-3-heptanol obtained is 72∶28, and their optical purities are around 48%.

The simple synthesis of optically active R,Z-14-methyl-8-hexadecenal (cis-trogodermal), the principal components of the aggregation pheromone ofTrogoderma granarium, was accomplished. The synthesis was based on the use of R-1-bromo-4-methylhexane and Z-4-chloro-2-buten-1-ol as the sources of the chiral and olefin fragments.

A short synthesis of R,E-14-methyl-8-hexadecenal (trans-trogodermal) from 1,1-ethylene-dioxy-8-acetoxy-9-decene and R-(-)-1-bromo-4-methylhexane was realized.

We have found that, in contrast to previously studied acetylenes, Z-l-hydroxy-3-methyl2-penten-4-yne (I) reacts with CH20 and i-Pr2NH, similarly to the CHaO-Et2NH and analogous systems, with the formation of previously unknown furan derivatives (IIa) and (IIb). Apparently, an acetylenic hydroxyamine is initially formed under Mannich reaction conditions as well as the product of its intramolecular cyclization, namely, a dihydrofuran derivative, which reacts with CH20-R2NH as an enophile to give (II) as the final product.