arXiv subjects

Liang Shi

Publications and source records attributed to Liang Shi.

At least 55 records · Page 3Linked to original sources

arXiv2023

Real Face Foundation Representation Learning for Generalized Deepfake Detection

The emergence of deepfake technologies has become a matter of social concern as they pose threats to individual privacy and public security. It is now of great significance to develop reliable deepfake detectors. However, with numerous face manipulation algorithms present, it is almost impossible to collect sufficient representative fake faces, and it is hard for existing detectors to generalize to all types of manipulation. Therefore, we turn to learn the distribution of real faces, and indirectly identify fake images that deviate from the real face distribution. In this study, we propose Real Face Foundation Representation Learning (RFFR), which aims to learn a general representation from large-scale real face datasets and detect potential artifacts outside the distribution of RFFR. Specifically, we train a model on real face datasets by masked image modeling (MIM), which results in a discrepancy between input faces and the reconstructed ones when applying the model on fake samples. This discrepancy reveals the low-level artifacts not contained in RFFR, making it easier to build a deepfake detector sensitive to all kinds of potential artifacts outside the distribution of RFFR. Extensive experiments demonstrate that our method brings about better generalization performance, as it significantly outperforms the state-of-the-art methods in cross-manipulation evaluations, and has the potential to further improve by introducing extra real faces for training RFFR.

cs.CV↗

arXiv2022

A photonic chip-based machine learning approach for the prediction of molecular properties

Machine learning methods have revolutionized the discovery process of new molecules and materials. However, the intensive training process of neural networks for molecules with ever-increasing complexity has resulted in exponential growth in computation cost, leading to long simulation time and high energy consumption. Photonic chip technology offers an alternative platform for implementing neural networks with faster data processing and lower energy usage compared to digital computers. Photonics technology is naturally capable of implementing complex-valued neural networks at no additional hardware cost. Here, we demonstrate the capability of photonic neural networks for predicting the quantum mechanical properties of molecules. To the best of our knowledge, this work is the first to harness photonic technology for machine learning applications in computational chemistry and molecular sciences, such as drug discovery and materials design. We further show that multiple properties can be learned simultaneously in a photonic chip via a multi-task regression learning algorithm, which is also the first of its kind as well, as most previous works focus on implementing a network in the classification task.

cs.ET↗

arXiv2022

MSRL: Distributed Reinforcement Learning with Dataflow Fragments

Reinforcement learning (RL) trains many agents, which is resource-intensive and must scale to large GPU clusters. Different RL training algorithms offer different opportunities for distributing and parallelising the computation. Yet, current distributed RL systems tie the definition of RL algorithms to their distributed execution: they hard-code particular distribution strategies and only accelerate specific parts of the computation (e.g. policy network updates) on GPU workers. Fundamentally, current systems lack abstractions that decouple RL algorithms from their execution. We describe MindSpore Reinforcement Learning (MSRL), a distributed RL training system that supports distribution policies that govern how RL training computation is parallelised and distributed on cluster resources, without requiring changes to the algorithm implementation. MSRL introduces the new abstraction of a fragmented dataflow graph, which maps Python functions from an RL algorithm's training loop to parallel computational fragments. Fragments are executed on different devices by translating them to low-level dataflow representations, e.g. computational graphs as supported by deep learning engines, CUDA implementations or multi-threaded CPU processes. We show that MSRL subsumes the distribution strategies of existing systems, while scaling RL training to 64 GPUs.

cs.LG↗

arXiv2022

Variational Quantum Simulations of Finite-Temperature Dynamical Properties via Thermofield Dynamics

The recent advancement of quantum computer hardware offers the potential to simulate quantum many-body systems beyond the capability of its classical counterparts. However, most current works focus on simulating the ground-state properties or pure-state quantum dynamics of quantum systems. Finite-temperature physics are much less explored, and existing literature mainly consider equilibrium properties. Here we present a variational quantum simulation protocol based on the thermofield dynamics formalism to simulate the dynamical and non-equilibrium finite-temperature properties of quantum systems with near-term quantum computers. Compared to previous approaches in computing the equilibrium dynamical properties, our approach does not require the computationally expensive initial state sampling. More importantly, our approach is capable of simulating non-equilibrium phenomena which have not been previously explored with quantum computers. Numerical simulations of molecular absorption spectrum and spin quenched dynamics affirm the capability of our method.

quant-ph↗

arXiv2022

LDP-IDS: Local Differential Privacy for Infinite Data Streams

Streaming data collection is essential to real-time data analytics in various IoTs and mobile device-based systems, which, however, may expose end users' privacy. Local differential privacy (LDP) is a promising solution to privacy-preserving data collection and analysis. However, existing few LDP studies over streams are either applicable to finite streams only or suffering from insufficient protection. This paper investigates this problem by proposing LDP-IDS, a novel $w$-event LDP paradigm to provide practical privacy guarantee for infinite streams at users end, and adapting the popular budget division framework in centralized differential privacy (CDP). By constructing a unified error analysi for LDP, we first develop two adatpive budget division-based LDP methods for LDP-IDS that can enhance data utility via leveraging the non-deterministic sparsity in streams. Beyond that, we further propose a novel population division framework that can not only avoid the high sensitivity of LDP noise to budget division but also require significantly less communication. Based on the framework, we also present two adaptive population division methods for LDP-IDS with theoretical analysis. We conduct extensive experiments on synthetic and real-world datasets to evaluate the effectiveness and efficiency pf our proposed frameworks and methods. Experimental results demonstrate that, despite the effectiveness of the adaptive budget division methods, the proposed population division framework and methods can further achieve much higher effectiveness and efficiency.

cs.DB↗

arXiv2021

Simulating Energy Transfer in Molecular Systems with Digital Quantum Computers

Quantum computers have the potential to simulate chemical systems beyond the capability of classical computers. Recent developments in hybrid quantum-classical approaches enable the determinations of the ground or low energy states of molecular systems. Here, we extend near-term quantum simulations of chemistry to time-dependent processes by simulating energy transfer in organic semiconducting molecules. We developed a multi-scale modeling workflow that combines conventional molecular dynamics and quantum chemistry simulations with hybrid variational quantum algorithm to compute the exciton dynamics in both the single excitation subspace (i.e. Frenkel Hamiltonian) and the full-Hilbert space (i.e. multi-exciton) regimes. Our numerical examples demonstrate the feasibility of our approach, and simulations on IBM Q devices capture the qualitative behaviors of exciton dynamics, but with considerable errors. We present an error mitigation technique that combines experimental results from the variational and Trotter algorithms, and obtain significantly improved quantum dynamics. Our approach opens up new opportunities for modeling quantum dynamics in chemical, biological and material systems with quantum computers.

quant-ph↗

arXiv2021

Graph Self-Supervised Learning for Optoelectronic Properties of Organic Semiconductors

The search for new high-performance organic semiconducting molecules is challenging due to the vastness of the chemical space, machine learning methods, particularly deep learning models like graph neural networks (GNNs), have shown promising potential to address such challenge. However, practical applications of GNNs for chemistry are often limited by the availability of labelled data. Meanwhile, unlabelled molecular data is abundant and could potentially be utilized to alleviate the scarcity issue of labelled data. Here, we advocate the use of self-supervised learning to improve the performance of GNNs by pre-training them with unlabeled molecular data. We investigate regression problems involving ground and excited state properties, both relevant for optoelectronic properties of organic semiconductors. Additionally, we extend the self-supervised learning strategy to molecules in non-equilibrium configurations which are important for studying the effects of disorder. In all cases, we obtain considerable performance improvement over results without pre-training, in particular when labelled training data is limited, and such improvement is attributed to the capability of self-supervised learning in identifying structural similarity among unlabeled molecules.

physics.chem-ph↗

arXiv2021

Variational Quantum Simulation of Chemical Dynamics with Quantum Computers

Classical simulation of real-space quantum dynamics is challenging due to the exponential scaling of computational cost with system dimensions. Quantum computer offers the potential to simulate quantum dynamics with polynomial complexity; however, existing quantum algorithms based on the split-operator techniques require large-scale fault-tolerant quantum computers that remain elusive in the near future. Here we present variational simulations of real-space quantum dynamics suitable for implementation in Noisy Intermediate-Scale Quantum (NISQ) devices. The Hamiltonian is first encoded onto qubits using a discrete variable representation (DVR) and binary encoding scheme. We show that direct application of real-time variational quantum algorithm based on the McLachlan's principle is inefficient as the measurement cost grows exponentially with the qubit number for general potential energy and extremely small time-step size is required to achieve accurate results. Motivated by the insights that most chemical dynamics occur in the low energy subspace, we propose a subspace expansion method by projecting the total Hamiltonian, including the time-dependent driving field, onto the system low-energy eigenstate subspace using quantum computers, the exact quantum dynamics within the subspace can then be solved classically. We show that the measurement cost of the subspace approach grows polynomially with dimensionality for general potential energy. Our numerical examples demonstrate the capability of our approach, even under intense laser fields. Our work opens the possibility of simulating chemical dynamics with NISQ hardware.

quant-ph↗

arXiv2021

Simulation of Condensed-Phase Spectroscopy with Near-term Digital Quantum Computer

Spectroscopy is an indispensable tool in understanding the structures and dynamics of molecular systems. However computational modelling of spectroscopy is challenging due to the exponential scaling of computational complexity with system sizes unless drastic approximations are made. Quantum computer could potentially overcome these classically intractable computational tasks, but existing approaches using quantum computers to simulate spectroscopy can only handle isolated and static molecules. In this work we develop a workflow that combines multi-scale modeling and time-dependent variational quantum algorithm to compute the linear spectroscopy of systems interacting with their condensed-phase environment via the relevant time correlation function. We demonstrate the feasibility of our approach by numerically simulating the UV-Vis absorption spectra of organic semiconductors. We show that our dynamical approach captures several spectral features that are otherwise overlooked by static methods. Our method can be directly used for other linear condensed-phase spectroscopy and could potentially be extended to nonlinear multi-dimensional spectroscopy.

quant-ph↗

arXiv2021

Pre-training with Meta Learning for Chinese Word Segmentation

Recent researches show that pre-trained models (PTMs) are beneficial to Chinese Word Segmentation (CWS). However, PTMs used in previous works usually adopt language modeling as pre-training tasks, lacking task-specific prior segmentation knowledge and ignoring the discrepancy between pre-training tasks and downstream CWS tasks. In this paper, we propose a CWS-specific pre-trained model METASEG, which employs a unified architecture and incorporates meta learning algorithm into a multi-criteria pre-training task. Empirical results show that METASEG could utilize common prior segmentation knowledge from different existing criteria and alleviate the discrepancy between pre-trained models and downstream CWS tasks. Besides, METASEG can achieve new state-of-the-art performance on twelve widely-used CWS datasets and significantly improve model performance in low-resource settings.

cs.CL↗

arXiv2020

Network experimentation at scale

We describe our framework, deployed at Facebook, that accounts for interference between experimental units through cluster-randomized experiments. We document this system, including the design and estimation procedures, and detail insights we have gained from the many experiments that have used this system at scale. We introduce a cluster-based regression adjustment that substantially improves precision for estimating global treatment effects as well as testing for interference as part of our estimation procedure. With this regression adjustment, we find that imbalanced clusters can better account for interference than balanced clusters without sacrificing accuracy. In addition, we show how logging exposure to a treatment can be used for additional variance reduction. Interference is a widely acknowledged issue with online field experiments, yet there is less evidence from real-world experiments demonstrating interference in online settings. We fill this gap by describing two case studies that capture significant network effects and highlight the value of this experimentation framework.

cs.SI↗

arXiv2020

Unified Multi-Criteria Chinese Word Segmentation with BERT

Multi-Criteria Chinese Word Segmentation (MCCWS) aims at finding word boundaries in a Chinese sentence composed of continuous characters while multiple segmentation criteria exist. The unified framework has been widely used in MCCWS and shows its effectiveness. Besides, the pre-trained BERT language model has been also introduced into the MCCWS task in a multi-task learning framework. In this paper, we combine the superiority of the unified framework and pretrained language model, and propose a unified MCCWS model based on BERT. Moreover, we augment the unified BERT-based MCCWS model with the bigram features and an auxiliary criterion classification task. Experiments on eight datasets with diverse criteria demonstrate that our methods could achieve new state-of-the-art results for MCCWS.

cs.CL↗

arXiv2020

nsCouette -- A high-performance code for direct numerical simulations of turbulent Taylor-Couette flow

We present nsCouette, a highly scalable software tool to solve the Navier-Stokes equations for incompressible fluid flow between differentially heated and independently rotating, concentric cylinders. It is based on a pseudospectral spatial discretization and dynamic time-stepping. It is implemented in modern Fortran with a hybrid MPI-OpenMP parallelization scheme and thus designed to compute turbulent flows at high Reynolds and Rayleigh numbers. An additional GPU implementation (C-CUDA) for intermediate problem sizes and a basic version for turbulent pipe flow (nsPipe) are also provided.

physics.flu-dyn↗

arXiv2019

Deep Learning for Optoelectronic Properties of Organic Semiconductors

Atomistic modeling of energetic disorder in organic semiconductors (OSCs) and its effects on the optoelectronic properties of OSCs requires a large number of excited-state electronic-structure calculations, a computationally daunting task for many OSC applications. In this work, we advocate the use of deep learning to address this challenge and demonstrate that state-of-the-art deep neural networks (DNNs) are capable of predicting the electronic properties of OSCs at an accuracy comparable with the quantum chemistry methods used for generating training data. We extensively investigate the performances of four recent DNNs (deep tensor neural network, SchNet, message passing neural network, and multilevel graph convolutional neural network) in predicting various electronic properties of an important class of OSCs, i.e., oligothiophenes (OTs), including their HOMO and LUMO energies, excited-state energies and associated transition dipole moments. We find that SchNet shows the best performance for OTs of different sizes (from bithiophene to sexithiophene), achieving average prediction errors in the range of 20-80meV compared to the results from (time-dependent) density functional theory. We show that SchNet also consistently outperforms shallow feed-forward neural networks, especially in difficult cases with large molecules or limited training data. We further show that SchNet could predict the transition dipole moment accurately, a task previously known to be difficult for feed-forward neural networks, and we ascribe the relatively large errors in transition dipole prediction seen for some OT configurations to the charge-transfer character of their excited states. Finally, we demonstrate the effectiveness of SchNet by modeling the UV-Vis absorption spectra of OTs in dichloromethane and a good agreement is observed between the calculated and experimental spectra.

physics.chem-ph↗

arXiv2017

Influence of subgrid scale models on the buffer sublayer in channel flow

Subgrid-scale (SGS) models are critical in large-eddy simulations (LES) of turbulent flows. In this paper we conduct a comparative study on different SGS models, including one-k-equation, wall-adapting local eddy-viscosity (WALE), Sigma and shear-constrained model. Wall-resolved LES simulations of channel flows are performed with a finite volume code at shear Reynolds number Re_τ= 395. In the simulations, the buffer sublayer turns out to be the most sensitive to the SGS model. Through the analysis of the mean velocity, the second-order moments, the SGS viscosity and the fluctuating vorticities, it is shown that the WALE and Sigma model outperform others significantly in terms of fluctuations while the constrained model improves slightly the mean velocity. The results also indicate that the SGS dissipation influences strongly the velocity fluctuations but not the mean flow nor the log-layer mismatch.

physics.flu-dyn↗

arXiv2017

The Enhancement of Interfacial Exciton Dissociation by Energetic Disorder is a Nonequilibrium Effect

The dissociation of excited electron-hole pairs is a microscopic process that is fundamental to the performance of photovoltaic systems. For this process to be successful, the oppositely charged electron and hole must overcome an electrostatic binding energy before they undergo ground state recombination. Here we use a simple model of charge dynamics to investigate the role of molecular disorder in this process. This model reveals that moderate spatial variations in electronic energy levels, such as those that arise in disordered molecular systems, can actually increase charge dissociation yields. We demonstrate that this is a nonequilibrium effect that is mediated by the dissipation driven formation of partially dissociated intermediate states that are long-lived because they cannot easily recombine. We present a kinetic model that incorporates these states and show that it is capable of reproducing similar behavior when it is parameterized with nonequilibrium rates.

cond-mat.mes-hall↗

arXiv2017

Hydrodynamic turbulence in quasi-Keplerian rotating flows

We report a direct-numerical-simulation study of Taylor-Couette flow in the quasi-Keplerian regime at shear Reynolds numbers up to $\mathcal{O}(10^5)$. Quasi-Keplerian rotating flow has been investigated for decades as a simplified model system to study the origin of turbulence in accretion disks that is not fully understood. The flow in this study is axially periodic and thus the experimental end-wall effects on the stability of the flow are avoided. Using optimal linear perturbations as initial conditions, our simulations find no sustained turbulence: the strong initial perturbations distort the velocity profile and trigger turbulence that eventually decays.

physics.flu-dyn↗

arXiv2017

A Flexible Phase Retrieval Framework for Flux-limited Coherent X-Ray Imaging

Coherent X-ray diffraction imaging (CXDI) experiments are intrinsically limited by shot noise, a lack of prior knowledge about the sample's support, and missing measurements due to the experimental geometry. We propose a flexible, iterative phase retrieval framework that allows for accurate modeling of Gaussian or Poissonian noise statistics, modified support updates, regularization of reconstructed signals, and handling of missing data in the observations. The proposed method is efficiently solved using alternating direction method of multipliers (ADMM) and is demonstrated to consistently outperform state-of-the-art algorithms for low-photon phase retrieval from CXDI experiments, both for simulated diffraction patterns and for experimental measurements.

physics.bio-ph↗