arXiv subjects

Tao Han

Publications and source records attributed to Tao Han.

At least 109 records · Page 6Linked to original sources

arXiv2024

Improving Global Weather and Ocean Wave Forecast with Large Artificial Intelligence Models

The rapid advancement of artificial intelligence technologies, particularly in recent years, has led to the emergence of several large parameter artificial intelligence weather forecast models. These models represent a significant breakthrough, overcoming the limitations of traditional numerical weather prediction models and indicating the emergence of profound potential tools for atmosphere-ocean forecasts. This study explores the evolution of these advanced artificial intelligence forecast models, and based on the identified commonalities, proposes the "Three Large Rules" to measure their development. We discuss the potential of artificial intelligence in revolutionizing numerical weather prediction, and briefly outlining the underlying reasons for its great potential. While acknowledging the high accuracy, computational efficiency, and ease of deployment of large artificial intelligence forecast models, we also emphasize the irreplaceable values of traditional numerical forecasts and explore the challenges in the future development of large-scale artificial intelligence atmosphere-ocean forecast models. We believe that the optimal future of atmosphere-ocean weather forecast lies in achieving a seamless integration of artificial intelligence and traditional numerical models. Such a synthesis is anticipated to offer a more advanced and reliable approach for improved atmosphere-ocean forecasts. Additionally, we illustrate how forecasters can adapt and leverage the advanced artificial intelligence model through an example by building a large artificial intelligence model for global ocean wave forecast.

cs.LG↗

arXiv2024

DiPrompT: Disentangled Prompt Tuning for Multiple Latent Domain Generalization in Federated Learning

Federated learning (FL) has emerged as a powerful paradigm for learning from decentralized data, and federated domain generalization further considers the test dataset (target domain) is absent from the decentralized training data (source domains). However, most existing FL methods assume that domain labels are provided during training, and their evaluation imposes explicit constraints on the number of domains, which must strictly match the number of clients. Because of the underutilization of numerous edge devices and additional cross-client domain annotations in the real world, such restrictions may be impractical and involve potential privacy leaks. In this paper, we propose an efficient and novel approach, called Disentangled Prompt Tuning (DiPrompT), a method that tackles the above restrictions by learning adaptive prompts for domain generalization in a distributed manner. Specifically, we first design two types of prompts, i.e., global prompt to capture general knowledge across all clients and domain prompts to capture domain-specific knowledge. They eliminate the restriction on the one-to-one mapping between source domains and local clients. Furthermore, a dynamic query metric is introduced to automatically search the suitable domain label for each sample, which includes two-substep text-image alignments based on prompt tuning without labor-intensive annotation. Extensive experiments on multiple datasets demonstrate that our DiPrompT achieves superior domain generalization performance over state-of-the-art FL methods when domain labels are not provided, and even outperforms many centralized learning methods using domain labels.

cs.LG↗

arXiv2024

Global Tropical Cyclone Intensity Forecasting with Multi-modal Multi-scale Causal Autoregressive Model

Accurate forecasting of Tropical cyclone (TC) intensity is crucial for formulating disaster risk reduction strategies. Current methods predominantly rely on limited spatiotemporal information from ERA5 data and neglect the causal relationships between these physical variables, failing to fully capture the spatial and temporal patterns required for intensity forecasting. To address this issue, we propose a Multi-modal multi-Scale Causal AutoRegressive model (MSCAR), which is the first model that combines causal relationships with large-scale multi-modal data for global TC intensity autoregressive forecasting. Furthermore, given the current absence of a TC dataset that offers a wide range of spatial variables, we present the Satellite and ERA5-based Tropical Cyclone Dataset (SETCD), which stands as the longest and most comprehensive global dataset related to TCs. Experiments on the dataset show that MSCAR outperforms the state-of-the-art methods, achieving maximum reductions in global and regional forecast errors of 9.52% and 6.74%, respectively. The code and dataset are publicly available at https://anonymous.4open.science/r/MSCAR.

physics.ao-ph↗

arXiv2024

Anomalous Tau Neutrino Appearance from Light Mediators in Short-Baseline Neutrino Experiments

We point out a new mechanism giving rise to anomalous tau neutrino appearance at the near detectors of beam-focused neutrino experiments, without extending the neutrino sector. The charged mesons ($π^\pm, K^\pm$) produced and focused in the target-horn system can decay to a (neutrino-philic) light mediator via the helicity-unsuppressed three-body decays. If such a mediator carries non-vanishing hadronic couplings, it can also be produced via the bremsstrahlung of the incident proton beam. The subsequent decay of the mediator to a tau neutrino pair results in tau neutrino detection at the near detectors, which is unexpected under the standard three-flavor neutrino oscillation paradigm. We argue that the signal flux from the charged meson decays can be significant enough to discover the light mediator signal at the on-axis liquid-argon near detector of the DUNE experiment, due to the focusing of charged mesons. In addition, we show that ICARUS-NuMI, an off-axis near detector of the NuMI beam, as well as DUNE, can observe a handful of tau neutrino events induced by beam-proton bremsstrahlung.

hep-ph↗

arXiv2024

FengWu-GHR: Learning the Kilometer-scale Medium-range Global Weather Forecasting

Kilometer-scale modeling of global atmosphere dynamics enables fine-grained weather forecasting and decreases the risk of disastrous weather and climate activity. Therefore, building a kilometer-scale global forecast model is a persistent pursuit in the meteorology domain. Active international efforts have been made in past decades to improve the spatial resolution of numerical weather models. Nonetheless, developing the higher resolution numerical model remains a long-standing challenge due to the substantial consumption of computational resources. Recent advances in data-driven global weather forecasting models utilize reanalysis data for model training and have demonstrated comparable or even higher forecasting skills than numerical models. However, they are all limited by the resolution of reanalysis data and incapable of generating higher-resolution forecasts. This work presents FengWu-GHR, the first data-driven global weather forecasting model running at the 0.09$^{\circ}$ horizontal resolution. FengWu-GHR introduces a novel approach that opens the door for operating ML-based high-resolution forecasts by inheriting prior knowledge from a pretrained low-resolution model. The hindcast of weather prediction in 2022 indicates that FengWu-GHR is superior to the IFS-HRES. Furthermore, evaluations on station observations and case studies of extreme events support the competitive operational forecasting skill of FengWu-GHR at the high resolution.

cs.LG↗

arXiv2023

Searching for Heavy Leptophilic $Z'$: from Lepton Colliders to Gravitational Waves

We study the phenomenology of leptophilic $Z'$ gauge bosons at the future high-energy $e^+e^-$ and $μ^+μ^-$ colliders, as well as at the gravitational wave observatories. The leptophilic $Z'$ model, although well-motivated, remains largely unconstrained from current low-energy and collider searches for $Z'$ masses above ${\cal O}(100~{\rm GeV})$, thus providing a unique opportunity for future lepton colliders. Taking $U(1)_{L_α-L_β}~(α,β=e,μ,τ)$ models as concrete examples, we show that future $e^+e^-$ and $μ^+μ^-$ colliders with multi-TeV center-of-mass energies provide unprecedented sensitivity to heavy leptophilic $Z'$ bosons. Moreover, if these $U(1)$ models are classically scale-invariant, the phase transition at the $U(1)$ symmetry-breaking scale tends to be strongly first-order with ultra-supercooling, and leads to observable stochastic gravitational wave signatures. We find that the future sensitivity of gravitational wave observatories, such as advanced LIGO-VIRGO and Cosmic Explorer, can be complementary to the collider experiments, probing higher $Z'$ masses up to ${\cal O}(10^4~{\rm TeV})$, while being consistent with naturalness and perturbativity considerations.

hep-ph↗

arXiv2023

Report of the 2021 U.S. Community Study on the Future of Particle Physics (Snowmass 2021) Summary Chapter

The 2021-22 High-Energy Physics Community Planning Exercise (a.k.a. ``Snowmass 2021'') was organized by the Division of Particles and Fields of the American Physical Society. Snowmass 2021 was a scientific study that provided an opportunity for the entire U.S. particle physics community, along with its international partners, to identify the most important scientific questions in High Energy Physics for the following decade, with an eye to the decade after that, and the experiments, facilities, infrastructure, and R&D needed to pursue them. This Snowmass summary report synthesizes the lessons learned and the main conclusions of the Community Planning Exercise as a whole and presents a community-informed synopsis of U.S. particle physics at the beginning of 2023. This document, along with the Snowmass reports from the various subfields, will provide input to the 2023 Particle Physics Project Prioritization Panel (P5) subpanel of the U.S. High-Energy Physics Advisory Panel (HEPAP), and will help to guide and inform the activity of the U.S. particle physics community during the next decade and beyond.

hep-ex↗

arXiv2023

Traffic Sign Interpretation in Real Road Scene

Most existing traffic sign-related works are dedicated to detecting and recognizing part of traffic signs individually, which fails to analyze the global semantic logic among signs and may convey inaccurate traffic instruction. Following the above issues, we propose a traffic sign interpretation (TSI) task, which aims to interpret global semantic interrelated traffic signs (e.g.,~driving instruction-related texts, symbols, and guide panels) into a natural language for providing accurate instruction support to autonomous or assistant driving. Meanwhile, we design a multi-task learning architecture for TSI, which is responsible for detecting and recognizing various traffic signs and interpreting them into a natural language like a human. Furthermore, the absence of a public TSI available dataset prompts us to build a traffic sign interpretation dataset, namely TSI-CN. The dataset consists of real road scene images, which are captured from the highway and the urban way in China from a driver's perspective. It contains rich location labels of texts, symbols, and guide panels, and the corresponding natural language description labels. Experiments on TSI-CN demonstrate that the TSI task is achievable and the TSI architecture can interpret traffic signs from scenes successfully even if there is a complex semantic logic among signs. The TSI-CN dataset and the source code of the TSI architecture will be publicly available after the revision process.

cs.CV↗

arXiv2023

Quantum Entanglement and Bell Inequality Violation in Semi-Leptonic Top Decays

Quantum entanglement is a fundamental property of quantum mechanics. Recently, studies have explored entanglement in the $t\bar{t}$ system at the Large Hadron Collider (LHC) when both the top quark and anti-top quark decay leptonically. Entanglement is detected via correlations between the polarizations of the top and anti-top and these polarizations are measured through the angles of the decay products of the top and anti-top. In this work, we propose searching for evidence of quantum entanglement in the semi-leptonic decay channel where the final state includes one lepton, one neutrino, two $b$-flavor tagged jets, and two light jets from the $W$ decay. We find that this channel is both easier to reconstruct and has a larger effective quantity of data than the fully leptonic channel. As a result, the semi-leptonic channel is $60\%$ more sensitive to quantum entanglement and a factor of 3 more sensitive to Bell inequality violation, compared to the leptonic channel. In $139~{\rm fb}^{-1}$ ($3~{\rm ab}^{-1}$) of data at the LHC (HL-LHC), it should be feasible to measure entanglement at a precision of $\lesssim 3\%\ (0.7\%)$. Detecting Bell inequality violation, on the other hand, is more challenging. With $300~{\rm fb}^{-1}$ ($3~{\rm ab}^{-1}$) of integrated luminosity at the LHC Run-3 (HL-LHC), we expect a sensitivity of $1.3σ$ ($4.1 σ$). In our study, we utilize a realistic parametric fitting procedure to optimally recover the true angular distributions from detector effects. Compared to unfolding this procedure yields more stable results.

hep-ph↗

arXiv2023

STEERER: Resolving Scale Variations for Counting and Localization via Selective Inheritance Learning

Scale variation is a deep-rooted problem in object counting, which has not been effectively addressed by existing scale-aware algorithms. An important factor is that they typically involve cooperative learning across multi-resolutions, which could be suboptimal for learning the most discriminative features from each scale. In this paper, we propose a novel method termed STEERER (\textbf{S}elec\textbf{T}iv\textbf{E} inh\textbf{ER}itance l\textbf{E}a\textbf{R}ning) that addresses the issue of scale variations in object counting. STEERER selects the most suitable scale for patch objects to boost feature extraction and only inherits discriminative features from lower to higher resolution progressively. The main insights of STEERER are a dedicated Feature Selection and Inheritance Adaptor (FSIA), which selectively forwards scale-customized features at each scale, and a Masked Selection and Inheritance Loss (MSIL) that helps to achieve high-quality density maps across all scales. Our experimental results on nine datasets with counting and localization tasks demonstrate the unprecedented scale generalization ability of STEERER. Code is available at \url{https://github.com/taohan10200/STEERER}.

cs.CV↗

arXiv2023

A Guide to Diagnosing Colored Resonances at Hadron Colliders

We present a comprehensive study on how to distinguish the properties of heavy dijet resonances at hadron colliders. A variety of spins, chiral couplings, charges, and QCD color representations are considered. Distinguishing the different color representations is particularly difficult at hadron colliders. To determine the QCD color structure, we consider a third jet radiated in a resonant dijet event. We show that the relative rates of three-jet versus two-jet processes are sensitive to the color representation of the resonance. We also show analytically that the antennae radiation pattern of soft radiation depends on the color structure of dijet events and develops an observable that is sensitive to the antennae patterns. Finally, we exploit a Convolutional Neural Network with Machine Learning techniques to differentiate the radiation patterns from different colored resonances and find encouraging results to discriminate them. We demonstrate our results numerically at a 14 TeV LHC, and the methodology presented here should be applicable to other future hadron colliders.

hep-ph↗

arXiv2023

Random Utterance Concatenation Based Data Augmentation for Improving Short-video Speech Recognition

One of limitations in end-to-end automatic speech recognition (ASR) framework is its performance would be compromised if train-test utterance lengths are mismatched. In this paper, we propose an on-the-fly random utterance concatenation (RUC) based data augmentation method to alleviate train-test utterance length mismatch issue for short-video ASR task. Specifically, we are motivated by observations that our human-transcribed training utterances tend to be much shorter for short-video spontaneous speech (~3 seconds on average), while our test utterance generated from voice activity detection front-end is much longer (~10 seconds on average). Such a mismatch can lead to suboptimal performance. Empirically, it's observed the proposed RUC method significantly improves long utterance recognition without performance drop on short one. Overall, it achieves 5.72% word error rate reduction on average for 15 languages and improved robustness to various utterance length.

eess.AS↗

arXiv2023

Prospects for constraining interacting dark energy models from gravitational wave and gamma ray burst joint observation

With the measurement of the electromagnetic (EM) counterpart, a gravitational wave (GW) event could be treated as a standard siren. As a novel cosmological probe, GW standard sirens will bring significant implications for cosmology. In this paper, by considering the coincident detections of GW and associated $γ$ ray burst (GRB), we find that only about 400 GW bright standard sirens from binary neutron star mergers could be detected in a 10-year observation of the Einstein Telescope and the THESEUS satellite mission. Based on this mock sample, we investigate the implications of GW standard sirens on the interaction between dark energy and dark matter. In our analysis, four viable interacting dark energy (IDE) models, with interaction forms $Q=3βH ρ_{\mathrm{de}}$ and $Q=3 βH ρ_{\mathrm{c}}$, are considered. Compared with the traditional EM observational data such as CMB, BAO, and SN Ia, the combination of both GW and EM observations could effectively break the degeneracies between different cosmological parameters and provide more stringent cosmological fits. We find that the GW data could play a more important role for determining the interaction in the models with $Q=3 βH ρ_{\mathrm{c}}$, compared with the models with $Q=3βH ρ_{\mathrm{de}}$. We also show that constraining IDE models with mock GW data based on different fiducial $H_0$ values yield different results, indicating that accurate determination of $H_0$ is significant for exploring the interaction between dark energy and dark matter.

astro-ph.CO↗

arXiv2023

DRPT: Disentangled and Recurrent Prompt Tuning for Compositional Zero-Shot Learning

Compositional Zero-shot Learning (CZSL) aims to recognize novel concepts composed of known knowledge without training samples. Standard CZSL either identifies visual primitives or enhances unseen composed entities, and as a result, entanglement between state and object primitives cannot be fully utilized. Admittedly, vision-language models (VLMs) could naturally cope with CZSL through tuning prompts, while uneven entanglement leads prompts to be dragged into local optimum. In this paper, we take a further step to introduce a novel Disentangled and Recurrent Prompt Tuning framework termed DRPT to better tap the potential of VLMs in CZSL. Specifically, the state and object primitives are deemed as learnable tokens of vocabulary embedded in prompts and tuned on seen compositions. Instead of jointly tuning state and object, we devise a disentangled and recurrent tuning strategy to suppress the traction force caused by entanglement and gradually optimize the token parameters, leading to a better prompt space. Notably, we develop a progressive fine-tuning procedure that allows for incremental updates to the prompts, optimizing the object first, then the state, and vice versa. Meanwhile, the optimization of state and object is independent, thus clearer features can be learned to further alleviate the issue of entangling misleading optimization. Moreover, we quantify and analyze the entanglement in CZSL and supplement entanglement rebalancing optimization schemes. DRPT surpasses representative state-of-the-art methods on extensive benchmark datasets, demonstrating superiority in both accuracy and efficiency.

cs.CV↗

arXiv2023

Higgs to $b\bar{b}$ from Vector Boson Fusion for High-Scale Physics

Vector boson fusion is arguably the most direct collider probe of electroweak symmetry breaking. Typically, the signature includes two forward/backward jets with low transverse momenta with a scale that is set by the mass of the vector boson. For this reason, an upper cut is used when searching for vector boson fusion processes in the Standard Model. Alternatively, the upper cut on the forward jets can be removed and the high-momentum exchange region of vector boson fusion can be studied. This phase space region has sensitivity to new physics via higher dimensional operators and form factors. In this work, we study the high-momentum region of the vector boson fusion channel where the Higgs decays to $b\bar{b}$. We show that, depending on the form of new physics, the limits on the new physics scale range from 0.5 TeV to 1.8 TeV.

hep-ph↗

arXiv2023

FengWu: Pushing the Skillful Global Medium-range Weather Forecast beyond 10 Days Lead

We present FengWu, an advanced data-driven global medium-range weather forecast system based on Artificial Intelligence (AI). Different from existing data-driven weather forecast methods, FengWu solves the medium-range forecast problem from a multi-modal and multi-task perspective. Specifically, a deep learning architecture equipped with model-specific encoder-decoders and cross-modal fusion Transformer is elaborately designed, which is learned under the supervision of an uncertainty loss to balance the optimization of different predictors in a region-adaptive manner. Besides this, a replay buffer mechanism is introduced to improve medium-range forecast performance. With 39-year data training based on the ERA5 reanalysis, FengWu is able to accurately reproduce the atmospheric dynamics and predict the future land and atmosphere states at 37 vertical levels on a 0.25° latitude-longitude resolution. Hindcasts of 6-hourly weather in 2018 based on ERA5 demonstrate that FengWu performs better than GraphCast in predicting 80\% of the 880 reported predictands, e.g., reducing the root mean square error (RMSE) of 10-day lead global z500 prediction from 733 to 651 $m^{2}/s^2$. In addition, the inference cost of each iteration is merely 600ms on NVIDIA Tesla A100 hardware. The results suggest that FengWu can significantly improve the forecast skill and extend the skillful global medium-range weather forecast out to 10.75 days lead (with ACC of z500 > 0.6) for the first time.

cs.AI↗

arXiv2023

Heavy Neutral Lepton Searches at the Electron-Ion Collider: A Snowmass Whitepaper

In this whitepaper, we consider the model of heavy neutral leptons (HNLs) as an example to explore the potential of new physics searches at the Electron-Ion Collider (EIC). We propose two broad categories of search strategies depending on the HNL lifetime: direct searches for the prompt decay of HNLs with a short lifetime and displaced vertex searches for long-lived ones. After identifying the most promising signals and the corresponding backgrounds, we perform a detailed simulation to estimate the sensitivity of the EIC to HNLs, accounting for detector thresholds, resolutions, and geometric acceptance. We derive projections for the EIC reach to the HNL squared mixing angle as a function of the HNL mass under the electron flavor mixing dominance hypothesis. Our findings indicate that the EIC can provide comparable sensitivity to the existing constraints for the prompt searches, while the displaced vertex searches can cover substantial new ground for HNLs in the 1-10 GeV mass range. Our proposed strategies are generally applicable to other new physics scenarios as well and motivate additional phenomenological exploration and dedicated future searches at the EIC.

hep-ph↗

arXiv2023

RoNet: Toward Robust Neural Assisted Mobile Network Configuration

Automating configuration is the key path to achieving zero-touch network management in ever-complicating mobile networks. Deep learning techniques show great potential to automatically learn and tackle high-dimensional networking problems. The vulnerability of deep learning to deviated input space, however, raises increasing deployment concerns under unpredictable variabilities and simulation-to-reality discrepancy in real-world networks. In this paper, we propose a novel RoNet framework to improve the robustness of neural-assisted configuration policies. We formulate the network configuration problem to maximize performance efficiency when serving diverse user applications. We design three integrated stages with novel normal training, learn-to-attack, and robust defense method for balancing the robustness and performance of policies. We evaluate RoNet via the NS-3 simulator extensively and the simulation results show that RoNet outperforms existing solutions in terms of robustness, adaptability, and scalability.

cs.NI↗