arXiv subjects

Cheng Zhang

Publications and source records attributed to Cheng Zhang.

At least 325 records · Page 18Linked to original sources

arXiv2023

SAM-RL: Sensing-Aware Model-Based Reinforcement Learning via Differentiable Physics-Based Simulation and Rendering

Model-based reinforcement learning (MBRL) is recognized with the potential to be significantly more sample-efficient than model-free RL. How an accurate model can be developed automatically and efficiently from raw sensory inputs (such as images), especially for complex environments and tasks, is a challenging problem that hinders the broad application of MBRL in the real world. In this work, we propose a sensing-aware model-based reinforcement learning system called SAM-RL. Leveraging the differentiable physics-based simulation and rendering, SAM-RL automatically updates the model by comparing rendered images with real raw images and produces the policy efficiently. With the sensing-aware learning pipeline, SAM-RL allows a robot to select an informative viewpoint to monitor the task process. We apply our framework to real world experiments for accomplishing three manipulation tasks: robotic assembly, tool manipulation, and deformable object manipulation. We demonstrate the effectiveness of SAM-RL via extensive experiments. Videos are available on our project webpage at https://sites.google.com/view/rss-sam-rl.

cs.RO↗

arXiv2023

Giant superlinear power dependence of photocurrent based on layered Ta$_2$NiS$_5$ photodetector

Photodetector based on two-dimensional (2D) materials is an ongoing quest in optoelectronics. These 2D photodetectors are generally efficient at low illuminating power but suffer severe recombination processes at high power, which results in the sublinear power dependence of photoresponse and lower optoelectronic efficiency. The desirable superlinear photocurrent is mostly achieved by sophisticated 2D heterostructures or device arrays, while 2D materials rarely show intrinsic superlinear photoresponse. Here, we report the giant superlinear power dependence of photocurrent based on multi-layer Ta$_2$NiS$_5$. While the fabricated photodetector exhibits good sensitivity ($3.1 mS/W$ per square) and fast photoresponse ($31 μ$$s$), the bias-, polarization-, and spatial-resolved measurements point to an intrinsic photoconductive mechanism. By increasing the incident power density from $1.5 μ$W/$μ$$m^{2}$ to $200 μ$W/$μ$$m^{2}$, the photocurrent power dependence varies from sublinear to superlinear. At higher illuminating conditions, a prominent superlinearity is observed with a giant power exponent of $γ=1.5$. The unusual photoresponse can be explained by a two-recombination-center model where the distinct density of states of the recombination centers effectively closes all recombination channels. The fabricated photodetector is integrated into camera for taking photos with enhanced contrast due to the superlinearity. Our work provides an effective route to enable higher optoelectronic efficiency at extreme conditions.

physics.app-ph↗

arXiv2023

Understanding Causality with Large Language Models: Feasibility and Opportunities

We assess the ability of large language models (LLMs) to answer causal questions by analyzing their strengths and weaknesses against three types of causal question. We believe that current LLMs can answer causal questions with existing causal knowledge as combined domain experts. However, they are not yet able to provide satisfactory answers for discovering new knowledge or for high-stakes decision-making tasks with high precision. We discuss possible future directions and opportunities, such as enabling explicit and implicit causal modules as well as deep causal-aware LLMs. These will not only enable LLMs to answer many different types of causal questions for greater impact but also enable LLMs to be more trustworthy and efficient in general.

cs.LG↗

arXiv2023

Automatic Generation of Multiple-Choice Questions

Creating multiple-choice questions to assess reading comprehension of a given article involves generating question-answer pairs (QAPs) and adequate distractors. We present two methods to tackle the challenge of QAP generations: (1) A deep-learning-based end-to-end question generation system based on T5 Transformer with Preprocessing and Postprocessing Pipelines (TP3). We use the finetuned T5 model for our downstream task of question generation and improve accuracy using a combination of various NLP tools and algorithms in preprocessing and postprocessing to select appropriate answers and filter undesirable questions. (2) A sequence-learning-based scheme to generate adequate QAPs via meta-sequence representations of sentences. A meta-sequence is a sequence of vectors comprising semantic and syntactic tags. we devise a scheme called MetaQA to learn meta sequences from training data to form pairs of a meta sequence for a declarative sentence and a corresponding interrogative sentence. The TP3 works well on unseen data, which is complemented by MetaQA. Both methods can generate well-formed and grammatically correct questions. Moreover, we present a novel approach to automatically generate adequate distractors for a given QAP. The method is a combination of part-of-speech tagging, named-entity tagging, semantic-role labeling, regular expressions, domain knowledge bases, word embeddings, word edit distance, WordNet, and other algorithms.

cs.CL↗

arXiv2023

Causal Reasoning in the Presence of Latent Confounders via Neural ADMG Learning

Latent confounding has been a long-standing obstacle for causal reasoning from observational data. One popular approach is to model the data using acyclic directed mixed graphs (ADMGs), which describe ancestral relations between variables using directed and bidirected edges. However, existing methods using ADMGs are based on either linear functional assumptions or a discrete search that is complicated to use and lacks computational tractability for large datasets. In this work, we further extend the existing body of work and develop a novel gradient-based approach to learning an ADMG with non-linear functional relations from observational data. We first show that the presence of latent confounding is identifiable under the assumptions of bow-free ADMGs with non-linear additive noise models. With this insight, we propose a novel neural causal model based on autoregressive flows for ADMG learning. This not only enables us to determine complex causal structural relationships behind the data in the presence of latent confounding, but also estimate their functional relationships (hence treatment effects) simultaneously. We further validate our approach via experiments on both synthetic and real-world datasets, and demonstrate the competitive performance against relevant baselines.

cs.LG↗

arXiv2023

Implementation and Evaluation of a Machine Learned Mesoscale Eddy Parameterization into a Numerical Ocean Circulation Model

We address the question of how to use a machine learned parameterization in a general circulation model, and assess its performance both computationally and physically. We take one particular machine learned parameterization \cite{Guillaumin1&Zanna-JAMES21} and evaluate the online performance in a different model from which it was previously tested. This parameterization is a deep convolutional network that predicts parameters for a stochastic model of subgrid momentum forcing by mesoscale eddies. We treat the parameterization as we would a conventional parameterization once implemented in the numerical model. This includes trying the parameterization in a different flow regime from that in which it was trained, at different spatial resolutions, and with other differences, all to test generalization. We assess whether tuning is possible, which is a common practice in general circulation model development. We find the parameterization, without modification or special treatment, to be stable and that the action of the parameterization to be diminishing as spatial resolution is refined. We also find some limitations of the machine learning model in implementation: 1) tuning of the outputs from the parameterization at various depths is necessary; 2) the forcing near boundaries is not predicted as well as in the open ocean; 3) the cost of the parameterization is prohibitively high on CPUs. We discuss these limitations, present some solutions to problems, and conclude that this particular ML parameterization does inject energy, and improve backscatter, as intended but it might need further refinement before we can use it in production mode in contemporary climate models.

physics.ao-ph↗

arXiv2023

Azimuthal asymmetries of muon pair production in ultraperipheral heavy ion collisions

In this paper we study azimuthal asymmetries of the muon pair production in ultraperipheral heavy ion collisions within the joint impact parameter and transverse momentum dependent framework. The final state QED radiation effects are resummed to all orders in perturbation theory, where the complete muon mass corrections are also taken into account. We further make numerical estimations for azimuthal asymmetries in the different kinematic regions accessible at RHIC and LHC with the derived resummation formula. We find that the lepton mass effects can give sizable corrections to the asymmetries at relatively large pair transverse momentum at RHIC energy.

hep-ph↗

arXiv2023

Learnable Topological Features for Phylogenetic Inference via Graph Neural Networks

Structural information of phylogenetic tree topologies plays an important role in phylogenetic inference. However, finding appropriate topological structures for specific phylogenetic inference tasks often requires significant design effort and domain expertise. In this paper, we propose a novel structural representation method for phylogenetic inference based on learnable topological features. By combining the raw node features that minimize the Dirichlet energy with modern graph representation learning techniques, our learnable topological features can provide efficient structural information of phylogenetic trees that automatically adapts to different downstream tasks without requiring domain expertise. We demonstrate the effectiveness and efficiency of our method on a simulated data tree probability estimation task and a benchmark of challenging real data variational Bayesian phylogenetic inference problems.

stat.ML↗

arXiv2023

Pointwise Weyl Laws for Schrödinger operators with singular potentials

We consider the Schrödinger operators $H_V=-Δ_g+V$ with singular potentials $V$ on general $n$-dimensional Riemannian manifolds and study whether various forms of pointwise Weyl law remain valid under this pertubation. We prove that the pointwise Weyl law holds for potentials in the Kato class, which is the minimal assumption to ensure that $H_V$ is essentially self-adjoint and bounded from below or has favorable heat kernel bounds. Moreover, we show that the pointwise Weyl law with the standard sharp error term $O(λ^{n-1})$ holds for potentials in $L^n(M)$.

math.AP↗

arXiv2023

Sharp Pointwise Weyl Laws for Schrödinger Operators with Singular Potentials on Flat Tori

The Weyl law of the Laplacian on the flat torus $\mathbb{T}^n$ is concerning the number of eigenvalues $\leλ^2$, which is equivalent to counting the lattice points inside the ball of radius $λ$ in $\mathbb{R}^n$. The leading term in the Weyl law is $c_nλ^n$, while the sharp error term $O(λ^{n-2})$ is only known in dimension $n\ge5$. Determining the sharp error term in lower dimensions is a famous open problem (e.g. Gauss circle problem). In this paper, we show that under a type of singular perturbations one can obtain the pointwise Weyl law with a sharp error term in any dimensions. Moreover, this result verifies the sharpness of the general theorems for the Schrödinger operators $H_V=-Δ_{g}+V$ in the previous work of the authors, and extends the 3-dimensional results of Frank-Sabin to any dimensions.

math.AP↗

arXiv2023

Causal-Discovery Performance of ChatGPT in the context of Neuropathic Pain Diagnosis

ChatGPT has demonstrated exceptional proficiency in natural language conversation, e.g., it can answer a wide range of questions while no previous large language models can. Thus, we would like to push its limit and explore its ability to answer causal discovery questions by using a medical benchmark (Tu et al. 2019) in causal discovery.

cs.CL↗

arXiv2023

Azimuthal decorrelation for photon induced dijet production in ultra-peripheral collisions of heavy ions

We study the azimuthal angular decorrelation of the dijet production via photon fusion in ultra-peripheral heavy ion collisions. The impact parameter dependent cross section of quark-antiquark pairs production is derived using the equivalent photon approximation, and the contribution from final-state QCD radiations to the azimuthal angular distribution are calculated within Soft-Collinear Effective Theory. We carry out the QCD resummation of large logarithms of the azimuthal angle as well as the jet radius at next-to-leading logarithmic accuracy. In the end we present the normalized differential cross section for azimuthal decorrelation of the dijet pair and find that our results are consistent with the measurements reported by the ATLAS collaboration.

hep-ph↗

arXiv2023

Nonlinear Schrödinger equation on the half-line without a conserved number of solitons

We explore the phenomena of absorption/emission of solitons by an integrable boundary for the nonlinear Schrödinger equation on the half-line. This is based on the investigation of time-dependent reflection matrices which satisfy the boundary zero curvature equation. In particular, this leads to absorption/emission processes at the boundary that can take place for solitons and higher-order solitons. As a consequence, the usual charges on the half-line are no longer conserved but we show explicitly how to restore an infinite set of conserved quantities by taking the boundary into account. The Hamiltonian description and Poisson structure of the model are presented, which allows us to derive for the first time a classical version of the boundary algebra used originally in the context of the quantum nonlinear Schrödinger equation.

nlin.SI↗

arXiv2022

Sharp $L^p$ estimates and size of nodal sets for generalized Steklov eigenfunctions

We prove sharp $L^p$ estimates for the Steklov eigenfunctions on compact manifolds with boundary in terms of their $L^2$ norms on the boundary. We prove it by establishing $L^p$ bounds for the harmonic extension operators as well as the spectral projection operators on the boundary. Moreover, we derive lower bounds on the size of nodal sets for a variation of the Steklov spectral problem. We consider a generalized version of the Steklov problem by adding a non-smooth potential on the boundary but some of our results are new even without potential.

math.AP↗

arXiv2022

Gravitational-wave bursts from spin-precessing black holes in binary systems

Gravitational waves from precessing binary black holes exhibit new features that are absent in non-precessionary systems. All current waveform models take into account only the modulation of the signal due to precession. In this letter, we find that this effect has its own signature, by gravitational emission of a short and transient signal, or burst. The frequency of the burst is comparable to that of the late stage of the inspiral. We show that under certain conditions, this signal is strong enough to be detected by Advanced LIGO. For third-generation detectors like the Einstein telescope, the calculated signal-to-noise ratio can reach higher values. Measurements of precession would provide valuable insights into the intrinsic structure of black holes, and therefore into astrophysical binary formation mechanisms.

gr-qc↗

arXiv2022

WSC-Trans: A 3D network model for automatic multi-structural segmentation of temporal bone CT

Cochlear implantation is currently the most effective treatment for patients with severe deafness, but mastering cochlear implantation is extremely challenging because the temporal bone has extremely complex and small three-dimensional anatomical structures, and it is important to avoid damaging the corresponding structures when performing surgery. The spatial location of the relevant anatomical tissues within the target area needs to be determined using CT prior to the procedure. Considering that the target structures are too small and complex, the time required for manual segmentation is too long, and it is extremely challenging to segment the temporal bone and its nearby anatomical structures quickly and accurately. To overcome this difficulty, we propose a deep learning-based algorithm, a 3D network model for automatic segmentation of multi-structural targets in temporal bone CT that can automatically segment the cochlea, facial nerve, auditory tubercle, vestibule and semicircular canal. The algorithm combines CNN and Transformer for feature extraction and takes advantage of spatial attention and channel attention mechanisms to further improve the segmentation effect, the experimental results comparing with the results of various existing segmentation algorithms show that the dice similarity scores, Jaccard coefficients of all targets anatomical structures are significantly higher while HD95 and ASSD scores are lower, effectively proving that our method outperforms other advanced methods.

eess.IV↗

arXiv2022

Discovering the Unknown Knowns: Turning Implicit Knowledge in the Dataset into Explicit Training Examples for Visual Question Answering

Visual question answering (VQA) is challenging not only because the model has to handle multi-modal information, but also because it is just so hard to collect sufficient training examples -- there are too many questions one can ask about an image. As a result, a VQA model trained solely on human-annotated examples could easily over-fit specific question styles or image contents that are being asked, leaving the model largely ignorant about the sheer diversity of questions. Existing methods address this issue primarily by introducing an auxiliary task such as visual grounding, cycle consistency, or debiasing. In this paper, we take a drastically different approach. We found that many of the "unknowns" to the learned VQA model are indeed "known" in the dataset implicitly. For instance, questions asking about the same object in different images are likely paraphrases; the number of detected or annotated objects in an image already provides the answer to the "how many" question, even if the question has not been annotated for that image. Building upon these insights, we present a simple data augmentation pipeline SimpleAug to turn this "known" knowledge into training examples for VQA. We show that these augmented examples can notably improve the learned VQA models' performance, not only on the VQA-CP dataset with language prior shifts but also on the VQA v2 dataset without such shifts. Our method further opens up the door to leverage weakly-labeled or unlabeled images in a principled way to enhance VQA models. Our code and data are publicly available at https://github.com/heendung/simpleAUG.

cs.CV↗

arXiv2022

Rhino: Deep Causal Temporal Relationship Learning With History-dependent Noise

Discovering causal relationships between different variables from time series data has been a long-standing challenge for many domains such as climate science, finance, and healthcare. Given the complexity of real-world relationships and the nature of observations in discrete time, causal discovery methods need to consider non-linear relations between variables, instantaneous effects and history-dependent noise (the change of noise distribution due to past actions). However, previous works do not offer a solution addressing all these problems together. In this paper, we propose a novel causal relationship learning framework for time-series data, called Rhino, which combines vector auto-regression, deep learning and variational inference to model non-linear relationships with instantaneous effects while allowing the noise distribution to be modulated by historical observations. Theoretically, we prove the structural identifiability of Rhino. Our empirical results from extensive synthetic experiments and two real-world benchmarks demonstrate better discovery performance compared to relevant baselines, with ablation studies revealing its robustness under model misspecification.

cs.LG↗