arXiv subjects

Jun Gao

Publications and source records attributed to Jun Gao.

At least 217 records · Page 12Linked to original sources

arXiv2022

Machine learning of log-likelihood functions in global analysis of parton distributions

Modern analysis on parton distribution functions (PDFs) requires calculations of the log-likelihood functions from thousands of experimental data points, and scans of multi-dimensional parameter space with tens of degrees of freedom. In conventional analysis the Hessian approximation has been widely used for the estimation of the PDF uncertainties.The Lagrange Multiplier (LM) scan while being a more faithful method is less used due to computational limitations, and is the main focus of this study. We propose to use Neural Networks (NNs) and machine learning techniques to model the profile of the log-likelihood functions or cross sections for multi-dimensional parameter space in order to overcome those limitations which work beyond the quadratic approximations and meanwhile ensures efficient scans of the full parameter space. We demonstrate the efficiency of the new approach in the framework of the CT18 global analysis of PDFs by constructing NNs for various target functions, and performing LM scans on PDFs and cross sections at hadron colliders. We further study the impact of the NOMAD dimuon data on constraining PDFs with the new approach, and find enhanced strange-quark distributions and reduced PDF uncertainties. Moreover, we show how the approach can be used to constrain new physics beyond the Standard Model (BSM) by a joint fit of both PDFs and Wilson coefficients of operators in the SM effective field theory.

hep-ph↗

arXiv2022

Thrust distribution in Higgs decays at the next-to-leading order and beyond

We present predictions for the thrust distribution in hadronic decays of the Higgs boson at the next-to-leading order and the approximate next-to-next-to-leading order. The approximate NNLO corrections are derived from a factorization formula in the soft/collinear phase-space regions. We find large corrections, especially for the gluon channel. The scale variations at the lowest orders tend to underestimate the genuine higher order contributions. The results of this paper is therefore necessary to control the perturbative uncertainties of the theoretical predictions. We also discuss on possible improvements to our results, such as a soft-gluon resummation for the 2-jets limit, and an exact next-to-next-to-leading order calculation for the multi-jets region.

hep-ph↗

arXiv2022

Decay of the charged Higgs boson and the top quark in two-Higgs-doublet model at NNLO in QCD

We present numerical calculations of the partial width of the charged Higgs boson decay into a top quark, $H^- \rightarrow \bar{t} + b + X$, and the partial width of the top quark decay into a light charged Higgs boson $t \rightarrow H^+ + b + X$ at next-to-next-to-leading order (NNLO) in QCD, based on a factorization formula of the jet mass. The NNLO corrections significantly reduce the renormalization scale dependence of the partial decay width in both cases. We show relative size of the NNLO corrections for different charged Higgs boson masses and for different renormalization scales. The NNLO corrections are about 16% (1%) of the leading order widths for the charged Higgs boson mass of 200GeV (2000GeV), while it is quite small for the top quark decay. Our analyses are independent of the detailed structure of the Yukawa couplings, and can be applied to various new physics models, as demonstrated by the decay branching ratio in different types of the two-Higgs-doublet models.

hep-ph↗

arXiv2022

Improving Semantic Segmentation in Transformers using Hierarchical Inter-Level Attention

Existing transformer-based image backbones typically propagate feature information in one direction from lower to higher-levels. This may not be ideal since the localization ability to delineate accurate object boundaries, is most prominent in the lower, high-resolution feature maps, while the semantics that can disambiguate image signals belonging to one object vs. another, typically emerges in a higher level of processing. We present Hierarchical Inter-Level Attention (HILA), an attention-based method that captures Bottom-Up and Top-Down Updates between features of different levels. HILA extends hierarchical vision transformer architectures by adding local connections between features of higher and lower levels to the backbone encoder. In each iteration, we construct a hierarchy by having higher-level features compete for assignments to update lower-level features belonging to them, iteratively resolving object-part relationships. These improved lower-level features are then used to re-update the higher-level features. HILA can be integrated into the majority of hierarchical architectures without requiring any changes to the base model. We add HILA into SegFormer and the Swin Transformer and show notable improvements in accuracy in semantic segmentation with fewer parameters and FLOPS. Project website and code: https://www.cs.toronto.edu/~garyleung/hila/

cs.CV↗

arXiv2022

Model Degradation Hinders Deep Graph Neural Networks

Graph Neural Networks (GNNs) have achieved great success in various graph mining tasks.However, drastic performance degradation is always observed when a GNN is stacked with many layers. As a result, most GNNs only have shallow architectures, which limits their expressive power and exploitation of deep neighborhoods.Most recent studies attribute the performance degradation of deep GNNs to the \textit{over-smoothing} issue. In this paper, we disentangle the conventional graph convolution operation into two independent operations: \textit{Propagation} (\textbf{P}) and \textit{Transformation} (\textbf{T}).Following this, the depth of a GNN can be split into the propagation depth ($D_p$) and the transformation depth ($D_t$). Through extensive experiments, we find that the major cause for the performance degradation of deep GNNs is the \textit{model degradation} issue caused by large $D_t$ rather than the \textit{over-smoothing} issue mainly caused by large $D_p$. Further, we present \textit{Adaptive Initial Residual} (AIR), a plug-and-play module compatible with all kinds of GNN architectures, to alleviate the \textit{model degradation} issue and the \textit{over-smoothing} issue simultaneously. Experimental results on six real-world datasets demonstrate that GNNs equipped with AIR outperform most GNNs with shallow architectures owing to the benefits of both large $D_p$ and $D_t$, while the time costs associated with AIR can be ignored.

cs.LG↗

arXiv2022

Experimental Quantum Simulation of Dynamic Localization on Curved Photonic Lattices

Dynamic localization, which originates from the phenomena of particle evolution suppression under an externally applied AC electric field, has been simulated by suppressed light evolution in periodically-curved photonic arrays. However, experimental studies on their quantitative dynamic transport properties and application for quantum information processing are rare. Here we fabricate one-dimensional and hexagonal two-dimensional arrays, both with sinusoidal curvature. We successfully observe the suppressed single-photon evolution patterns, and for the first time measure the variances to study their transport properties. For one-dimensional arrays, the measured variances match both the analytical electric field calculation and the quantum walk Hamiltonian engineering approach. For hexagonal arrays, as anisotropic effective couplings in four directions are mutually dependent, the analytical approach suffers, while quantum walk conveniently incorporates all anisotropic coupling coefficients in the Hamiltonian and solves its exponential as a whole, yielding consistent variances with our experimental results. Furthermore, we implement a nearly complete localization to show that it can preserve both the initial injection and the wave-packet after some evolution, acting as a memory of a flexible time scale in integrated photonics. We demonstrate a useful quantum simulation of dynamic localization for studying their anisotropic transport properties, and a promising application of dynamic localization as a building block for quantum information processing in integrated photonics.

quant-ph↗

arXiv2022

Interpretable Proof Generation via Iterative Backward Reasoning

We present IBR, an Iterative Backward Reasoning model to solve the proof generation tasks on rule-based Question Answering (QA), where models are required to reason over a series of textual rules and facts to find out the related proof path and derive the final answer. We handle the limitations of existed works in two folds: 1) enhance the interpretability of reasoning procedures with detailed tracking, by predicting nodes and edges in the proof path iteratively backward from the question; 2) promote the efficiency and accuracy via reasoning on the elaborate representations of nodes and history paths, without any intermediate texts that may introduce external noise during proof generation. There are three main modules in IBR, QA and proof strategy prediction to obtain the answer and offer guidance for the following procedure; parent node prediction to determine a node in the existing proof that a new child node will link to; child node prediction to find out which new node will be added to the proof. Experiments on both synthetic and paraphrased datasets demonstrate that IBR has better in-domain performance as well as cross-domain transferability than several strong baselines. Our code and models are available at https://github.com/find-knowledge/IBR .

cs.CL↗

arXiv2022

Understanding PDF uncertainty on the $W$ boson mass measurements in CT18 global analysis

We study the dependence of the transverse mass distribution of the charged lepton and the missing energies on the parton distributions (PDFs) adapted to the $W$ boson mass measurements at the CDF and ATLAS experiments. We compare the shape variations of the distribution induced by different PDFs and find that spread of predictions from different PDF sets can be much larger than the PDF uncertainty predicted by a specific PDF set. We suggest analyzing the experimental data using up-to-date PDFs for a better understanding of the PDF uncertainties in the $W$ boson mass measurements. We further carry out a series of Lagrange multiplier scans to identify the constraints on the transverse mass distribution imposed by individual data sets in the CT18 global analysis. In the case of CDF measurement, the distribution is mostly sensitive to the $d$-quark PDFs at the intermediate $x$ region that is largely constrained by the DIS and Drell-Yan data on the deuteron target, as well as the Tevatron lepton charge asymmetry data.

hep-ph↗

arXiv2022

General heavy-flavor mass scheme for charged-current DIS at NNLO and beyond

Incompleteness in current knowledge of neutrino interactions with nuclear matter imposes a primary limitation in searches for leptonic CP violation carried out at long-baseline neutrino experiments. In this paper, we present a new computation that elevates the theoretical accuracy to next-to-next-to-leading order (NNLO) in QCD for charged-current deeply-inelastic scattering (DIS) processes relevant for ongoing and future neutrino programs. Mass-dependent quark contributions are consistently included across a wide range of momentum transfers in the SACOT-$χ$ general-mass scheme. When appropriate, we further include N$^3$LO corrections in the zero-mass scheme. We show theoretical predictions for several experiments with neutrinos over a wide range of energies and at the upcoming Electron-Ion Collider. Our prediction reduces perturbative uncertainties to $\sim\!1\%$, sufficient for the high-precision objectives of future charged-current DIS measurements, and provides important theoretical inputs to experimental studies of leptonic mixing and CP violations.

hep-ph↗

arXiv2022

Self-Supervised Light Field Depth Estimation Using Epipolar Plane Images

Exploiting light field data makes it possible to obtain dense and accurate depth map. However, synthetic scenes with limited disparity range cannot contain the diversity of real scenes. By training in synthetic data, current learning-based methods do not perform well in real scenes. In this paper, we propose a self-supervised learning framework for light field depth estimation. Different from the existing end-to-end training methods using disparity label per pixel, our approach implements network training by estimating EPI disparity shift after refocusing, which extends the disparity range of epipolar lines. To reduce the sensitivity of EPI to noise, we propose a new input mode called EPI-Stack, which stacks EPIs in the view dimension. This method is less sensitive to noise scenes than traditional input mode and improves the efficiency of estimation. Compared with other state-of-the-art methods, the proposed method can also obtain higher quality results in real-world scenarios, especially in the complex occlusion and depth discontinuity.

cs.CV↗

arXiv2022

Fine Detailed Texture Learning for 3D Meshes with Generative Models

This paper presents a method to reconstruct high-quality textured 3D models from both multi-view and single-view images. The reconstruction is posed as an adaptation problem and is done progressively where in the first stage, we focus on learning accurate geometry, whereas in the second stage, we focus on learning the texture with a generative adversarial network. In the generative learning pipeline, we propose two improvements. First, since the learned textures should be spatially aligned, we propose an attention mechanism that relies on the learnable positions of pixels. Secondly, since discriminator receives aligned texture maps, we augment its input with a learnable embedding which improves the feedback to the generator. We achieve significant improvements on multi-view sequences from Tripod dataset as well as on single-view image datasets, Pascal 3D+ and CUB. We demonstrate that our method achieves superior 3D textured models compared to the previous works. Please visit our web-page for 3D visuals.

cs.CV↗

arXiv2022

REAM$\sharp$: An Enhancement Approach to Reference-based Evaluation Metrics for Open-domain Dialog Generation

The lack of reliable automatic evaluation metrics is a major impediment to the development of open-domain dialogue systems. Various reference-based metrics have been proposed to calculate a score between a predicted response and a small set of references. However, these metrics show unsatisfactory correlations with human judgments. For a reference-based metric, its reliability mainly depends on two factors: its ability to measure the similarity between the predicted response and the reference response, as well as the reliability of the given reference set. Yet, there are few discussions on the latter. Our work attempts to fill this vacancy. We first clarify an assumption on reference-based metrics that, if more high-quality references are added into the reference set, the reliability of the metric will increase. Next, we present REAM$\sharp$: an enhancement approach to Reference-based EvAluation Metrics for open-domain dialogue systems. A prediction model is designed to estimate the reliability of the given reference set. We show how its predicted results can be helpful to augment the reference set, and thus improve the reliability of the metric. Experiments validate both the effectiveness of our prediction model and that the reliability of reference-based metrics improves with the augmented reference sets.

cs.CL↗

arXiv2022

Improving Event Representation via Simultaneous Weakly Supervised Contrastive Learning and Clustering

Representations of events described in text are important for various tasks. In this work, we present SWCC: a Simultaneous Weakly supervised Contrastive learning and Clustering framework for event representation learning. SWCC learns event representations by making better use of co-occurrence information of events. Specifically, we introduce a weakly supervised contrastive learning method that allows us to consider multiple positives and multiple negatives, and a prototype-based clustering method that avoids semantically related events being pulled apart. For model training, SWCC learns representations by simultaneously performing weakly supervised contrastive learning and prototype-based clustering. Experimental results show that SWCC outperforms other baselines on Hard Similarity and Transitive Sentence Similarity tasks. In addition, a thorough analysis of the prototype-based clustering method demonstrates that the learned prototype vectors are able to implicitly capture various relations between events.

cs.CL↗

arXiv2022

NNLO constraints on proton PDFs from the SeaQuest and STAR experiments and other developments in the CTEQ-TEA global analysis

We review progress in the global QCD analysis by the CTEQ-TEA group since the publication of CT18 parton distribution functions (PDFs) in the proton. Specifically, we discuss comparisons of CT18 NNLO predictions with the LHC 13 TeV measurements as well as with the FNAL SeaQuest and BNL STAR data on lepton pair production. The specialized CT18X PDFs approximating saturation effects are compared with the CT18sx PDFs obtained using NLL/NLO small-$x$ resummation. Short summaries are presented for the special CT18 parton distributions with fitted charm and with lattice QCD inputs. A recent comparative analysis of the impact of deuteron nuclear effects on the parton distributions by the CTEQ-JLab and CTEQ-TEA groups is summarized.

hep-ph↗

arXiv2022

JuCify: A Step Towards Android Code Unification for Enhanced Static Analysis

Native code is now commonplace within Android app packages where it co-exists and interacts with Dex bytecode through the Java Native Interface to deliver rich app functionalities. Yet, state-of-the-art static analysis approaches have mostly overlooked the presence of such native code, which, however, may implement some key sensitive, or even malicious, parts of the app behavior. This limitation of the state of the art is a severe threat to validity in a large range of static analyses that do not have a complete view of the executable code in apps. To address this issue, we propose a new advance in the ambitious research direction of building a unified model of all code in Android apps. The JuCify approach presented in this paper is a significant step towards such a model, where we extract and merge call graphs of native code and bytecode to make the final model readily-usable by a common Android analysis framework: in our implementation, JuCify builds on the Soot internal intermediate representation. We performed empirical investigations to highlight how, without the unified model, a significant amount of Java methods called from the native code are "unreachable" in apps' call-graphs, both in goodware and malware. Using JuCify, we were able to enable static analyzers to reveal cases where malware relied on native code to hide invocation of payment library code or of other sensitive code in the Android framework. Additionally, JuCify's model enables state-of-the-art tools to achieve better precision and recall in detecting data leaks through native code. Finally, we show that by using JuCify we can find sensitive data leaks that pass through native code.

cs.SE↗

arXiv2021

Deep Marching Tetrahedra: a Hybrid Representation for High-Resolution 3D Shape Synthesis

We introduce DMTet, a deep 3D conditional generative model that can synthesize high-resolution 3D shapes using simple user guides such as coarse voxels. It marries the merits of implicit and explicit 3D representations by leveraging a novel hybrid 3D representation. Compared to the current implicit approaches, which are trained to regress the signed distance values, DMTet directly optimizes for the reconstructed surface, which enables us to synthesize finer geometric details with fewer artifacts. Unlike deep 3D generative models that directly generate explicit representations such as meshes, our model can synthesize shapes with arbitrary topology. The core of DMTet includes a deformable tetrahedral grid that encodes a discretized signed distance function and a differentiable marching tetrahedra layer that converts the implicit signed distance representation to the explicit surface mesh representation. This combination allows joint optimization of the surface geometry and topology as well as generation of the hierarchy of subdivisions using reconstruction and adversarial losses defined explicitly on the surface mesh. Our approach significantly outperforms existing work on conditional shape synthesis from coarse voxel inputs, trained on a dataset of complex 3D animal shapes. Project page: https://nv-tlabs.github.io/DMTet/.

cs.CV↗

arXiv2021

DIB-R++: Learning to Predict Lighting and Material with a Hybrid Differentiable Renderer

We consider the challenging problem of predicting intrinsic object properties from a single image by exploiting differentiable renderers. Many previous learning-based approaches for inverse graphics adopt rasterization-based renderers and assume naive lighting and material models, which often fail to account for non-Lambertian, specular reflections commonly observed in the wild. In this work, we propose DIBR++, a hybrid differentiable renderer which supports these photorealistic effects by combining rasterization and ray-tracing, taking the advantage of their respective strengths -- speed and realism. Our renderer incorporates environmental lighting and spatially-varying material models to efficiently approximate light transport, either through direct estimation or via spherical basis functions. Compared to more advanced physics-based differentiable renderers leveraging path tracing, DIBR++ is highly performant due to its compact and expressive shading model, which enables easy integration with learning frameworks for geometry, reflectance and lighting prediction from a single image without requiring any ground-truth. We experimentally demonstrate that our approach achieves superior material and lighting disentanglement on synthetic and real data compared to existing rasterization-based approaches and showcase several artistic applications including material editing and relighting.

cs.CV↗

arXiv2021

Hadronic decays of Higgs boson at NNLO matched with parton shower

We present predictions for hadronic decays of the Higgs boson at next-to-next-to-leading order (NNLO) in QCD matched with parton shower based on the POWHEG framework. Those include decays into bottom quarks with full bottom-quark mass dependence, light quarks, and gluons in the heavy top quark effective theory. Our calculations describe exclusive decays of the Higgs boson with leading logarithmic accuracy in the Sudakov region and next-to-leading order (NLO) accuracy matched with parton shower in the three-jet region, with normalizations fixed to the partial width at NNLO. We estimated remaining perturbative uncertainties taking typical event shape variables as an example and demonstrated the need of future improvements on both parton shower and matrix element calculations. The calculations can be used immediately in evaluations of the physics performances of detector designs for future Higgs factories.

hep-ph↗