arXiv subjects

Tong Li

Publications and source records attributed to Tong Li.

At least 181 records · Page 10Linked to original sources

arXiv2022

Mid-level Representation Enhancement and Graph Embedded Uncertainty Suppressing for Facial Expression Recognition

Facial expression is an essential factor in conveying human emotional states and intentions. Although remarkable advancement has been made in facial expression recognition (FER) task, challenges due to large variations of expression patterns and unavoidable data uncertainties still remain. In this paper, we propose mid-level representation enhancement (MRE) and graph embedded uncertainty suppressing (GUS) addressing these issues. On one hand, MRE is introduced to avoid expression representation learning being dominated by a limited number of highly discriminative patterns. On the other hand, GUS is introduced to suppress the feature ambiguity in the representation space. The proposed method not only has stronger generalization capability to handle different variations of expression patterns but also more robustness to capture expression representations. Experimental evaluation on Aff-Wild2 have verified the effectiveness of the proposed method.

cs.CV↗

arXiv2022

Enhancing Collaborative Filtering Recommender with Prompt-Based Sentiment Analysis

Collaborative Filtering(CF) recommender is a crucial application in the online market and ecommerce. However, CF recommender has been proven to suffer from persistent problems related to sparsity of the user rating that will further lead to a cold-start issue. Existing methods address the data sparsity issue by applying token-level sentiment analysis that translate text review into sentiment scores as a complement of the user rating. In this paper, we attempt to optimize the sentiment analysis with advanced NLP models including BERT and RoBERTa, and experiment on whether the CF recommender has been further enhanced. We build the recommenders on the Amazon US Reviews dataset, and tune the pretrained BERT and RoBERTa with the traditional fine-tuned paradigm as well as the new prompt-based learning paradigm. Experimental result shows that the recommender enhanced with the sentiment ratings predicted by the fine-tuned RoBERTa has the best performance, and achieved 30.7% overall gain by comparing MAP, NDCG and precision at K to the baseline recommender. Prompt-based learning paradigm, although superior to traditional fine-tune paradigm in pure sentiment analysis, fail to further improve the CF recommender.

cs.IR↗

arXiv2022

Medical Scientific Table-to-Text Generation with Human-in-the-Loop under the Data Sparsity Constraint

Structured (tabular) data in the preclinical and clinical domains contains valuable information about individuals and an efficient table-to-text summarization system can drastically reduce manual efforts to condense this data into reports. However, in practice, the problem is heavily impeded by the data paucity, data sparsity and inability of the state-of-the-art natural language generation models (including T5, PEGASUS and GPT-Neo) to produce accurate and reliable outputs. In this paper, we propose a novel table-to-text approach and tackle these problems with a novel two-step architecture which is enhanced by auto-correction, copy mechanism and synthetic data augmentation. The study shows that the proposed approach selects salient biomedical entities and values from structured data with improved precision (up to 0.13 absolute increase) of copying the tabular values to generate coherent and accurate text for assay validation reports and toxicology reports. Moreover, we also demonstrate a light-weight adaptation of the proposed system to new datasets by fine-tuning with as little as 40\% training examples. The outputs of our model are validated by human experts in the Human-in-the-Loop scenario.

cs.CL↗

arXiv2022

A Scalable Workflow to Build Machine Learning Classifiers with Clinician-in-the-Loop to Identify Patients in Specific Diseases

Clinicians may rely on medical coding systems such as International Classification of Diseases (ICD) to identify patients with diseases from Electronic Health Records (EHRs). However, due to the lack of detail and specificity as well as a probability of miscoding, recent studies suggest the ICD codes often cannot characterise patients accurately for specific diseases in real clinical practice, and as a result, using them to find patients for studies or trials can result in high failure rates and missing out on uncoded patients. Manual inspection of all patients at scale is not feasible as it is highly costly and slow. This paper proposes a scalable workflow which leverages both structured data and unstructured textual notes from EHRs with techniques including NLP, AutoML and Clinician-in-the-Loop mechanism to build machine learning classifiers to identify patients at scale with given diseases, especially those who might currently be miscoded or missed by ICD codes. Case studies in the MIMIC-III dataset were conducted where the proposed workflow demonstrates a higher classification performance in terms of F1 scores compared to simply using ICD codes on gold testing subset to identify patients with Ovarian Cancer (0.901 vs 0.814), Lung Cancer (0.859 vs 0.828), Cancer Cachexia (0.862 vs 0.650), and Lupus Nephritis (0.959 vs 0.855). Also, the proposed workflow that leverages unstructured notes consistently outperforms the baseline that uses structured data only with an increase of F1 (Ovarian Cancer 0.901 vs 0.719, Lung Cancer 0.859 vs 0.787, Cancer Cachexia 0.862 vs 0.838 and Lupus Nephritis 0.959 vs 0.785). Experiments on the large testing set also demonstrate the proposed workflow can find more patients who are miscoded or missed by ICD codes. Moreover, interpretability studies are also conducted to clinically validate the top impact features of the classifiers.

cs.CL↗

arXiv2022

Interpretation of XENON1T excess with MeV boosted dark matter

The XENON1T excess of keV electron recoil events may be induced by the scattering of electrons and long-lived particles with MeV mass and high-speed. We consider a tangible model composed of two scalar MeV dark matter (DM) particles $S_A$ and $S_B$ to interpret the XENON1T keV excess via boosted $S_B$. A small mass splitting $m_{S_A}-m_{S_B}>0$ is introduced and the boosted $S_B$ can be produced by the dark annihilation process of $S_A S_A^\dagger \to ϕ\to S_B S_B^\dagger$ via a resonant scalar $ϕ$. The $S_B-$electron scattering is intermediated by a vector boson $X$. Although the constraints from BBN, CMB and low-energy experiments set the $X-$mediated $S_B-$electron scattering cross section to be $\lesssim 10^{-35} \mathrm{cm}^2$, the MeV scale DM with a resonance enhanced dark annihilation today can still provide enough boosted $S_B$ and induce the XENON1T keV excess. The relic density of $S_B$ is significantly reduced by the $s$-wave process of $S_B S_B^\dagger \to X X$ which is allowed by the constraints from CMB and 21-cm absorption. A very small relic fraction of $S_B$ is compatible with the stringent bounds on un-boosted $S_B$-electron scattering in DM direct detection and the $S_A$-electron scattering is also allowed.

hep-ph↗

arXiv2022

Dark magnetic dipole property in fermionic absorption by nucleus and electrons

The fermionic dark matter (DM) absorption by nucleus or electron targets provides a distinctive signal to search for sub-GeV DM. We consider a Dirac fermion DM charged under a dark gauge group and with the dark magnetic dipole operator. The DM field mixes with right-handed neutrino and interacts with the ordinary electromagnetic charge current via the kinetic mixing term of gauge fields. As a result, the incoming DM is absorbed and converted into neutrino in final state through the dipole-charge interaction. For the DM absorption by nucleus, the recoil energy spectrum exhibit a peak at $m_χ^2/2m_N$ for each isotope in the target. XENON1T can probe the DM mass above 27 MeV and the projected constraint on the inelastic DM-nucleon cross section becomes $10^{-49}$ cm$^2$. CRESSTIII with lower energy threshold would be sensitive to the DM mass above 2 MeV. We also check that the contribution from the nuclear magnetic dipole is negligible for $^{131}{\rm Xe}$ target. The absorption of DM by bound electron target induces ionization signal and is sensitive to sub-MeV DM mass. The involvement of the ionization form factor spreads out the localized recoil energy. We show the future prospect for the constraint on the magnetic dipole coupling from the electron ionization of $^{131}{\rm Xe}$.

hep-ph↗

arXiv2022

Three-dimensional Skyrme Hartree-Fock-Bogoliubov solver in coordinate-space representation

The coordinate-space representation of the Hartree-Fock-Bogoliubov theory is the method of choice to study weakly bound nuclei whose properties are affected by the quasiparticle continuum space. To describe such systems, we developed a three-dimensional Skyrme-Hartree-Fock-Bogoliubov solver HFBFFT based on the existing, highly optimized and parallelized Skyrme-Hartree-Fock code Sky3D. The code does not impose any self-consistent spatial symmetries such as mirror inversions or parity. The underlying equations are solved in HFBFFT directly in the canonical basis using the fast Fourier transform. To remedy the problems with pairing collapse, we implemented the soft energy cutoff and pairing annealing. The convergence of HFB solutions was improved by a sub-iteration method. The Hermiticity violation of differential operators brought by Fourier-transform-based differentiation has also been solved. The accuracy and performance of HFBFFT were tested by benchmarking it against other HFB codes, both spherical and deformed, for a set of nuclei, both well-bound and weakly-bound.

nucl-th↗

arXiv2022

Measurement of Higgs boson self-couplings through $2\rightarrow 3$ vector bosons scattering in future muon colliders

The $2\rightarrow 3$ scattering of longitudinal vector bosons (VBS) has been proven to be a complementary channel to measure the Higgs self-couplings in the Standard Model (SM) and the SM effective field theory (SMEFT) at colliders. We perform the first comprehensive study of all main $2\rightarrow 3$ VBS processes at high-energy muon colliders, especially including background analysis. The main contributing channels turn out to be the scattering of $W^+W^-\rightarrow W^+W^-h$, $ZZh$ and $hhh$. We obtain the constraints on $c_6$ and $c_{Φ_1}$ which are the Wilson coefficients of the dimension-6 operators relevant for Higgs self-couplings in the SMEFT. With the center-of-mass energies of 10 TeV and 30 TeV, we find the expected sensitivity to the coefficients $c_6/Λ^2$ and $c_{Φ_1}/Λ^2$ can reach the level of 0.01 TeV$^{-2}$. The tightest constraints come from the $hhh$ channel, while the constraints from $WWh$ are also comparable. Our results crucially depend on selecting the longitudinal polarizations for the final $W$ and $Z$ bosons. We then study how the sensitivities change by varying the efficiency of tagging longitudinal polarizations, and find that the significance remains consistently high.

hep-ph↗

arXiv2022

Axion-Like Particles at High Energy Muon Colliders -- A White paper for Snowmass 2021

We study the discovery potential for heavy axion-like particles (ALPs) and the perspectives for determining their coupling properties at a muon collider. Focusing on their couplings to the Standard Model (SM) gauge bosons $γ, Z, W^\pm$, we show that a high-energy muon collider can substantially extend the mass coverage, essentially reaching the kinematic limit of the collider energy. The unique kinematics allow for non-ambiguous determination of the individual coupling strengths. The associated production via $μ^+μ^-$ annihilation and the VBF processes with the tagged outgoing muons can be utilized to verify the CP property of the ALPs. We illustrate our results for a muon collider running at 3 TeV and 10 TeV.

hep-ph↗

arXiv2021

A new Direct Detection Strategy for the Cosmic Neutrino Background

The direct detection of cosmic neutrino background (CNB) has been a longstanding challenge in particle physics, due to its low number density and tiny neutrino masses. In this work, we consider the spectrum of the CNB boosted by cosmic rays via the neutrino self-interaction, and calculate the event rate of the boosted CNB-plasmon scattering in term of the dielectric response, which accounts for in-medium screening effect of a condensed matter target. This can be taken as the new direct detection strategy for the CNB in complementary to the traditional one, which captures the CNB on a $β$-unstable nucleus. Our result shows that one can either see the event of the CNB for the exposure of per kg$\cdot$year, or puts a strong constraint on the neutrino self-interaction. We further explore the background induced by the sub-MeV dark matter and the boosted super-light dark matter.

hep-ph↗

arXiv2021

Loop effect with vector mediator in the coherent neutrino-nucleus scattering

The observation of the coherent elastic neutrino-nucleus scattering (CE$ν$NS) provides us opportunities to explore a wide class of new physics. In the Standard Model (SM), the CE$ν$NS process arises from the vector and axial-vector neutral currents through the exchange of $Z$ boson and the axial-vector current contribution turns out to be subdominant. It is thus natural to consider the extra contributions to CE$ν$NS from more generic new physics beyond the SM with (axial-)vector interactions associated with a new vector mediator $Z'$. Besides the ordinary CE$ν$NS, the active neutrinos can convert into a new exotic fermion $χ$ through the process $νN\to χN$ mediated by $Z'$ without violating the coherence. It would be interesting to consider the implication of this conversion for the new fermion sector beyond the SM. We consider the framework of a simplified neutrino model in which a new Dirac fermion $χ$ interacts with active neutrinos and a leptophobic vector mediator $Z'$. We evaluate both the tree-level and loop-level contributions to the CE$ν$NS and in particular the loop diagrams produce active neutrino elastic scattering process $νN\to νN$ with the fermion $χ$ inside the loops. When the interaction between $Z'$ and the SM quarks is vector type and axial-vector type, the CE$ν$NS processes are respectively dominated by the tree-level and loop-level contributions. We investigate the constraints on the model parameters by fitting to the COHERENT data, assuming a wide range of $m_χ$. The parameter space with $m_χ$ larger than the maximal energy of incoming neutrinos can be constrained by including the loop-level contribution. More importantly, the inclusion of loop diagrams can place constraint on axial-vector interaction whose tree-level process is absent in the coherent neutrino-nucleus scattering.

hep-ph↗

arXiv2021

Self-Supervised Detection of Contextual Synonyms in a Multi-Class Setting: Phenotype Annotation Use Case

Contextualised word embeddings is a powerful tool to detect contextual synonyms. However, most of the current state-of-the-art (SOTA) deep learning concept extraction methods remain supervised and underexploit the potential of the context. In this paper, we propose a self-supervised pre-training approach which is able to detect contextual synonyms of concepts being training on the data created by shallow matching. We apply our methodology in the sparse multi-class setting (over 15,000 concepts) to extract phenotype information from electronic health records. We further investigate data augmentation techniques to address the problem of the class sparsity. Our approach achieves a new SOTA for the unsupervised phenotype concept annotation on clinical text on F1 and Recall outperforming the previous SOTA with a gain of up to 4.5 and 4.0 absolute points, respectively. After fine-tuning with as little as 20\% of the labelled data, we also outperform BioBERT and ClinicalBERT. The extrinsic evaluation on three ICU benchmarks also shows the benefit of using the phenotypes annotated by our model as features.

cs.CL↗

arXiv2021

The implication of $J/ψ\to (γ+ ){\rm invisible}$ for the effective field theories of neutrino and dark matter

We study the implication of $J/ψ$ decay into invisible particles for light sterile neutrino and sub-GeV dark matter (DM). The low-energy effective field theories (EFTs) are used for the description of general neutrino interactions and the Dirac fermion DM coupled to charm quark. For $J/ψ\to γ+{\rm invisible}$, we perform the likelihood fits for the individual neutrino and DM operators with distinct Lorentz structures and photon spectra. The limits on the decay branching fractions are obtained for different neutrino or DM scenarios and then converted to the lower bounds on the new energy scales. The most stringent bounds on the energy scale in neutrino and DM EFTs are 12.8 GeV and 11.6 GeV, respectively. The purely invisible decay $J/ψ\to {\rm invisible}$ provides complementary constraints on the effective operators. The relevant bound on the energy scale is above 100 GeV for the dipole operators. We also evaluate the limit on the DM-nucleon scattering cross section converted from $J/ψ$ data. The data of $J/ψ$ invisible decays are sensitive to the light DM mass range where the DM direct detection experiments cannot probe yet. The future Super Tau Charm Factory after one year run can push the limits down by two orders of magnitude.

hep-ph↗

arXiv2021

Charged lepton flavor violation in light of the muon magnetic moment anomaly and colliders

Any observation of charged lepton flavor violation (CLFV) implies the existence of new physics beyond the SM in charged lepton sector. CLFV interactions may also contribute to the muon magnetic moment and explain the discrepancy between the SM prediction and the recent muon $g-2$ precision measurement at Fermilab. We consider the most general SM gauge invariant Lagrangian of $ΔL=0$ bileptons with CLFV couplings and investigate the interplay of low-energy precision experiments and colliders in light of the muon magnetic moment anomaly. We go beyond previous work by demonstrating the sensitivity of the LHC, the MACE experiment, a proposed muonium-antimuonium conversion experiment, and a muon collider. Currently-available LHC data is already able to probe unexplored parameter space via the CLFV process $pp\toγ^*/Z^*\to \ell_1^\pm \ell_1^\pm \ell_2^\mp \ell_2^\mp$.

hep-ph↗

arXiv2021

Connecting Primordial Black Hole to boosted sub-GeV Dark Matter through neutrino

The explorations of alternative dark matter (DM) candidates beyond WIMP motivated primordial black holes (PBHs) or sub-GeV DM particle in the Milky Way. Neutrinos from PBH evaporation at the present times play as a novel medium boosting sub-GeV DM and leaving signatures in the terrestrial experiments. We explore the boosted DM by the neutrino flux from PBH evaporation (PBH$ν$BDM) so as to connect the macroscopic PBHs to sub-GeV DM particle. We consider this PBH$ν$BDM scenario to interpret the XENON1T keV excess. The projected bounds on the sub-GeV DM-electron scattering cross section and the fraction of DM composed of PBHs $f_{\rm PBH}$ are imposed for future experiments.

hep-ph↗

arXiv2021

Self-Adaptive Transfer Learning for Multicenter Glaucoma Classification in Fundus Retina Images

The early diagnosis and screening of glaucoma are important for patients to receive treatment in time and maintain eyesight. Nowadays, deep learning (DL) based models have been successfully used for computer-aided diagnosis (CAD) of glaucoma from retina fundus images. However, a DL model pre-trained using a dataset from one hospital center may have poor performance on a dataset from another new hospital center and therefore its applications in the real scene are limited. In this paper, we propose a self-adaptive transfer learning (SATL) strategy to fill the domain gap between multicenter datasets. Specifically, the encoder of a DL model that is pre-trained on the source domain is used to initialize the encoder of a reconstruction model. Then, the reconstruction model is trained using only unlabeled image data from the target domain, which makes the encoder in the model adapt itself to extract useful high-level features both for target domain images encoding and glaucoma classification, simultaneously. Experimental results demonstrate that the proposed SATL strategy is effective in the domain adaptation task between one private and two public glaucoma diagnosis datasets, i.e. pri-RFG, REFUGE, and LAG. Moreover, the proposed strategy is completely independent of the source domain data, which meets the real scene application and the privacy protection policy.

eess.IV↗

arXiv2021

Searching for heavy neutrino in terms of tau lepton at future hadron collider

The tau lepton plays important role in the correlation between the low-energy neutrino oscillation data and the lepton flavor structure in heavy neutrino decay. We investigate the lepton flavor signatures with tau lepton at hadron collider through lepton number violating (LNV) processes. In the Type I Seesaw with U$(1)_{\rm B-L}$ extension, we study the pair production of heavy neutrinos via a $Z'$ resonance. We present a detailed assessment of the search sensitivity to the channels with tau lepton in the subsequent decay of heavy neutrinos. For the benchmark model with $Z'$ only coupled to the third generation fermions, we find that the future circular collider (FCC-hh) can discover the LNV signal with tau lepton for $M_{Z'}$ up to 2.2 (3) TeV with the gauge coupling $g'=0.6$ and the integrated luminosity of 3 (30) ab$^{-1}$. The test on the flavor combinations of SM charged leptons would reveal the specific nature of different heavy neutrinos.

hep-ph↗

arXiv2021

Constraints on general neutrino interactions with exotic fermion from neutrino-electron scattering experiments

The couplings between the neutrinos and exotic fermion can be probed in both neutrino scattering experiments and dark matter direct detection experiments. We present a detailed analysis of the general neutrino interactions with an exotic fermion and electrons at neutrino-electron scattering experiments. We obtain the constraints on the coupling coefficients of the scalar, pseudoscalar, vector, axialvector, tensor and electromagnetic dipole interactions from the CHARM-II, TEXONO and Borexino experiments. For the flavor-universal interactions, we find that the Borexino experiment sets the strongest bounds in the low mass region for the electromagnetic dipole interactions, and the CHARM-II experiment dominates the bounds for other scenarios. If the interactions are flavor dependent, the bounds from the CHARM-II or TEXONO experiment can be avoided, and there are correlations between the flavored coupling coefficients for the Borexino experiment. We also discuss the detection of sub-MeV DM absorbed by bound electron targets and illustrate that the vector coefficients preferred by XENON1T data is allowed by the neutrino experiments.

hep-ph↗