Searcharxiv⌕ Search

arXiv subjects

Han Wang

Publications and source records attributed to Han Wang.

At least 343 records · Page 19Linked to original sources
arXiv2023

Efficient Numerical Algorithm for Large-Scale Damped Natural Gradient Descent

We propose a new algorithm for efficiently solving the damped Fisher matrix in large-scale scenarios where the number of parameters significantly exceeds the number of available samples. This problem is fundamental for natural gradient descent and stochastic reconfiguration. Our algorithm is based on Cholesky decomposition and is generally applicable. Benchmark results show that the algorithm is significantly faster than existing methods.

cs.LG↗
arXiv2023

Toward Incompatible Quantum Limits on Multiparameter Estimation

Achieving the ultimate precisions for multiple parameters simultaneously is an outstanding challenge in quantum physics, because the optimal measurements for incompatible parameters cannot be performed jointly due to the Heisenberg uncertainty principle. In this work, a criterion proposed for multiparameter estimation provides a possible way to beat this curse. According to this criterion, it is possible to mitigate the influence of incompatibility meanwhile improve the ultimate precisions by increasing the variances of the parameter generators simultaneously. For demonstration, a scheme involving high-order Hermite-Gaussian states as probes is proposed for estimating the spatial displacement and angular tilt of light at the same time, and precisions up to 1.45 nm and 4.08 nrad are achieved in experiment simultaneously. Consequently, our findings provide a deeper insight into the role of Heisenberg uncertainty principle in multiparameter estimation, and contribute in several ways to the applications of quantum metrology.

quant-ph↗
arXiv2023

Improved Communication Efficiency in Federated Natural Policy Gradient via ADMM-based Gradient Updates

Federated reinforcement learning (FedRL) enables agents to collaboratively train a global policy without sharing their individual data. However, high communication overhead remains a critical bottleneck, particularly for natural policy gradient (NPG) methods, which are second-order. To address this issue, we propose the FedNPG-ADMM framework, which leverages the alternating direction method of multipliers (ADMM) to approximate global NPG directions efficiently. We theoretically demonstrate that using ADMM-based gradient updates reduces communication complexity from ${O}({d^{2}})$ to ${O}({d})$ at each iteration, where $d$ is the number of model parameters. Furthermore, we show that achieving an $ε$-error stationary convergence requires ${O}(\frac{1}{(1-γ)^{2}ε})$ iterations for discount factor $γ$, demonstrating that FedNPG-ADMM maintains the same convergence rate as the standard FedNPG. Through evaluation of the proposed algorithms in MuJoCo environments, we demonstrate that FedNPG-ADMM maintains the reward performance of standard FedNPG, and that its convergence rate improves when the number of federated agents increases.

cs.LG↗
arXiv2023

Deep Variational Free Energy Approach to Dense Hydrogen

We developed a deep generative model-based variational free energy approach to the equations of state of dense hydrogen. We employ a normalizing flow network to model the proton Boltzmann distribution and a fermionic neural network to model the electron wave function at given proton positions. By jointly optimizing the two neural networks we reached a comparable variational free energy to the previous coupled electron-ion Monte Carlo calculation. The predicted equation of state of dense hydrogen under planetary conditions is denser than the findings of ab initio molecular dynamics calculation and empirical chemical model. Moreover, direct access to the entropy and free energy of dense hydrogen opens new opportunities in planetary modeling and high-pressure physics research.

cond-mat.str-el↗
arXiv2023

Anomalous thermal transport across the superionic transition in ice

Superionic ices with highly mobile protons within the stable oxygen sub-lattice occupy an important proportion of the phase diagram of ice and widely exist in the interior of icy giants and throughout the universe. Understanding the thermal transport in superionic ice is vital for the thermal evolution of icy planets. However, it is highly challenging due to the extreme thermodynamic conditions and dynamical nature of protons, beyond the capability of the traditional lattice dynamics and empirical potential molecular dynamics approaches. In this work, by utilizing the deep potential molecular dynamics approach, we investigate the thermal conductivity of ice-VII and superionic ice-VII" along the isobar of $p = 30\ \rm{GPa}$. A non-monotonic trend of thermal conductivity with elevated temperature is observed. Through heat flux decomposition and trajectory-based spectra analysis, we show that the thermally-activated proton diffusion in ice-VII and superionic ice-VII" contribute significantly to heat convection, while the broadening in vibrational energy peaks and significant softening of transverse acoustic branches lead to a reduction in heat conduction. The competition between proton diffusion and phonon scattering results in anomalous thermal transport across the superionic transition in ice. This work unravels the important role of proton diffusion in the thermal transport of high-pressure ice. Our approach provides new insights into modeling the thermal transport and atomistic dynamics in superionic materials.

physics.comp-ph↗
arXiv2023

1D-confined crystallization routes for tungsten phosphides

Topological materials confined in one-dimension (1D) can transform computing technologies, such as 1D topological semimetals for nanoscale interconnects and 1D topological superconductors for fault-tolerant quantum computing. As such, understanding crystallization of 1D-confined topological materials is critical. Here, we demonstrate 1D-confined crystallization routes during template-assisted nanowire synthesis where we observe diameter-dependent phase selectivity for topological metal tungsten phosphides. A phase bifurcation occurs to produce tungsten monophosphide and tungsten diphosphide at the cross-over nanowire diameter of ~ 35 nm. Four-dimensional scanning transmission electron microscopy was used to identify the two phases and to map crystallographic orientations of grains at a few nm resolution. The 1D-confined phase selectivity is attributed to the minimization of the total surface energy, which depends on the nanowire diameter and chemical potentials of precursors. Theoretical calculations were carried out to construct the diameter-dependent phase diagram, which agrees with experimental observations. Our find-ings suggest a new crystallization route to stabilize topological materials confined in 1D.

cond-mat.mtrl-sci↗
arXiv2023

Oracle Complexity Reduction for Model-free LQR: A Stochastic Variance-Reduced Policy Gradient Approach

We investigate the problem of learning an $ε$-approximate solution for the discrete-time Linear Quadratic Regulator (LQR) problem via a Stochastic Variance-Reduced Policy Gradient (SVRPG) approach. Whilst policy gradient methods have proven to converge linearly to the optimal solution of the model-free LQR problem, the substantial requirement for two-point cost queries in gradient estimations may be intractable, particularly in applications where obtaining cost function evaluations at two distinct control input configurations is exceptionally costly. To this end, we propose an oracle-efficient approach. Our method combines both one-point and two-point estimations in a dual-loop variance-reduced algorithm. It achieves an approximate optimal solution with only $O\left(\log\left(1/ε\right)^β\right)$ two-point cost information for $β\in (0,1)$.

math.OC↗
arXiv2023

DPA-1: Pretraining of Attention-based Deep Potential Model for Molecular Simulation

Machine learning assisted modeling of the inter-atomic potential energy surface (PES) is revolutionizing the field of molecular simulation. With the accumulation of high-quality electronic structure data, a model that can be pretrained on all available data and finetuned on downstream tasks with a small additional effort would bring the field to a new stage. Here we propose DPA-1, a Deep Potential model with a novel attention mechanism, which is highly effective for representing the conformation and chemical spaces of atomic systems and learning the PES. We tested DPA-1 on a number of systems and observed superior performance compared with existing benchmarks. When pretrained on large-scale datasets containing 56 elements, DPA-1 can be successfully applied to various downstream tasks with a great improvement of sample efficiency. Surprisingly, for different elements, the learned type embedding parameters form a $spiral$ in the latent space and have a natural correspondence with their positions on the periodic table, showing interesting interpretability of the pretrained DPA-1 model.

physics.chem-ph↗
arXiv2023

Robust Markov Decision Processes without Model Estimation

Robust Markov Decision Processes (MDPs) are receiving much attention in learning a robust policy which is less sensitive to environment changes. There are an increasing number of works analyzing sample-efficiency of robust MDPs. However, there are two major barriers to applying robust MDPs in practice. First, most works study robust MDPs in a model-based regime, where the transition probability needs to be estimated and requires a large amount of memories $\mathcal{O}(|\mathcal{S}|^2|\mathcal{A}|)$. Second, prior work typically assumes a strong oracle to obtain the optimal solution as an intermediate step to solve robust MDPs. However, in practice, such an oracle does not exist usually. To remove the oracle, we transform the original robust MDPs into an alternative form, which allows us to use stochastic gradient methods to solve the robust MDPs. Moreover, we prove the alternative form still plays a similar role as the original form. With this new formulation, we devise a sample-efficient algorithm to solve the robust MDPs in a model-free regime, which does not require an oracle and trades off a lower storage requirement $\mathcal{O}(|\mathcal{S}||\mathcal{A}|)$ with being able to generate samples from a generative model or Markovian chain. Finally, we validate our theoretical findings via numerical experiments, showing the efficiency with the alternative form of robust MDPs.

stat.ML↗
arXiv2023

Learning Personalized Models with Clustered System Identification

We address the problem of learning linear system models from observing multiple trajectories from different system dynamics. This framework encompasses a collaborative scenario where several systems seeking to estimate their dynamics are partitioned into clusters according to their system similarity. Thus, the systems within the same cluster can benefit from the observations made by the others. Considering this framework, we present an algorithm where each system alternately estimates its cluster identity and performs an estimation of its dynamics. This is then aggregated to update the model of each cluster. We show that under mild assumptions, our algorithm correctly estimates the cluster identities and achieves an approximate sample complexity that scales inversely with the number of systems in the cluster, thus facilitating a more efficient and personalized system identification process.

math.OC↗
arXiv2023

Evaluating GPT-3 Generated Explanations for Hateful Content Moderation

Recent research has focused on using large language models (LLMs) to generate explanations for hate speech through fine-tuning or prompting. Despite the growing interest in this area, these generated explanations' effectiveness and potential limitations remain poorly understood. A key concern is that these explanations, generated by LLMs, may lead to erroneous judgments about the nature of flagged content by both users and content moderators. For instance, an LLM-generated explanation might inaccurately convince a content moderator that a benign piece of content is hateful. In light of this, we propose an analytical framework for examining hate speech explanations and conducted an extensive survey on evaluating such explanations. Specifically, we prompted GPT-3 to generate explanations for both hateful and non-hateful content, and a survey was conducted with 2,400 unique respondents to evaluate the generated explanations. Our findings reveal that (1) human evaluators rated the GPT-generated explanations as high quality in terms of linguistic fluency, informativeness, persuasiveness, and logical soundness, (2) the persuasive nature of these explanations, however, varied depending on the prompting strategy employed, and (3) this persuasiveness may result in incorrect judgments about the hatefulness of the content. Our study underscores the need for caution in applying LLM-generated explanations for content moderation. Code and results are available at https://github.com/Social-AI-Studio/GPT3-HateEval.

cs.CL↗
arXiv2023

Stochastic Optimization of Coupled Power Distribution-Urban Transportation Network Operations with Autonomous Mobility on Demand Systems

Autonomous mobility on demand systems (AMoDS) will significantly affect the operation of coupled power distribution-urban transportation networks (PTNs) by the optimal dispatch of electric vehicles (EVs). This paper proposes an uncertainty method to analyze the operational states of PTNs with AMoDS. First, a PTN operation framework is designed considering the controllable EVs dispatched by AMoDS as well as the uncontrollable driving behaviors of other vehicle users. Then, a bi-level power-traffic flow (PTF) model is proposed to characterize the interaction of power distribution networks (PDNs) and urban transportation networks (UTNs). In the upper level, a social optimum model is established to minimize the operating cost of PDNs and UTNs embedded with controllable EVs. In the lower level, a stochastic user equilibrium (SUE) model is established to minimize the operating cost of uncontrollable EVs and gasoline vehicles (GVs) in UTNs. Finally, a probabilistic PTF analysis method is developed to evaluate PTN operations under environmental and human uncertainties. A regional sensitivity analysis method is proposed to identify the critical uncertainties and quantify the impacts of their distribution ranges on PTN operations. The effectiveness of the proposed method is verified by the PTN consisting of a 21-bus PDN and a 20-node UTN.

eess.SY↗
arXiv2023

Deep Learning Illuminates Spin and Lattice Interaction in Magnetic Materials

Atomistic simulations hold significant value in clarifying crucial phenomena such as phase transitions and energy transport in materials science. Their success stems from the presence of potential energy functions capable of accurately depicting the relationship between system energy and lattice changes. In magnetic materials, two atomic scale degrees of freedom come into play: the lattice and the spin. However, accurately tracing the simultaneous evolution of both lattice and spin in magnetic materials at an atomic scale is a substantial challenge. This is largely due to the complexity involved in depicting the interaction energy precisely, and its influence on lattice and spin-driving forces, such as atomic force and magnetic torque, which continues to be a daunting task in computational science. Addressing this deficit, we present DeepSPIN, a versatile approach that generates high-precision predictive models of energy, atomic forces, and magnetic torque in magnetic systems. This is achieved by integrating first-principles calculations of magnetic excited states with deep learning techniques via active learning. We thoroughly explore the methodology, accuracy, and scalability of our proposed model in this paper. Our technique adeptly connects first-principles computations and atomic-scale simulations of magnetic materials. This synergy presents opportunities to utilize these calculations in devising and tackling theoretical and practical obstacles concerning magnetic materials.

cond-mat.mtrl-sci↗
arXiv2023

Blinkverse: A Database of Fast Radio Bursts

The volume of research on fast radio bursts (FRBs) observation have been seeing a dramatic growth. To facilitate the systematic analysis of the FRB population, we established a database platform, Blinkverse (https://blinkverse.alkaidos.cn), as a central inventory of FRBs from various observatories and with published properties, particularly dynamic spectra from FAST, CHIME, GBT, Arecibo, etc. Blinkverse thus not only forms a superset of FRBCAT, TNS, and CHIME/FRB, but also provides convenient access to thousands of FRB dynamic spectra from FAST, some of which were not available before. Blinkverse is regularly maintained and will be updated by external users in the future. Data entries of FRBs can be retrieved through parameter searches through FRB location, fluence, etc., and their logical combinations. Interactive visualization was built into the platform. We analyzed the energy distribution, period analysis, and classification of FRBs based on data downloaded from Blinkverse. The energy distributions of repeaters and non-repeaters are found to be distinct from one another.

astro-ph.HE↗
arXiv2023

New Covert and Side Channels Based on Retirement

Intel processors utilize the retirement to orderly retire the micro-ops that have been executed out of order. To enhance retirement utilization, the retirement is dynamically shared between two logical cores on the same physical core. However, this shared retirement mechanism creates a potential vulnerability wherein an attacker can exploit the competition for retirement to infer the data of a victim on another logical core on the same physical core. Based on this leakage, we propose two new covert channels: the Different Instructions (DI) covert channel using different instructions for information transmission, and the Same Instructions (SI) covert channel using the same instructions to transmit information. The DI covert channel can achieve 98.5% accuracy with a bandwidth of 1450 Kbps, while the SI covert channel can achieve 94.85% accuracy with a bandwidth of 483.33 Kbps. Furthermore, this paper explores additional applications of retirement: Firstly, retirement is applied to Spectre attacks, resulting in a new variant of Spectre v1, which can achieve 94.17% accuracy with a bandwidth of 29 Kbps; Secondly, retirement is leveraged to infer the programs being executed by the victim, which can infer 10 integer benchmarks of SPEC with 89.28% accuracy. Finally, we discuss possible protection against new covert channels.

cs.CR↗
arXiv2023

Accelerating galaxy dynamical modeling using a neural network for joint lensing and kinematics analyses

Strong gravitational lensing is a powerful tool to provide constraints on galaxy mass distributions and cosmological parameters, such as the Hubble constant, $H_0$. Nevertheless, inference of such parameters from images of lensing systems is not trivial as parameter degeneracies can limit the precision in the measured lens mass and cosmological results. External information on the mass of the lens, in the form of kinematic measurements, is needed to ensure a precise and unbiased inference. Traditionally, such kinematic information has been included in the inference after the image modeling, using spherical Jeans approximations to match the measured velocity dispersion integrated within an aperture. However, as spatially resolved kinematic measurements become available via IFU data, more sophisticated dynamical modeling is necessary. Such kinematic modeling is expensive, and constitutes a computational bottleneck which we aim to overcome with our Stellar Kinematics Neural Network (SKiNN). SKiNN emulates axisymmetric modeling using a neural network, quickly synthesizing from a given mass model a kinematic map which can be compared to the observations to evaluate a likelihood. With a joint lensing plus kinematic framework, this likelihood constrains the mass model at the same time as the imaging data. We show that SKiNN's emulation of a kinematic map is accurate to considerably better precision than can be measured (better than $1\%$ in almost all cases). Using SKiNN speeds up the likelihood evaluation by a factor of $\sim 200$. This speedup makes dynamical modeling economical, and enables lens modelers to make effective use of modern data quality in the JWST era.

astro-ph.GA↗
arXiv2023

Contracting with Heterogeneous Researchers

We study the design of contracts that incentivize a researcher to conduct a costly experiment, extending the work of Yoder (2022) from binary states to a general state space. The cost is private information of the researcher. When the experiment is observable, we find the optimal contract and show that higher types choose more costly experiments, but not necessarily more Blackwell informative ones. When only the experiment result is observable, the principal can still achieve the same optimal outcome if and only if a certain monotonicity condition with respect to types holds. Our analysis demonstrates that the general case is qualitatively different than the binary one, but that the contracting problem remains tractable.

econ.TH↗
arXiv2023

Measuring and Mitigating Interference in Reinforcement Learning

Catastrophic interference is common in many network-based learning systems, and many proposals exist for mitigating it. Before overcoming interference we must understand it better. In this work, we provide a definition and novel measure of interference for value-based reinforcement learning methods such as Fitted Q-Iteration and DQN. We systematically evaluate our measure of interference, showing that it correlates with instability in control performance, across a variety of network architectures. Our new interference measure allows us to ask novel scientific questions about commonly used deep learning architectures and study learning algorithms which mitigate interference. Lastly, we outline a class of algorithms which we call online-aware that are designed to mitigate interference, and show they do reduce interference according to our measure and that they improve stability and performance in several classic control environments.

cs.LG↗