arXiv subjects

Peng Sun

Publications and source records attributed to Peng Sun.

At least 181 records · Page 10Linked to original sources

arXiv2023

Local-to-Global Information Communication for Real-Time Semantic Segmentation Network Search

Neural Architecture Search (NAS) has shown great potentials in automatically designing neural network architectures for real-time semantic segmentation. Unlike previous works that utilize a simplified search space with cell-sharing way, we introduce a new search space where a lightweight model can be more effectively searched by replacing the cell-sharing manner with cell-independent one. Based on this, the communication of local to global information is achieved through two well-designed modules. For local information exchange, a graph convolutional network (GCN) guided module is seamlessly integrated as a communication deliver between cells. For global information aggregation, we propose a novel dense-connected fusion module (cell) which aggregates long-range multi-level features in the network automatically. In addition, a latency-oriented constraint is endowed into the search process to balance the accuracy and latency. We name the proposed framework as Local-to-Global Information Communication Network Search (LGCNet). Extensive experiments on Cityscapes and CamVid datasets demonstrate that LGCNet achieves the new state-of-the-art trade-off between accuracy and speed. In particular, on Cityscapes dataset, LGCNet achieves the new best performance of 74.0\% mIoU with the speed of 115.2 FPS on Titan Xp.

cs.CV↗

arXiv2023

Event-Based Fusion for Motion Deblurring with Cross-modal Attention

Traditional frame-based cameras inevitably suffer from motion blur due to long exposure times. As a kind of bio-inspired camera, the event camera records the intensity changes in an asynchronous way with high temporal resolution, providing valid image degradation information within the exposure time. In this paper, we rethink the eventbased image deblurring problem and unfold it into an end-to-end two-stage image restoration network. To effectively fuse event and image features, we design an event-image cross-modal attention module applied at multiple levels of our network, which allows to focus on relevant features from the event branch and filter out noise. We also introduce a novel symmetric cumulative event representation specifically for image deblurring as well as an event mask gated connection between the two stages of our network which helps avoid information loss. At the dataset level, to foster event-based motion deblurring and to facilitate evaluation on challenging real-world images, we introduce the Real Event Blur (REBlur) dataset, captured with an event camera in an illumination controlled optical laboratory. Our Event Fusion Network (EFNet) sets the new state of the art in motion deblurring, surpassing both the prior best-performing image-based method and all event-based methods with public implementations on the GoPro dataset (by up to 2.47dB) and on our REBlur dataset, even in extreme blurry conditions. The code and our REBlur dataset will be made publicly available.

cs.CV↗

arXiv2023

Adjoint optimization of polarization-splitting grating couplers

We have designed a polarization-splitting grating coupler (PSGC) in silicon-oninsulator (SOI) that has 1.2 dB peak loss in numerical simulations, which is the best simulated performance of PSGCs without a bottom reflector to the best of our knowledge. Adjoint method-based shape optimization enables us to explore complex geometries that are intractable with conventional design approaches. Physics-based process-independent knowledge of PSGCs is extracted from the adjoint optimization and can be transferred to other platforms with a minimum of effort.

physics.optics↗

arXiv2022

Lattice gauge ensembles and data management

The generation of ensembles of gauge configurations is a considerable expense. The preservation and curation of these ensembles constitutes a valuable shared resource for the lattice field theory community. The organizers of Lattice 2022 dedicated a parallel session to the presentation of gauge ensembles and their generation, plans for ensemble publication and data management/storage activities of different collaborations. A summary of the twelve contributions is presented here.

hep-lat↗

arXiv2022

Distance between various discretized fermion actions

We present the leading order mixed-action effect $Δ_{\rm mix}\equiv m_{π,{\rm vs}}^2-\frac{m_{π,{\rm vv}}^2+m_{π,{\rm ss}}^2}{2}$ using HISQ, clover or overlap valence fermion actions on gauge ensembles using various sea fermion actions across a widely-used lattice spacing range $a\in [0.04,0.19]$~fm. The results suggest that $Δ_{\rm mix}$ decreases as the fourth order of the lattice spacing on the gauge ensembles with dynamical chiral sea fermions, such as Domain wall or HISQ fermions. When a clover sea fermion action which has explicit chiral symmetry breaking is used in the ensemble, $Δ_{\rm mix}$ can be much larger regardless of the valence fermion action used.

hep-lat↗

arXiv2022

Probing the $Hγγ$ coupling via Higgs exclusive decay into quarkonia plus a photon at the HL-LHC

It is well know that the precise measurement of $H\toγγ$ can generate two degenerate parameter spaces of $Hγγ$ anomalous coupling. We propose to utilize the exclusive Higgs rare decays $H\toΥ(ns)+γ$ to break above mentioned degeneracy and further to constrain the $Hγγ$ anomalous coupling at the HL-LHC. We demonstrate that the branching ratios of $H\toΥ(ns)+γ$ can be significantly enhanced in the non-SM-like parameter space from $H\toγγ$ measurement, due to the destructive interference between the direct and indirect production of $Υ(ns)$ in the SM. By applying the NRQCD factorization formalism, we calculate the partial decay widths of $H\toΥ(ns)+γ$ at the NLO accuracy of $α_s$. We show that it is hopeful to break such degenerate parameter space at the HL-LHC if we can further highly suppress the background and enhance the signal efficiency compared to the ATLAS preliminary simulation.

hep-ph↗

arXiv2022

First observation of the hidden-charm pentaquarks on lattice

The s-wave scattering of $Σ_c \bar{D}$ and $Σ_c \bar{D}^*$ in the $I(J^P) = \frac{1}{2}(\frac{1}{2}^-)$ channel is calculated in lattice QCD using two ensembles with different volumes but the same lattice spacing $a\sim 0.08\mathrm{fm}$ and pion mass $M_π\sim 294\mathrm{MeV}$. The scattering amplitudes near threshold are obtained by Lüscher's finite volume method. We find bound state poles in both $Σ_c \bar{D}$ and $Σ_c \bar{D}^*$ channels, which are possibly related to the $P_c(4312)$ and $P_c(4440) / P_c(4457)$ pentaquarks observed in experiments. The binding energy is $6(2)(2)$MeV for $Σ_c \bar{D}$ and $7(3)(1)$MeV for $Σ_c \bar{D}^*$, where the first error is the statistical error and the second is the systematic error due to the lattice artifacts.

hep-lat↗

arXiv2022

Ergodic measures of intermediate entropies for dynamical systems with the approximate product property

For a dynamical system satisfying the approximate product property and asymptotically entropy expansiveness, we characterize a delicate structrue of the space of invariant measures: The ergodic measures of intermediate entropies and intermediate pressures are generic in certain subspaces. This proves a conjecture of Katok for a broad class of systems and extends a sequence of known results.

math.DS↗

arXiv2022

Controllability of a Class of Swarm Signaling Networks

In this paper, we propose closed-form analytical expressions to determine the minimum number of driver nodes that is needed to control a specific class of networks. We consider swarm signaling networks with regular out-degree distribution where a fraction $p$ of the links is unavailable. We further apply our method to networks with bi-modal out-degree distributions. Our approximations are validated through intensive simulations. Results show that our approximations have high accuracy when compared with simulation results for both types of out-degree distribution.

eess.SY↗

arXiv2022

Deep Learning Workload Scheduling in GPU Datacenters: Taxonomy, Challenges and Vision

Deep learning (DL) shows its prosperity in a wide variety of fields. The development of a DL model is a time-consuming and resource-intensive procedure. Hence, dedicated GPU accelerators have been collectively constructed into a GPU datacenter. An efficient scheduler design for such GPU datacenter is crucially important to reduce the operational cost and improve resource utilization. However, traditional approaches designed for big data or high performance computing workloads can not support DL workloads to fully utilize the GPU resources. Recently, substantial schedulers are proposed to tailor for DL workloads in GPU datacenters. This paper surveys existing research efforts for both training and inference workloads. We primarily present how existing schedulers facilitate the respective workloads from the scheduling objectives and resource consumption features. Finally, we prospect several promising future research directions. More detailed summary with the surveyed paper and code links can be found at our project website: https://github.com/S-Lab-System-Group/Awesome-DL-Scheduling-Papers

cs.DC↗

arXiv2022

Unramified Cohomology of Quadrics in Characteristic Two

Let $F$ be a field of characteristic 2 and let $X$ be a smooth projective quadric of dimension $\ge 1$ over $F$. We study the unramified cohomology groups with 2-primary torsion coefficients of $X$ in degrees 2 and 3. We determine completely the kernel and the cokernel of the natural map from the cohomology of $F$ to the unramified cohomology of $X$. This extends the results in characteristic different from 2 obtained by Kahn, Rost and Sujatha in the nineteen-nineties.

math.NT↗

arXiv2022

Wireless Image Transmission Using Deep Source Channel Coding With Attention Modules

Recent research on joint source channel coding (JSCC) for wireless communications has achieved great success owing to the employment of deep learning (DL). However, the existing work on DL based JSCC usually trains the designed network to operate under a specific signal-to-noise ratio (SNR) regime, without taking into account that the SNR level during the deployment stage may differ from that during the training stage. A number of networks are required to cover the scenario with a broad range of SNRs, which is computational inefficiency (in the training stage) and requires large storage. To overcome these drawbacks our paper proposes a novel method called Attention DL based JSCC (ADJSCC) that can successfully operate with different SNR levels during transmission. This design is inspired by the resource assignment strategy in traditional JSCC, which dynamically adjusts the compression ratio in source coding and the channel coding rate according to the channel SNR. This is achieved by resorting to attention mechanisms because these are able to allocate computing resources to more critical tasks. Instead of applying the resource allocation strategy in traditional JSCC, the ADJSCC uses the channel-wise soft attention to scaling features according to SNR conditions. We compare the ADJSCC method with the state-of-the-art DL based JSCC method through extensive experiments to demonstrate its adaptability, robustness and versatility. Compared with the existing methods, the proposed method takes less storage and is more robust in the presence of channel mismatch.

cs.IT↗

arXiv2022

Probing the $Zb\bar{b}$ anomalous couplings via exclusive $Z$ boson decay

We propose to utilize the exclusive $Z$-boson rare decays $Z\to Υ(ns)+γ$ to constrain the $Zb\bar{b}$ couplings at the HL-LHC and 100 TeV proton-proton collider. We demonstrate that the event yield of the proposed processes is sensitive to the axial-vector component of the $Zb\bar{b}$ coupling and can provide complementary information to the jet-charge weighted single-spin asymmetry measurement at the EIC and the $gg\to Zh$ production rate measurement at the LHC. By applying the NRQCD factorization formalism, we calculate the partial decay width of $Z\to Υ(ns)+γ$ to the NLO accuracy in strong interaction, which is found to agree with those obtained from the light-cone distribution amplitude approach. We show that the HL-LHC can break the degeneracy of the $Zb\bar{b}$ couplings, as implied by the precision electroweak data at LEP and SLC, if the signal efficiency can be improved by a factor of 1.7 from the present ATLAS analysis at the 13 TeV LHC with an integrated luminosity of $36.1~{\rm fb}^{-1}$.

hep-ph↗

arXiv2022

Nonperturbative Determination of Collins-Soper Kernel from Quasi Transverse-Momentum Dependent Wave Functions

In the framework of large-momentum effective theory at one-loop matching accuracy, we perform a lattice calculation of the Collins-Soper kernel which governs the rapidity evolution of transverse-momentum-dependent (TMD) distributions. We first obtain the quasi TMD wave functions at three different meson momenta on a lattice with valence clover quarks on a dynamical HISQ sea and lattice spacing $a=0.12$~fm from MILC, and renormalize the pertinent linear divergences using Wilson loops. Through one-loop matching to the light-cone wave functions, we determine the Collins-Soper kernel with transverse separation up to 0.6~fm. We study the systematic uncertainties from operator mixing and scale dependence, as well as the impact from higher power corrections. Our results potentially allow for a determination of the soft function and other transverse-momentum dependent quantities at one-loop accuracy.

hep-lat↗

arXiv2022

TrajGen: Generating Realistic and Diverse Trajectories with Reactive and Feasible Agent Behaviors for Autonomous Driving

Realistic and diverse simulation scenarios with reactive and feasible agent behaviors can be used for validation and verification of self-driving system performance without relying on expensive and time-consuming real-world testing. Existing simulators rely on heuristic-based behavior models for background vehicles, which cannot capture the complex interactive behaviors in real-world scenarios. To bridge the gap between simulation and the real world, we propose TrajGen, a two-stage trajectory generation framework, which can capture more realistic behaviors directly from human demonstration. In particular, TrajGen consists of the multi-modal trajectory prediction stage and the reinforcement learning based trajectory modification stage. In the first stage, we propose a novel auxiliary RouteLoss for the trajectory prediction model to generate multi-modal diverse trajectories in the drivable area. In the second stage, reinforcement learning is used to track the predicted trajectories while avoiding collisions, which can improve the feasibility of generated trajectories. In addition, we develop a data-driven simulator I-Sim that can be used to train reinforcement learning models in parallel based on naturalistic driving data. The vehicle model in I-Sim can guarantee that the generated trajectories by TrajGen satisfy vehicle kinematic constraints. Finally, we give comprehensive metrics to evaluate generated trajectories for simulation scenarios, which shows that TrajGen outperforms either trajectory prediction or inverse reinforcement learning in terms of fidelity, reactivity, feasibility, and diversity.

cs.RO↗

arXiv2022

Generative Adversarial Exploration for Reinforcement Learning

Exploration is crucial for training the optimal reinforcement learning (RL) policy, where the key is to discriminate whether a state visiting is novel. Most previous work focuses on designing heuristic rules or distance metrics to check whether a state is novel without considering such a discrimination process that can be learned. In this paper, we propose a novel method called generative adversarial exploration (GAEX) to encourage exploration in RL via introducing an intrinsic reward output from a generative adversarial network, where the generator provides fake samples of states that help discriminator identify those less frequently visited states. Thus the agent is encouraged to visit those states which the discriminator is less confident to judge as visited. GAEX is easy to implement and of high training efficiency. In our experiments, we apply GAEX into DQN and the DQN-GAEX algorithm achieves convincing performance on challenging exploration problems, including the game Venture, Montezuma's Revenge and Super Mario Bros, without further fine-tuning on complicate learning algorithms. To our knowledge, this is the first work to employ GAN in RL exploration problems.

cs.LG↗

arXiv2022

First Lattice QCD determination of semileptonic decays of charmed-strange baryons $Ξ_c$

While the standard model is the most successfully theory to describe all interactions and constituents in elementary particle physics, it has been constantly examined for over four decades. Weak decays of charm quarks can measure the coupling strength of quarks in different families and serve as an ideal probe for CP violation. As the lowest charm-strange baryons with three different flavors, $Ξ_c$ baryons (made of $csu$ or $csd$) have been extensively studied in experiments at the large hadron collider and in electron-positron collision. However the lack of reliable knowledge in theory becomes the unavoidable obstacle in the way. In this work, we use the state-of-the-art Lattice QCD techniques, and generate 2+1 clover fermion ensembles with two lattice spacings, $a=(0.108{\rm fm},0.080{\rm fm})$. We then present the first {\it ab-initio} lattice QCD determination of form factors governing $Ξ_{c}\to Ξ\ell^+ν_{\ell}$, analogous with the notable $β$-decay of nuclei. Our theoretical results for decay widths are consistent with and about two times more precise than the latest measurements by ALICE and Belle collaborations. Together with experimental measurements, we independently determine the quark-mixing matrix element $|V_{cs}|$, which is found in good agreement with other determinations.

hep-lat↗

arXiv2022

Distribution Amplitudes of $K^*$ and $ϕ$ at Physical Pion Mass from Lattice QCD

We present the first lattice QCD calculation of the distribution amplitudes of longitudinally and transversely polarized vector mesons $K^*$ and $ϕ$ using large momentum effective theory. We use the clover fermion action on three ensembles with 2+1+1 flavors of highly improved staggered quarks (HISQ) action, generated by MILC collaboration, at physical pion mass and \{0.06, 0.09, 0.12\} fm lattice spacings, and choose three different hadron momenta $P_z=\{1.29, 1.72, 2.15\}$ GeV. The resulting lattice matrix elements are nonperturbatively renormalized in a hybrid scheme proposed recently. An extrapolation to the continuum and infinite momentum limit is carried out. We find that while the longitudinal distribution amplitudes tend to be close to the asymptotic form, the transverse ones deviate rather significantly from the asymptotic form. Our final results provide crucial {\it ab initio} theory inputs for analyzing pertinent exclusive processes.

hep-lat↗