Showing papers by "Tongji University published in 2019"

Session-Based Recommendation with Graph Neural Networks

Abstract: The problem of session-based recommendation aims to predict user actions based on anonymous sessions. Previous methods model a session as a sequence and estimate user representations besides item representations to make recommendations. Though achieved promising results, they are insufficient to obtain accurate user vectors in sessions and neglect complex transitions of items. To obtain accurate item embedding and take complex transitions of items into account, we propose a novel method, i.e. Session-based Recommendation with Graph Neural Networks, SR-GNN for brevity. In the proposed method, session sequences are modeled as graphstructured data. Based on the session graph, GNN can capture complex transitions of items, which are difficult to be revealed by previous conventional sequential methods. Each session is then represented as the composition of the global preference and the current interest of that session using an attention network. Extensive experiments conducted on two real datasets show that SR-GNN evidently outperforms the state-of-the-art session-based recommendation methods consistently.

Dendritic Neuron Model With Effective Learning Algorithms for Classification, Approximation, and Prediction

Abstract: An artificial neural network (ANN) that mimics the information processing mechanisms and procedures of neurons in human brains has achieved a great success in many fields, e.g., classification, prediction, and control. However, traditional ANNs suffer from many problems, such as the hard understanding problem, the slow and difficult training problems, and the difficulty to scale them up. These problems motivate us to develop a new dendritic neuron model (DNM) by considering the nonlinearity of synapses, not only for a better understanding of a biological neuronal system, but also for providing a more useful method for solving practical problems. To achieve its better performance for solving problems, six learning algorithms including biogeography-based optimization, particle swarm optimization, genetic algorithm, ant colony optimization, evolutionary strategy, and population-based incremental learning are for the first time used to train it. The best combination of its user-defined parameters has been systemically investigated by using the Taguchi’s experimental design method. The experiments on 14 different problems involving classification, approximation, and prediction are conducted by using a multilayer perceptron and the proposed DNM. The results suggest that the proposed learning algorithms are effective and promising for training DNM and thus make DNM more powerful in solving classification, approximation, and prediction problems.

Shape Robust Text Detection With Progressive Scale Expansion Network

Abstract: Scene text detection has witnessed rapid progress especially with the recent development of convolutional neural networks. However, there still exists two challenges which prevent the algorithm into industry applications. On the one hand, most of the state-of-art algorithms require quadrangle bounding box which is in-accurate to locate the texts with arbitrary shape. On the other hand, two text instances which are close to each other may lead to a false detection which covers both instances. Traditionally, the segmentation-based approach can relieve the first problem but usually fail to solve the second challenge. To address these two challenges, in this paper, we propose a novel Progressive Scale Expansion Network (PSENet), which can precisely detect text instances with arbitrary shapes. More specifically, PSENet generates the different scale of kernels for each text instance, and gradually expands the minimal scale kernel to the text instance with the complete shape. Due to the fact that there are large geometrical margins among the minimal scale kernels, our method is effective to split the close text instances, making it easier to use segmentation-based methods to detect arbitrary-shaped text instances. Extensive experiments on CTW1500, Total-Text, ICDAR 2015 and ICDAR 2017 MLT validate the effectiveness of PSENet. Notably, on CTW1500, a dataset full of long curve texts, PSENet achieves a F-measure of 74.3% at 27 FPS, and our best F-measure (82.2%) outperforms state-of-art algorithms by 6.6%. The code will be released in the future.

Progress in high-strain perovskite piezoelectric ceramics

Abstract: High strain piezoelectric ceramics are the state-of-the-art materials for high precision, positioning devices. A comprehensive review of the latest developments of the various types of perovskite piezoelectric ceramic systems is presented herein, with special attention given to three promising families of lead-free perovskite ferroelectrics: the barium titanate, alkaline niobate and bismuth perovskites. Included in this review are details of phase transition behavior, strain enhancement approaches, material reliabilities as well as the status of some promising applications. This current review describes both compositional and structural engineering approaches that are intended to achieve enhanced strain properties in perovskite piezoelectric ceramics. The factors that affect the strain behavior of high-strain perovskite piezoelectric ceramics are addressed. The reliability characteristics of these high-strain ferroelectrics as well as the recent approaches to the long-term electrical, thermal and time-stability enhancement are summarized. Several promising applications of high-strain perovskite materials are introduced, which take advantages of their characteristics; examples include high-energy storage, pyroelectric and electro-caloric effect and luminescent properties.

Cistrome Data Browser: expanded datasets and new tools for gene regulatory analysis.

Abstract: The Cistrome Data Browser (DB) is a resource of human and mouse cis-regulatory information derived from ChIP-seq, DNase-seq and ATAC-seq chromatin profiling assays, which map the genome-wide locations of transcription factor binding sites, histone post-translational modifications and regions of chromatin accessible to endonuclease activity. Currently, the Cistrome DB contains approximately 47,000 human and mouse samples with about 24,000 newly collected datasets compared to the previous release two years ago. Furthermore, the Cistrome DB has a new Toolkit module with several features that allow users to better utilize the large-scale ChIP-seq, DNase-seq, and ATAC-seq data. First, users can query the factors which are likely to regulate a specific gene of interest. Second, the Cistrome DB Toolkit facilitates searches for factor binding, histone modifications, and chromatin accessibility in any given genomic interval shorter than 2Mb. Third, the Toolkit can determine the most similar ChIP-seq, DNase-seq, and ATAC-seq samples in terms of genomic interval overlaps with user-provided genomic interval sets. The Cistrome DB is a user-friendly, up-to-date, and well maintained resource, and the new tools will greatly benefit the biomedical research community. The database is freely available at http://cistrome.org/db, and the Toolkit is at http://dbtoolkit.cistrome.org.

The Chinese Society of Clinical Oncology (CSCO): clinical guidelines for the diagnosis and treatment of gastric cancer.

Abstract: China is one of the countries with the highest incidence of gastric cancer. There are differences in epidemiological characteristics, clinicopathological features, tumor biological characteristics, treatment patterns, and drug selection between gastric cancer patients from the Eastern and Western countries. Non-Chinese guidelines cannot specifically reflect the diagnosis and treatment characteristics for the Chinese gastric cancer patients. The Chinese Society of Clinical Oncology (CSCO) arranged for a panel of senior experts specializing in all sub-specialties of gastric cancer to compile, discuss, and revise the guidelines on the diagnosis and treatment of gastric cancer based on the findings of evidence-based medicine in China and abroad. By referring to the opinions of industry experts, taking into account of regional differences, giving full consideration to the accessibility of diagnosis and treatment resources, these experts have conducted experts’ consensus judgement on relevant evidence and made various grades of recommendations for the clinical diagnosis and treatment of gastric cancer to reflect the value of cancer treatment and meeting health economic indexes. This guideline uses tables and is complemented by explanatory and descriptive notes covering the diagnosis, comprehensive treatment, and follow-up visits for gastric cancer.

Boltzmann generators: Sampling equilibrium states of many-body systems with deep learning

Abstract: Computing equilibrium states in condensed-matter many-body systems, such as solvated proteins, is a long-standing challenge. Lacking methods for generating statistically independent equilibrium samples in "one shot," vast computational effort is invested for simulating these systems in small steps, e.g., using molecular dynamics. Combining deep learning and statistical mechanics, we developed Boltzmann generators, which are shown to generate unbiased one-shot equilibrium samples of representative condensed-matter systems and proteins. Boltzmann generators use neural networks to learn a coordinate transformation of the complex configurational equilibrium distribution to a distribution that can be easily sampled. Accurate computation of free-energy differences and discovery of new configurations are demonstrated, providing a statistical mechanics tool that can avoid rare events during sampling without prior knowledge of reaction coordinates.

Intercalation-conversion hybrid cathodes enabling Li–S full-cell architectures with jointly superior gravimetric and volumetric energy densities

Abstract: A common practise in the research of Li–S batteries is to use high electrode porosity and excessive electrolytes to boost sulfur-specific capacity. Here we propose a class of dense intercalation-conversion hybrid cathodes by combining intercalation-type Mo6S8 with conversion-type sulfur to realize a Li–S full cell. The mechanically hard Mo6S8 with fast Li-ion transport ability, high electronic conductivity, active capacity contribution and high affinity for lithium polysulfides is shown to be an ideal backbone to immobilize the sulfur species and unlock their high gravimetric capacity. Cycling stability and rate capability are reported under realistic conditions of low carbon content (~10 wt%), low electrolyte/active material ratio (~1.2 µl mg−1), low cathode porosity (~55 vol%) and high mass loading (>10 mg cm−2). A pouch cell assembled based on the hybrid cathode and a 2× excess Li metal anode is able to simultaneously deliver a gravimetric energy density of 366 Wh kg−1 and a volumetric energy density of 581 Wh l−1. Despite tremendous progress in the development of LiS batteries, their performance at the full-cell level is not as competitive as state-of-the-art Li-ion batteries. Here the authors report a full-cell architecture making use of a hybrid intercalation-conversion cathode, enabling both high volumetric and gravimetric energy densities.

Progressive Differentiable Architecture Search: Bridging the Depth Gap Between Search and Evaluation

Abstract: Recently, differentiable search methods have made major progress in reducing the computational costs of neural architecture search. However, these approaches often report lower accuracy in evaluating the searched architecture or transferring it to another dataset. This is arguably due to the large gap between the architecture depths in search and evaluation scenarios. In this paper, we present an efficient algorithm which allows the depth of searched architectures to grow gradually during the training procedure. This brings two issues, namely, heavier computational overheads and weaker search stability, which we solve using search space approximation and regularization, respectively. With a significantly reduced search time (~7 hours on a single GPU), our approach achieves state-of-the-art performance on both the proxy dataset (CIFAR10 or CIFAR100) and the target dataset (ImageNet). Code is available at https://github.com/chenxin061/pdarts

The emerging treatment landscape of targeted therapy in non-small-cell lung cancer.

Abstract: Lung cancer is one of the most common cancer in the world. In 2018, there were over 2 million new cases of lung cancer and over 1.7 million deaths were attributed to lung cancer. Targeted therapy has emerged as an important mean of the disease management for patients with non-small-cell lung cancer (NSCLC). Herein, we review and analyze recent literature, discuss the targeting pathways and ongoing clinical trials in lung cancer. Chemotherapy is no longer the best available treatment for all patients. Therapeutic decisions should be guided by an understanding of the molecular features of patient’s tumor tissues. The future gains will likely emerge from finding optimal ways of combining targeted therapy, immunotherapy, and chemotherapy.

Progressive Differentiable Architecture Search: Bridging the Depth Gap between Search and Evaluation

Abstract: Recently, differentiable search methods have made major progress in reducing the computational costs of neural architecture search. However, these approaches often report lower accuracy in evaluating the searched architecture or transferring it to another dataset. This is arguably due to the large gap between the architecture depths in search and evaluation scenarios. In this paper, we present an efficient algorithm which allows the depth of searched architectures to grow gradually during the training procedure. This brings two issues, namely, heavier computational overheads and weaker search stability, which we solve using search space approximation and regularization, respectively. With a significantly reduced search time (~7 hours on a single GPU), our approach achieves state-of-the-art performance on both the proxy dataset (CIFAR10 or CIFAR100) and the target dataset (ImageNet). Code is available at this https URL.

Hydrogen Evolution and Oxidation: Mechanistic Studies and Material Advances

Abstract: Electrochemical energy storage and conversion through hydrogen is essential for a clean and sustainable energy system. Highly efficient hydrogen electrocatalysts play a key role in the electrochemical transformation reactions. A comprehensive understanding of the hydrogen reaction kinetics and mechanisms is critical for the catalyst design and development. Especially pH-dependent hydrogen evolution and oxidation reaction (HER/HOR) kinetics receives increasing interest, and understanding its origin adds new knowledge to fundamental hydrogen electrocatalysis. Here, a detailed description of kinetic analysis and reaction mechanisms for HER/HOR, and a brief summary about recent development of highly efficient and cost-effective hydrogen electrocatalysts are presented. Lastly, recent advances in the fundamental understanding of pH-dependent hydrogen electrocatalysis are discussed.

Checkpoint blockade and nanosonosensitizer-augmented noninvasive sonodynamic therapy combination reduces tumour growth and metastases in mice

Abstract: Combined checkpoint blockade (e.g., PD1/PD-L1) with traditional clinical therapies can be hampered by side effects and low tumour-therapeutic outcome, hindering broad clinical translation. Here we report a combined tumour-therapeutic modality based on integrating nanosonosensitizers-augmented noninvasive sonodynamic therapy (SDT) with checkpoint-blockade immunotherapy. All components of the nanosonosensitizers (HMME/R837@Lip) are clinically approved, wherein liposomes act as carriers to co-encapsulate sonosensitizers (hematoporphyrin monomethyl ether (HMME)) and immune adjuvant (imiquimod (R837)). Using multiple tumour models, we demonstrate that combining nanosonosensitizers-augmented SDT with anti-PD-L1 induces an anti-tumour response, which not only arrests primary tumour progression, but also prevents lung metastasis. Furthermore, the combined treatment strategy offers a long-term immunological memory function, which can protect against tumour rechallenge after elimination of the initial tumours. Therefore, this work represents a proof-of-concept combinatorial tumour therapeutics based on noninvasive tumours-therapeutic modality with immunotherapy.

Macrophage Polarization in Physiological and Pathological Pregnancy.

Abstract: The immunology of pregnancy is complex and poorly defined. During the complex process of pregnancy, macrophages secrete many cytokines/chemokines and play pivotal roles in the maintenance of maternal-fetal tolerance. Here, we summarized the current knowledge of macrophage polarization and the mechanisms involved in physiological or pathological pregnancy processes, including miscarriage, preeclampsia, and preterm birth. Although current evidence provides a compelling argument that macrophages are important in pregnancy, our understanding of the roles and mechanisms of macrophages in pregnancy is still rudimentary. Since macrophages exhibit functional plasticity, they may be ideal targets for therapeutic manipulation during pathological pregnancy. Additional studies are needed to better define the functions and mechanisms of various macrophage subsets in both normal and pathological pregnancy.

Superior electromagnetic interference shielding 3D graphene nanoplatelets/reduced graphene oxide foam/epoxy nanocomposites with high thermal conductivity

Abstract: How to rationally design the microstructure of polymer nanocomposites to significantly improve their electromagnetic interference shielding effectiveness (EMI SE) is still a great challenge. Herein, we developed a template method for fabricating 3D porous graphene nanoplatelets/reduced graphene oxide foam/epoxy (GNPs/rGO/EP) nanocomposites, in which 3D rGO foam embedded with GNPs constructs a 3D electrical and thermal conductive network in the EP matrix. The 3D rGO framework resolves the agglomeration problem of GNPs, acts as an efficient bunch of channels for electrical transport and attenuates the entered electromagnetic wave. Benefiting from this 3D nanohybrid framework, the GNPs/rGO/EP nanocomposites containing 0.1 wt% rGO and 20.4 wt% GNPs exhibit an EMI SE value of 51 dB in the X-band range, an almost 292% improvement relative to the rGO/EP nanocomposites (∼13 dB) and 240% enhancement compared with the GNPs/EP nanocomposites without 3D microstructures (∼15 dB) and an excellent thermal conductivity of 1.56 W mK−1 and electrical conductivity up to 179.2 S m−1. This work provides a new strategy for the design of muti-functional epoxy nanocomposites for EMI shielding and efficient heat dissipation.

High-performance bifunctional oxygen electrocatalysts for zinc-air batteries over mesoporous Fe/Co-N-C nanofibers with embedding FeCo alloy nanoparticles

Abstract: Mesoporous Fe/Co-N-C nanofibers with embedding FeCo nanoparticles (denote as FeCo@MNC) have been prepared from electrospun Fe/Co-N coordination compounds with bicomponent polymers consisting of polyvinylpyrrolidone (PVP) and polyacrylonitrile (PAN). The as-fabricated hybrid nanofibers exhibited one-dimensional mesoporous morphology, a large BET surface area and uniformly distributed active sites (e.g FeCo alloy, Fe/Co-N). It elucidated that the co-existence of FeCo alloy and Fe/Co-N active sites could promote the catalytic activity of ORR and OER simultaneously. Essentially, the unique one-dimensional of nanofiber with observable porous morphology has indispensable contribution to charge transportation and exposure of active sites to O2 adsorption when assembled into a rechargeable zinc-air battery. As a result, FeCo@MNC exhibited a low discharge-charge voltage gap (e.g. 0.9 V, discharge-charge at 20 mA cm−2), higher power density (e.g. 115 mW cm−2, at 143 mA cm−2) and stability.

J-Aggregates of Cyanine Dye for NIR-II in Vivo Dynamic Vascular Imaging beyond 1500 nm

Abstract: Light in the second near-infrared window, especially beyond 1500 nm, shows enhanced tissue transparency for high-resolution in vivo optical bioimaging due to decreased tissue scattering, absorption, and autofluorescence. Despite some inorganic luminescent nanoparticles have been developed to improve the bioimaging around 1500 nm, it is still a great challenge to synthesize organic molecules with the absorption and emission toward this region. Here, we present J-aggregates with 1360 nm absorption and 1370 nm emission formed by self-assembly of amphiphilic cyanine dye FD-1080 and 1,2-dimyristoyl-sn-glycero-3-phosphocholine. Molecular dynamics simulations were further employed to illustrate the self-assembly process. Superior spatial resolution and high signal-to-background ratio of J-aggregates were demonstrated for noninvasive brain and hindlimb vasculature bioimaging beyond 1500 nm. The efficacy evaluation of the clinically used hypotensor is successfully achieved by high-resolution in vivo dynamic vascular imaging with J-aggregates.

High performance n-type Ag 2 Se film on nylon membrane for flexible thermoelectric power generator

Abstract: Researches on flexible thermoelectric materials usually focus on conducting polymers and conducting polymer-based composites; however, it is a great challenge to obtain high thermoelectric properties comparable to inorganic counterparts. Here, we report an n-type Ag2Se film on flexible nylon membrane with an ultrahigh power factor ~987.4 ± 104.1 μWm−1K−2 at 300 K and an excellent flexibility (93% of the original electrical conductivity retention after 1000 bending cycles around a 8-mm diameter rod). The flexibility is attributed to a synergetic effect of the nylon membrane and the Ag2Se film intertwined with numerous high-aspect-ratio Ag2Se grains. A thermoelectric prototype composed of 4-leg of the Ag2Se film generates a voltage and a maximum power of 18 mV and 460 nW, respectively, at a temperature difference of 30 K. This work opens opportunities of searching for high performance thermoelectric film for flexible thermoelectric devices. Although flexible thermoelectric materials based on conducting polymers are attractive for energy harvesting, their performance is inferior to their inorganic counterparts. Here, the authors present a facile method to deliver inorganic nanowire films with high power factor and flexibility.

PC-DARTS: Partial Channel Connections for Memory-Efficient Architecture Search

Abstract: Differentiable architecture search (DARTS) provided a fast solution in finding effective network architectures, but suffered from large memory and computing overheads in jointly training a super-network and searching for an optimal architecture. In this paper, we present a novel approach, namely, Partially-Connected DARTS, by sampling a small part of super-network to reduce the redundancy in exploring the network space, thereby performing a more efficient search without comprising the performance. In particular, we perform operation search in a subset of channels while bypassing the held out part in a shortcut. This strategy may suffer from an undesired inconsistency on selecting the edges of super-net caused by sampling different channels. We alleviate it using edge normalization, which adds a new set of edge-level parameters to reduce uncertainty in search. Thanks to the reduced memory cost, PC-DARTS can be trained with a larger batch size and, consequently, enjoys both faster speed and higher training stability. Experimental results demonstrate the effectiveness of the proposed method. Specifically, we achieve an error rate of 2.57% on CIFAR10 with merely 0.1 GPU-days for architecture search, and a state-of-the-art top-1 error rate of 24.2% on ImageNet (under the mobile setting) using 3.8 GPU-days for search. Our code has been made available at: this https URL.

One‐Dimensional Fe2P Acts as a Fenton Agent in Response to NIR II Light and Ultrasound for Deep Tumor Synergetic Theranostics

Abstract: The stringent reaction conditions for an effective Fenton reaction (pH range of 3-4) hinders its application in cancer therapy. Therefore, how to improve the efficiency of the Fenton reaction in a tumor site has been the main obstacle in chemodynamic therapy (CDT). Herein, we report biocompatible one-dimensional (1D) ferrous phosphide nanorods (FP NRs) with ultrasound (US)- and photothermal (PT)-enhanced Fenton properties and excellent photothermal conversion efficiency (56.6 %) in the NIR II window, showing synergistic therapeutic properties. Additionally, the high photothermal conversion efficiency and excellent traverse relaxivity (277.79 mm-1 s-1 ) of the FP NRs means they are excellent photoacoustic imaging (PAI) and magnetic resonance imaging (MRI) agents. This is the first report on exploiting the response of metallic phosphides to NIR II laser (1064 nm) and ultrasound to improve the CDT effect with a high therapeutic effect and PA/MR imaging.

Conductive diamond: synthesis, properties, and electrochemical applications

Abstract: Conductive diamond possesses unique features as compared to other solid electrodes, such as a wide electrochemical potential window, a low and stable background current, relatively rapid rates of electron-transfer for soluble redox systems without conventional pretreatment, long-term responses, stability, biocompatibility, and a rich surface chemistry. Conductive diamond microcrystalline and nanocrystalline films, structures and particles have been prepared using a variety of approaches. Given these highly desirable attributes, conductive diamond has found extensive use as an enabling electrode across a variety of fields encompassing chemical and biochemical sensing, environmental degradation, electrosynthesis, electrocatalysis, and energy storage and conversion. This review provides an overview of the fundamental properties and highlights recent progress and achievements in the growth of boron-doped (metal-like) and nitrogen and phosphorus-doped (semi-conducting) diamond and hydrogen-terminated undoped diamond electrodes. Applications in electroanalysis, environmental degradation, electrosynthesis electrocatalysis, and electrochemical energy storage are also discussed. Diamond electrochemical devices utilizing micro-scale, ultramicro-scale, and nano-scale electrodes as well as their counterpart arrays are viewed. The challenges and future research directions of conductive diamond are discussed and outlined. This review will be important and informative for chemists, biochemists, physicists, materials scientists, and engineers engaged in the use of these novel forms of carbon.