Showing papers by "Academia Sinica published in 2020"

YOLOv4: Optimal Speed and Accuracy of Object Detection

Abstract: There are a huge number of features which are said to improve Convolutional Neural Network (CNN) accuracy. Practical testing of combinations of such features on large datasets, and theoretical justification of the result, is required. Some features operate on certain models exclusively and for certain problems exclusively, or only for small-scale datasets; while some features, such as batch-normalization and residual-connections, are applicable to the majority of models, tasks, and datasets. We assume that such universal features include Weighted-Residual-Connections (WRC), Cross-Stage-Partial-connections (CSP), Cross mini-Batch Normalization (CmBN), Self-adversarial-training (SAT) and Mish-activation. We use new features: WRC, CSP, CmBN, SAT, Mish activation, Mosaic data augmentation, CmBN, DropBlock regularization, and CIoU loss, and combine some of them to achieve state-of-the-art results: 43.5% AP (65.7% AP50) for the MS COCO dataset at a realtime speed of ~65 FPS on Tesla V100. Source code is at this https URL

The ATLAS Experiment at the CERN Large Hadron Collider

Abstract: The ATLAS detector as installed in its experimental cavern at point 1 at CERN is described in this paper. A brief overview of the expected performance of the detector when the Large Hadron Collider begins operation is also presented.

CSPNet: A New Backbone that can Enhance Learning Capability of CNN

Abstract: Neural networks have enabled state-of-the-art approaches to achieve incredible results on computer vision tasks such as object detection. However, such success greatly relies on costly computation resources, which hinders people with cheap devices from appreciating the advanced technology. In this paper, we propose Cross Stage Partial Network (CSPNet) to mitigate the problem that previous works require heavy inference computations from the network architecture perspective. We attribute the problem to the duplicate gradient information within network optimization. The proposed networks respect the variability of the gradients by integrating feature maps from the beginning and the end of a network stage, which, in our experiments, reduces computations by 20% with equivalent or even superior accuracy on the ImageNet dataset, and significantly outperforms state-of-the-art approaches in terms of AP 50 on the MS COCO object detection dataset. The CSPNet is easy to implement and general enough to cope with architectures based on ResNet, ResNeXt, and DenseNet.

Pan-cancer analysis of whole genomes

Abstract: Cancer is driven by genetic change, and the advent of massively parallel sequencing has enabled systematic documentation of this variation at the whole-genome scale1,2,3. Here we report the integrative analysis of 2,658 whole-cancer genomes and their matching normal tissues across 38 tumour types from the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium of the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA). We describe the generation of the PCAWG resource, facilitated by international data sharing using compute clouds. On average, cancer genomes contained 4–5 driver mutations when combining coding and non-coding genomic elements; however, in around 5% of cases no drivers were identified, suggesting that cancer driver discovery is not yet complete. Chromothripsis, in which many clustered structural variants arise in a single catastrophic event, is frequently an early event in tumour evolution; in acral melanoma, for example, these events precede most somatic point mutations and affect several cancer-associated genes simultaneously. Cancers with abnormal telomere maintenance often originate from tissues with low replicative activity and show several mechanisms of preventing telomere attrition to critical levels. Common and rare germline variants affect patterns of somatic mutation, including point mutations, structural variants and somatic retrotransposition. A collection of papers from the PCAWG Consortium describes non-coding mutations that drive cancer beyond those in the TERT promoter4; identifies new signatures of mutational processes that cause base substitutions, small insertions and deletions and structural variation5,6; analyses timings and patterns of tumour evolution7; describes the diverse transcriptional consequences of somatic mutation on splicing, expression levels, fusion genes and promoter activity8,9; and evaluates a range of more-specialized features of cancer genomes8,10,11,12,13,14,15,16,17,18.

Development of therapeutic antibodies for the treatment of diseases

Abstract: It has been more than three decades since the first monoclonal antibody was approved by the United States Food and Drug Administration (US FDA) in 1986, and during this time, antibody engineering has dramatically evolved. Current antibody drugs have increasingly fewer adverse effects due to their high specificity. As a result, therapeutic antibodies have become the predominant class of new drugs developed in recent years. Over the past five years, antibodies have become the best-selling drugs in the pharmaceutical market, and in 2018, eight of the top ten bestselling drugs worldwide were biologics. The global therapeutic monoclonal antibody market was valued at approximately US$115.2 billion in 2018 and is expected to generate revenue of $150 billion by the end of 2019 and $300 billion by 2025. Thus, the market for therapeutic antibody drugs has experienced explosive growth as new drugs have been approved for treating various human diseases, including many cancers, autoimmune, metabolic and infectious diseases. As of December 2019, 79 therapeutic mAbs have been approved by the US FDA, but there is still significant growth potential. This review summarizes the latest market trends and outlines the preeminent antibody engineering technologies used in the development of therapeutic antibody drugs, such as humanization of monoclonal antibodies, phage display, the human antibody mouse, single B cell antibody technology, and affinity maturation. Finally, future applications and perspectives are also discussed.

The 16th Data Release of the Sloan Digital Sky Surveys: First Release from the APOGEE-2 Southern Survey and Full Release of eBOSS Spectra

Abstract: This paper documents the sixteenth data release (DR16) from the Sloan Digital Sky Surveys; the fourth and penultimate from the fourth phase (SDSS-IV). This is the first release of data from the southern hemisphere survey of the Apache Point Observatory Galactic Evolution Experiment 2 (APOGEE-2); new data from APOGEE-2 North are also included. DR16 is also notable as the final data release for the main cosmological program of the Extended Baryon Oscillation Spectroscopic Survey (eBOSS), and all raw and reduced spectra from that project are released here. DR16 also includes all the data from the Time Domain Spectroscopic Survey (TDSS) and new data from the SPectroscopic IDentification of ERosita Survey (SPIDERS) programs, both of which were co-observed on eBOSS plates. DR16 has no new data from the Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) survey (or the MaNGA Stellar Library "MaStar"). We also preview future SDSS-V operations (due to start in 2020), and summarize plans for the final SDSS-IV data release (DR17).

Scaled-YOLOv4: Scaling Cross Stage Partial Network

Abstract: We show that the YOLOv4 object detection neural network based on the CSP approach, scales both up and down and is applicable to small and large networks while maintaining optimal speed and accuracy. We propose a network scaling approach that modifies not only the depth, width, resolution, but also structure of the network. YOLOv4-large model achieves state-of-the-art results: 55.4% AP (73.3% AP50) for the MS COCO dataset at a speed of 15 FPS on Tesla V100, while with the test time augmentation, YOLOv4-large achieves 55.8% AP (73.2 AP50). To the best of our knowledge, this is currently the highest accuracy on the COCO dataset among any published work. The YOLOv4-tiny model achieves 22.0% AP (42.0% AP50) at a speed of 443 FPS on RTX 2080Ti, while by using TensorRT, batch size = 4 and FP16-precision the YOLOv4-tiny achieves 1774 FPS.

Future Physics Programme of BESIII

Abstract: There has recently been a dramatic renewal of interest in hadron spectroscopy and charm physics. This renaissance has been driven in part by the discovery of a plethora of charmonium-like XYZ states at BESIII and B factories, and the observation of an intriguing proton-antiproton threshold enhancement and the possibly related X(1835) meson state at BESIII, as well as the threshold measurements of charm mesons and charm baryons. We present a detailed survey of the important topics in tau-charm physics and hadron physics that can be further explored at BESIII during the remaining operation period of BEPCII. This survey will help in the optimization of the data-taking plan over the coming years, and provides physics motivation for the possible upgrade of BEPCII to higher luminosity.

Search for electroweak production of charginos and sleptons decaying into final states with two leptons and missing transverse momentum in √s = 13 TeV pp collisions using the ATLAS detector

Abstract: A search for the electroweak production of charginos and sleptons decaying into final states with two electrons or muons is presented. The analysis is based on 139 fb$^{-1}$ of proton–proton collisions recorded by the ATLAS detector at the Large Hadron Collider at $\sqrt{s}=13$ $\text {TeV}$. Three R-parity-conserving scenarios where the lightest neutralino is the lightest supersymmetric particle are considered: the production of chargino pairs with decays via either W bosons or sleptons, and the direct production of slepton pairs. The analysis is optimised for the first of these scenarios, but the results are also interpreted in the others. No significant deviations from the Standard Model expectations are observed and limits at 95% confidence level are set on the masses of relevant supersymmetric particles in each of the scenarios. For a massless lightest neutralino, masses up to 420 $\text {Ge}\text {V}$ are excluded for the production of the lightest-chargino pairs assuming W-boson-mediated decays and up to 1 $\text {TeV}$ for slepton-mediated decays, whereas for slepton-pair production masses up to 700 $\text {Ge}\text {V}$ are excluded assuming three generations of mass-degenerate sleptons.

Structural basis for the neutralization of SARS-CoV-2 by an antibody from a convalescent patient.

Abstract: The COVID-19 pandemic has had an unprecedented health and economic impact and there are currently no approved therapies. We have isolated an antibody, EY6A, from an individual convalescing from COVID-19 and have shown that it neutralizes SARS-CoV-2 and cross-reacts with SARS-CoV-1. EY6A Fab binds the receptor binding domain (RBD) of the viral spike glycoprotein tightly (KD of 2 nM), and a 2.6-A-resolution crystal structure of an RBD-EY6A Fab complex identifies the highly conserved epitope, away from the ACE2 receptor binding site. Residues within this footprint are key to stabilizing the pre-fusion spike. Cryo-EM analyses of the pre-fusion spike incubated with EY6A Fab reveal a complex of the intact spike trimer with three Fabs bound and two further multimeric forms comprising the destabilized spike attached to Fab. EY6A binds what is probably a major neutralizing epitope, making it a candidate therapeutic for COVID-19.

Quantum-limit Chern topological magnetism in TbMn6Sn6.

Abstract: The quantum-level interplay between geometry, topology and correlation is at the forefront of fundamental physics1–15. Kagome magnets are predicted to support intrinsic Chern quantum phases owing to their unusual lattice geometry and breaking of time-reversal symmetry14,15. However, quantum materials hosting ideal spin–orbit-coupled kagome lattices with strong out-of-plane magnetization are lacking16–21. Here, using scanning tunnelling microscopy, we identify a new topological kagome magnet, TbMn6Sn6, that is close to satisfying these criteria. We visualize its effectively defect-free, purely manganese-based ferromagnetic kagome lattice with atomic resolution. Remarkably, its electronic state shows distinct Landau quantization on application of a magnetic field, and the quantized Landau fan structure features spin-polarized Dirac dispersion with a large Chern gap. We further demonstrate the bulk–boundary correspondence between the Chern gap and the topological edge state, as well as the Berry curvature field correspondence of Chern gapped Dirac fermions. Our results point to the realization of a quantum-limit Chern phase in TbMn6Sn6, and may enable the observation of topological quantum phenomena in the RMn6Sn6 (where R is a rare earth element) family with a variety of magnetic structures. Our visualization of the magnetic bulk–boundary–Berry correspondence covering real space and momentum space demonstrates a proof-of-principle method for revealing topological magnets. Scanning tunnelling microscopy is used to reveal a new topological kagome magnet with an intrinsic Chern quantum phase, which shows a distinct Landau fan structure with a large Chern gap.

Identification of type 2 diabetes loci in 433,540 East Asian individuals

Abstract: Meta-analyses of genome-wide association studies (GWAS) have identified more than 240 loci that are associated with type 2 diabetes (T2D)1,2; however, most of these loci have been identified in analyses of individuals with European ancestry. Here, to examine T2D risk in East Asian individuals, we carried out a meta-analysis of GWAS data from 77,418 individuals with T2D and 356,122 healthy control individuals. In the main analysis, we identified 301 distinct association signals at 183 loci, and across T2D association models with and without consideration of body mass index and sex, we identified 61 loci that are newly implicated in predisposition to T2D. Common variants associated with T2D in both East Asian and European populations exhibited strongly correlated effect sizes. Previously undescribed associations include signals in or near GDAP1, PTF1A, SIX3, ALDH2, a microRNA cluster, and genes that affect the differentiation of muscle and adipose cells3. At another locus, expression quantitative trait loci at two overlapping T2D signals affect two genes-NKX6-3 and ANK1-in different tissues4-6. Association studies in diverse populations identify additional loci and elucidate disease-associated genes, biology, and pathways.

The causes of sea-level rise since 1900

Abstract: The rate of global-mean sea-level rise since 1900 has varied over time, but the contributing factors are still poorly understood1 Previous assessments found that the summed contributions of ice-mass loss, terrestrial water storage and thermal expansion of the ocean could not be reconciled with observed changes in global-mean sea level, implying that changes in sea level or some contributions to those changes were poorly constrained2,3 Recent improvements to observational data, our understanding of the main contributing processes to sea-level change and methods for estimating the individual contributions, mean another attempt at reconciliation is warranted Here we present a probabilistic framework to reconstruct sea level since 1900 using independent observations and their inherent uncertainties The sum of the contributions to sea-level change from thermal expansion of the ocean, ice-mass loss and changes in terrestrial water storage is consistent with the trends and multidecadal variability in observed sea level on both global and basin scales, which we reconstruct from tide-gauge records Ice-mass loss-predominantly from glaciers-has caused twice as much sea-level rise since 1900 as has thermal expansion Mass loss from glaciers and the Greenland Ice Sheet explains the high rates of global sea-level rise during the 1940s, while a sharp increase in water impoundment by artificial reservoirs is the main cause of the lower-than-average rates during the 1970s The acceleration in sea-level rise since the 1970s is caused by the combination of thermal expansion of the ocean and increased ice-mass loss from Greenland Our results reconcile the magnitude of observed global-mean sea-level rise since 1900 with estimates based on the underlying processes, implying that no additional processes are required to explain the observed changes in sea level since 1900

Deeply conserved synteny resolves early events in vertebrate evolution

Abstract: Although it is widely believed that early vertebrate evolution was shaped by ancient whole-genome duplications, the number, timing and mechanism of these events remain elusive. Here, we infer the history of vertebrates through genomic comparisons with a new chromosome-scale sequence of the invertebrate chordate amphioxus. We show how the karyotypes of amphioxus and diverse vertebrates are derived from 17 ancestral chordate linkage groups (and 19 ancestral bilaterian groups) by fusion, rearrangement and duplication. We resolve two distinct ancient duplications based on patterns of chromosomal conserved synteny. All extant vertebrates share the first duplication, which occurred in the mid/late Cambrian by autotetraploidization (that is, direct genome doubling). In contrast, the second duplication is found only in jawed vertebrates and occurred in the mid–late Ordovician by allotetraploidization (that is, genome duplication following interspecific hybridization) from two now-extinct progenitors. This complex genomic history parallels the diversification of vertebrate lineages in the fossil record. Genomic comparisons with a new amphioxus chromosome-scale genome assembly reveal details of the early evolution of vertebrate genomes.

Metalens-array-based high-dimensional and multiphoton quantum source.

Abstract: The development of two-dimensional metasurfaces has shown great potential in quantum-optical technologies because of the excellent flexibility in light-field manipulation. By integrating a metalens array with a nonlinear crystal, we demonstrate a 100-path spontaneous parametric down-conversion photon-pair source in a 10 × 10 array, which shows promise for high-dimensional entanglement and multiphoton-state generation. We demonstrate two-, three- and four-dimensional two-photon path entanglement with different phases encoded by metalenses with fidelities of 98.4, 96.6, and 95.0%, respectively. Furthermore, four-photon and six-photon generation is observed with high indistinguishability of photons generated from different metalenses. Our metalens-array-based quantum photon source is compact, stable, and controllable, indicating a new platform for integrated quantum devices.

Gravitational Test beyond the First Post-Newtonian Order with the Shadow of the M87 Black Hole

Abstract: The 2017 Event Horizon Telescope (EHT) observations of the central source in M87 have led to the first measurement of the size of a black-hole shadow. This observation offers a new and clean gravitational test of the black-hole metric in the strong-field regime. We show analytically that spacetimes that deviate from the Kerr metric but satisfy weak-field tests can lead to large deviations in the predicted black-hole shadows that are inconsistent with even the current EHT measurements. We use numerical calculations of regular, parametric, non-Kerr metrics to identify the common characteristic among these different parametrizations that control the predicted shadow size. We show that the shadow-size measurements place significant constraints on deviation parameters that control the second post-Newtonian and higher orders of each metric and are, therefore, inaccessible to weak-field tests. The new constraints are complementary to those imposed by observations of gravitational waves from stellar-mass sources.

Performance of electron and photon triggers in ATLAS during LHC Run 2

Abstract: Electron and photon triggers covering transverse energies from 5 GeV to several TeV are essential for the ATLAS experiment to record signals for a wide variety of physics: from Standard Model processes to searches for new phenomena in both proton–proton and heavy-ion collisions. To cope with a fourfold increase of peak LHC luminosity from 2015 to 2018 (Run 2), to 2.1×1034cm-2s-1, and a similar increase in the number of interactions per beam-crossing to about 60, trigger algorithms and selections were optimised to control the rates while retaining a high efficiency for physics analyses. For proton–proton collisions, the single-electron trigger efficiency relative to a single-electron offline selection is at least 75% for an offline electron of 31 GeV, and rises to 96% at 60 GeV; the trigger efficiency of a 25 GeV leg of the primary diphoton trigger relative to a tight offline photon selection is more than 96% for an offline photon of 30 GeV. For heavy-ion collisions, the primary electron and photon trigger efficiencies relative to the corresponding standard offline selections are at least 84% and 95%, respectively, at 5 GeV above the corresponding trigger threshold.

Search for Heavy Higgs Bosons Decaying into Two Tau Leptons with the ATLAS Detector Using pp Collisions at s =13 TeV

Abstract: A search for heavy neutral Higgs bosons is performed using the LHC Run 2 data, corresponding to an integrated luminosity of 139 fb^{-1} of proton-proton collisions at sqrt[s]=13 TeV recorded with the ATLAS detector. The search for heavy resonances is performed over the mass range 0.2-2.5 TeV for the τ^{+}τ^{-} decay with at least one τ-lepton decaying into final states with hadrons. The data are in good agreement with the background prediction of the standard model. In the M_{h}^{125} scenario of the minimal supersymmetric standard model, values of tanβ>8 and tanβ>21 are excluded at the 95% confidence level for neutral Higgs boson masses of 1.0 and 1.5 TeV, respectively, where tanβ is the ratio of the vacuum expectation values of the two Higgs doublets.

Ancient DNA indicates human population shifts and admixture in northern and southern China

Abstract: Human genetic history in East Asia is poorly understood. To clarify population relationships, we obtained genome-wide data from 26 ancient individuals from northern and southern East Asia spanning 9500 to 300 years ago. Genetic differentiation in this region was higher in the past than the present, which reflects a major episode of admixture involving northern East Asian ancestry spreading across southern East Asia after the Neolithic, thereby transforming the genetic ancestry of southern China. Mainland southern East Asian and Taiwan Strait island samples from the Neolithic show clear connections with modern and ancient individuals with Austronesian-related ancestry, which supports an origin in southern China for proto-Austronesians. Connections among Neolithic coastal groups from Siberia and Japan to Vietnam indicate that migration and gene flow played an important role in the prehistory of coastal Asia.

The Trans-Ancestral Genomic Architecture of Glycaemic Traits

Abstract: Glycaemic traits are used to diagnose and monitor type 2 diabetes, and cardiometabolic health. To date, most genetic studies of glycaemic traits have focused on individuals of European ancestry. Here, we aggregated genome-wide association studies in up to 281,416 individuals without diabetes (30% non-European ancestry) with fasting glucose, 2h-glucose post-challenge, glycated haemoglobin, and fasting insulin data. Trans-ancestry and single-ancestry meta-analyses identified 242 loci (99 novel; P

Erratum: The Belle II Physics Book (Progress of Theoretical and Experimental Physics (2019) 2019 (123C01) DOI: 10.1093/ptep/ptz106)

Abstract: In the original version of this manuscript, an error was introduced on pp352. '2.7nb:1.6nb' has been corrected to '2.4nb:1.3nb' in the current online and printed version. doi:10.1093/ptep/ptz106.

Pop Music Transformer: Beat-based Modeling and Generation of Expressive Pop Piano Compositions

Abstract: A great number of deep learning based models have been recently proposed for automatic music composition. Among these models, the Transformer stands out as a prominent approach for generating expressive classical piano performance with a coherent structure of up to one minute. The model is powerful in that it learns abstractions of data on its own, without much human-imposed domain knowledge or constraints. In contrast with this general approach, this paper shows that Transformers can do even better for music modeling, when we improve the way a musical score is converted into the data fed to a Transformer model. In particular, we seek to impose a metrical structure in the input data, so that Transformers can be more easily aware of the beat-bar-phrase hierarchical structure in music. The new data representation maintains the flexibility of local tempo changes, and provides hurdles to control the rhythmic and harmonic structure of music. With this approach, we build a Pop Music Transformer that composes Pop piano music with better rhythmic structure than existing Transformer models.

High-Safety and High-Energy-Density Lithium Metal Batteries in a Novel Ionic-Liquid Electrolyte.

Abstract: Rechargeable lithium metal batteries are next generation energy storage devices with high energy density, but face challenges in achieving high energy density, high safety, and long cycle life. Here, lithium metal batteries in a novel nonflammable ionic-liquid (IL) electrolyte composed of 1-ethyl-3-methylimidazolium (EMIm) cations and high-concentration bis(fluorosulfonyl)imide (FSI) anions, with sodium bis(trifluoromethanesulfonyl)imide (NaTFSI) as a key additive are reported. The Na ion participates in the formation of hybrid passivation interphases and contributes to dendrite-free Li deposition and reversible cathode electrochemistry. The electrolyte of low viscosity allows practically useful cathode mass loading up to ≈16 mg cm-2 . Li anodes paired with lithium cobalt oxide (LiCoO2 ) and lithium nickel cobalt manganese oxide (LiNi0.8 Co0.1 Mn0.1 O2 , NCM 811) cathodes exhibit 99.6-99.9% Coulombic efficiencies, high discharge voltages up to 4.4 V, high specific capacity and energy density up to ≈199 mAh g-1 and ≈765 Wh kg-1 respectively, with impressive cycling performances over up to 1200 cycles. Highly stable passivation interphases formed on both electrodes in the novel IL electrolyte are the key to highly reversible lithium metal batteries, especially for Li-NMC 811 full batteries.

Construction hierarchically mesoporous/microporous materials based on block copolymer and covalent organic framework

Abstract: Hierarchically ordered mesoporous materials have gained significant scientific attention due to their high surface areas, uniform porosity over various lengths scales, high-volume storage capability, shape selectivity, enhanced mass transport and diffusion. These materials have been widely applied in the fields of photocatalysis, separation, adsorption, photovoltaic solar cells, energy storage and conversion, chemical sensing, and drug delivery. In general, hierarchically porous materials can be generated in two ways: through sol-gel template approaches and a combination of surfactant-assisted procedures. In this review, we discuss recent progress in the preparation, properties, and potential applications of four hierarchically ordered porous materials: hierarchical mesoporous silica, mesoporous phenolic/carbon, nitrogen-doped mesoporous carbon, and mesoporous/microporous covalent organic frameworks.

Highly efficient and tumor-selective nanoparticles for dual-targeted immunogene therapy against cancer

Abstract: While immunotherapy holds great promise for combating cancer, the limited efficacy due to an immunosuppressive tumor microenvironment and systemic toxicity hinder the broader application of cancer immunotherapy. Here, we report a combinatorial immunotherapy approach that uses a highly efficient and tumor-selective gene carrier to improve anticancer efficacy and circumvent the systemic toxicity. In this study, we engineered tumor-targeted lipid-dendrimer-calcium-phosphate (TT-LDCP) nanoparticles (NPs) with thymine-functionalized dendrimers that exhibit not only enhanced gene delivery capacity but also immune adjuvant properties by activating the stimulator of interferon genes (STING)–cGAS pathway. TT-LDCP NPs delivered siRNA against immune checkpoint ligand PD-L1 and immunostimulatory IL-2–encoding plasmid DNA to hepatocellular carcinoma (HCC), increased tumoral infiltration and activation of CD8+ T cells, augmented the efficacy of cancer vaccine immunotherapy, and suppressed HCC progression. Our work presents nanotechnology-enabled dual delivery of siRNA and plasmid DNA that selectively targets and reprograms the immunosuppressive tumor microenvironment to improve cancer immunotherapy.