arXiv subjects

Rui Wen

Publications and source records attributed to Rui Wen.

70 records · Page 4Linked to original sources

arXiv2022

Finding Influential Instances for Distantly Supervised Relation Extraction

Distant supervision (DS) is a strong way to expand the datasets for enhancing relation extraction (RE) models but often suffers from high label noise. Current works based on attention, reinforcement learning, or GAN are black-box models so they neither provide meaningful interpretation of sample selection in DS nor stability on different domains. On the contrary, this work proposes a novel model-agnostic instance sampling method for DS by influence function (IF), namely REIF. Our method identifies favorable/unfavorable instances in the bag based on IF, then does dynamic instance sampling. We design a fast influence sampling algorithm that reduces the computational complexity from $\mathcal{O}(mn)$ to $\mathcal{O}(1)$, with analyzing its robustness on the selected sampling function. Experiments show that by simply sampling the favorable instances during training, REIF is able to win over a series of baselines that have complicated architectures. We also demonstrate that REIF can support interpretable instance selection.

cs.LG↗

arXiv2021

ML-Doctor: Holistic Risk Assessment of Inference Attacks Against Machine Learning Models

Inference attacks against Machine Learning (ML) models allow adversaries to learn sensitive information about training data, model parameters, etc. While researchers have studied, in depth, several kinds of attacks, they have done so in isolation. As a result, we lack a comprehensive picture of the risks caused by the attacks, e.g., the different scenarios they can be applied to, the common factors that influence their performance, the relationship among them, or the effectiveness of possible defenses. In this paper, we fill this gap by presenting a first-of-its-kind holistic risk assessment of different inference attacks against machine learning models. We concentrate on four attacks -- namely, membership inference, model inversion, attribute inference, and model stealing -- and establish a threat model taxonomy. Our extensive experimental evaluation, run on five model architectures and four image datasets, shows that the complexity of the training dataset plays an important role with respect to the attack's performance, while the effectiveness of model stealing and membership inference attacks are negatively correlated. We also show that defenses like DP-SGD and Knowledge Distillation can only mitigate some of the inference attacks. Our analysis relies on a modular re-usable software, ML-Doctor, which enables ML model owners to assess the risks of deploying their models, and equally serves as a benchmark tool for researchers and practitioners.

cs.CR↗

arXiv2021

On the Fairness of Randomized Trials for Recommendation with Heterogeneous Demographics and Beyond

Observed events in recommendation are consequence of the decisions made by a policy, thus they are usually selectively labeled, namely the data are Missing Not At Random (MNAR), which often causes large bias to the estimate of true outcomes risk. A general approach to correct MNAR bias is performing small Randomized Controlled Trials (RCTs), where an additional uniform policy is employed to randomly assign items to each user. In this work, we concentrate on the fairness of RCTs under both homogeneous and heterogeneous demographics, especially analyzing the bias for the least favorable group on the latter setting. Considering RCTs' limitations, we propose a novel Counterfactual Robust Risk Minimization (CRRM) framework, which is totally free of expensive RCTs, and derive its theoretical generalization error bound. At last, empirical experiments are performed on synthetic tasks and real-world data sets, substantiating our method's superiority both in fairness and generalization.

cs.LG↗

arXiv2021

High-order baryon number fluctuations within the fRG approach

We compute high-order baryon number fluctuations at finite temperature and density within a QCD-assisted low energy effective field theory. Quantum, thermal and density fluctuations are incorporated with the functional renormalization group approach. Quantum and in-medium fluctuations are encoded via the evolution of renormalization group flow equations. The resulting fourth- and sixth-order baryon number fluctuations meet the lattice benchmark results at vanishing density. They are consistent with experimental measurements, and in particular, the non-monotonic dependence of the kurtosis of net-baryon number distributions on the collision energy is observed in our calculations. This non-monotonicity arises from the increasingly sharpened chiral crossover with the decrease of collision energy.

hep-ph↗

arXiv2021

Exact Quench Dynamics from Algebraic Geometry

We develop a systematic approach to compute physical observables of integrable spin chains with finite length. Our method is based on Bethe ansatz solution of the integrable spin chain and computational algebraic geometry. The final results are analytic and no longer depend on Bethe roots. The computation is purely algebraic and does not rely on further assumptions or numerics. This method can be applied to compute a broad family of physical quantities in integrable quantum spin chains. We demonstrate the power of the method by computing two important quantities in quench dynamics: the diagonal entropy and the Loschmidt echo and obtain new analytic results.

hep-th↗

arXiv2021

PRGC: Potential Relation and Global Correspondence Based Joint Relational Triple Extraction

Joint extraction of entities and relations from unstructured texts is a crucial task in information extraction. Recent methods achieve considerable performance but still suffer from some inherent limitations, such as redundancy of relation prediction, poor generalization of span-based extraction and inefficiency. In this paper, we decompose this task into three subtasks, Relation Judgement, Entity Extraction and Subject-object Alignment from a novel perspective and then propose a joint relational triple extraction framework based on Potential Relation and Global Correspondence (PRGC). Specifically, we design a component to predict potential relations, which constrains the following entity extraction to the predicted relation subset rather than all relations; then a relation-specific sequence tagging component is applied to handle the overlapping problem between subjects and objects; finally, a global correspondence component is designed to align the subject and object into a triple with low-complexity. Extensive experiments show that PRGC achieves state-of-the-art performance on public benchmarks with higher efficiency and delivers consistent performance gain on complex scenarios of overlapping triples.

cs.CL↗

arXiv2021

Lifelong Learning based Disease Diagnosis on Clinical Notes

Current deep learning based disease diagnosis systems usually fall short in catastrophic forgetting, i.e., directly fine-tuning the disease diagnosis model on new tasks usually leads to abrupt decay of performance on previous tasks. What is worse, the trained diagnosis system would be fixed once deployed but collecting training data that covers enough diseases is infeasible, which inspires us to develop a lifelong learning diagnosis system. In this work, we propose to adopt attention to combine medical entities and context, embedding episodic memory and consolidation to retain knowledge, such that the learned model is capable of adapting to sequential disease-diagnosis tasks. Moreover, we establish a new benchmark, named Jarvis-40, which contains clinical notes collected from various hospitals. Our experiments show that the proposed method can achieve state-of-the-art performance on the proposed benchmark.

cs.AI↗

arXiv2021

Online Disease Self-diagnosis with Inductive Heterogeneous Graph Convolutional Networks

We propose a Healthcare Graph Convolutional Network (HealGCN) to offer disease self-diagnosis service for online users based on Electronic Healthcare Records (EHRs). Two main challenges are focused in this paper for online disease diagnosis: (1) serving cold-start users via graph convolutional networks and (2) handling scarce clinical description via a symptom retrieval system. To this end, we first organize the EHR data into a heterogeneous graph that is capable of modeling complex interactions among users, symptoms and diseases, and tailor the graph representation learning towards disease diagnosis with an inductive learning paradigm. Then, we build a disease self-diagnosis system with a corresponding EHR Graph-based Symptom Retrieval System (GraphRet) that can search and provide a list of relevant alternative symptoms by tracing the predefined meta-paths. GraphRet helps enrich the seed symptom set through the EHR graph when confronting users with scarce descriptions, hence yield better diagnosis accuracy. At last, we validate the superiority of our model on a large-scale EHR dataset.

cs.LG↗

arXiv2021

Node-Level Membership Inference Attacks Against Graph Neural Networks

Many real-world data comes in the form of graphs, such as social networks and protein structure. To fully utilize the information contained in graph data, a new family of machine learning (ML) models, namely graph neural networks (GNNs), has been introduced. Previous studies have shown that machine learning models are vulnerable to privacy attacks. However, most of the current efforts concentrate on ML models trained on data from the Euclidean space, like images and texts. On the other hand, privacy risks stemming from GNNs remain largely unstudied. In this paper, we fill the gap by performing the first comprehensive analysis of node-level membership inference attacks against GNNs. We systematically define the threat models and propose three node-level membership inference attacks based on an adversary's background knowledge. Our evaluation on three GNN structures and four benchmark datasets shows that GNNs are vulnerable to node-level membership inference even when the adversary has minimal background knowledge. Besides, we show that graph density and feature similarity have a major impact on the attack's success. We further investigate two defense mechanisms and the empirical results indicate that these defenses can reduce the attack performance but with moderate utility loss.

cs.CR↗

arXiv2021

Critical behaviors of the $O(4)$ and $Z(2)$ symmetries in the QCD phase diagram

In this work we have studied the QCD phase structure and critical dynamics related to the 3-$d$ $O(4)$ and $Z(2)$ symmetry universality classes in the two-flavor quark-meson low energy effective theory within the functional renormalization group approach. We have employed the expansion of Chebyshev polynomials to solve the flow equation for the order-parameter potential. The chiral phase transition line of $O(4)$ symmetry in the chiral limit, and the $Z(2)$ line of critical end points related to the explicit chiral symmetry breaking are depicted in the phase diagram. Various critical exponents related to the order parameter, chiral susceptibilities and correlation lengths have been calculated for the 3-$d$ $O(4)$ and $Z(2)$ universality classes in the phase diagram, respectively. We find that the critical exponents obtained in the computation, where a field-dependent mesonic nontrivial dispersion relation is taken into account, are in quantitative agreement with results from other approaches, e.g., the conformal bootstrap, Monte Carlo simulations and $d=3$ perturbation expansion, etc. Moreover, the size of the critical regime in the QCD phase diagram is found to be very small.

hep-ph↗

arXiv2020

Information Theoretic Counterfactual Learning from Missing-Not-At-Random Feedback

Counterfactual learning for dealing with missing-not-at-random data (MNAR) is an intriguing topic in the recommendation literature since MNAR data are ubiquitous in modern recommender systems. Missing-at-random (MAR) data, namely randomized controlled trials (RCTs), are usually required by most previous counterfactual learning methods for debiasing learning. However, the execution of RCTs is extraordinarily expensive in practice. To circumvent the use of RCTs, we build an information-theoretic counterfactual variational information bottleneck (CVIB), as an alternative for debiasing learning without RCTs. By separating the task-aware mutual information term in the original information bottleneck Lagrangian into factual and counterfactual parts, we derive a contrastive information loss and an additional output confidence penalty, which facilitates balanced learning between the factual and counterfactual domains. Empirical evaluation on real-world datasets shows that our CVIB significantly enhances both shallow and deep models, which sheds light on counterfactual learning in recommendation that goes beyond RCTs.

cs.LG↗

arXiv2020

Correlations of conserved charges and QCD phase structure

Correlations of conserved charges, i.e., the baryon number, electric charge and the strangeness, have been calculated at finite temperature and chemical potential up to the fourth order. The calculations are done in a 2+1 flavor low energy effective theory, where quantum and thermal fluctuations are encoded through the evolution of flow equations within the functional renormalization group approach. Strangeness neutrality and a fixed ratio of the electric charge to the baryon number density are implemented throughout the computation. We find that higher-order correlations carry more sensitive critical dynamics in comparison to the quadratic ones, and a non-monotonic dependence of the fourth-order correlations between the baryon number and strangeness, $-χ^{BS}_{31}/χ^{S}_{2}$ and $χ^{BS}_{22}/χ^{S}_{2}$, on the collision energy is found.

hep-ph↗

arXiv2019

Mesonic dynamics and QCD phase transition

We study the nontrivial dispersion relation of mesons resulting from the splitting of the transversal and longitudinal mesonic wave function renormalizations, and its influences on the QCD phase transition, equation of state, and the fluctuations of baryon number. The calculations are performed in a two-flavor low energy effective model within the functional renormalization group approach. We find that influences of the splitting mesonic wave function renormalizations on the equilibrium thermodynamical bulk properties are mild. Furthermore, we have compare the fixed and running expansion for the effective potential in detail, through which the role of the field-dependent meson and quark wave function renormalizations could be inferred.

hep-ph↗

arXiv2019

Baryon number fluctuations in the 2+1 flavor low energy effective model

High order cumulants of the baryon number distribution are calculated in a 2+1 flavor low energy effective model. Quantum fluctuations are encoded through the functional renormalization group approach. The chiral and deconfinement phase transitions are investigated at finite temperature and baryon chemical potential. The equation of state for the QCD matter, fluctuations of the baryon number and the strangeness up to sixth order are calculated, and one finds the results are consistent with those from lattice QCD.

hep-ph↗

arXiv2019

Holographic (a,c)-charges and Their Universal Relation in d=6 from Massless Higher-order Gravities

Recent studies of holographic properties of massless higher-order gravities, whose linear spectrum contains only the (massless) graviton, yielded some universal relations in $d=4$ dimensions between the holographic $a$, $c$ charges and the overall coefficient factor ${\cal C}_T$ of the energy-momentum tensor two-point function, namely $c=\frac{1}{d-1} \ell\frac{\partial a}{\partial \ell}={\cal C}_T$, where $\ell$ is the AdS radius. The second equality was shown to be valid in all dimensions. In this paper, we establish the first equality in $d=6$ by examining these quantities from $D=7$ higher-order gravities. Consequently the overall coefficient of the two-point function of the energy-momentum tensor is proportional to this $c$ charge, generalizing the well-known $d=4$ result. We identify the relevant $c$ charge as the coefficient of a specific combination of the three type-B anomalies. Modulo total derivatives, this combination involves Riemann tensor at most linearly.

hep-th↗

arXiv2018

Baryon number probability distribution at finite temperature

The probability distribution of the net baryon number is investigated within the functional renormalization group approach. We find that the Roberge-Weiss periodicity related to the Z(3) symmetry of the gluon fields results directly in that, states of the net baryon number $N_B=N\pm 1/3$ with $N\in\mathbb{Z}$ are prohibited, and only those of $N_B=N$ are possible. By employing the probability distribution of the net baryon number, we also compute the cumulants of the baryon number distribution, which are found to be well consistent with those obtained from the generalized susceptibilities. A question about the relation between the color confinement and the probability distribution of net baryon number is put forward.

hep-ph↗