Searcharxiv⌕ Search

arXiv subjects

Yi Zhang

Publications and source records attributed to Yi Zhang.

At least 811 records · Page 45Linked to original sources
arXiv2023

NatCS: Eliciting Natural Customer Support Dialogues

Despite growing interest in applications based on natural customer support conversations, there exist remarkably few publicly available datasets that reflect the expected characteristics of conversations in these settings. Existing task-oriented dialogue datasets, which were collected to benchmark dialogue systems mainly in written human-to-bot settings, are not representative of real customer support conversations and do not provide realistic benchmarks for systems that are applied to natural data. To address this gap, we introduce NatCS, a multi-domain collection of spoken customer service conversations. We describe our process for collecting synthetic conversations between customers and agents based on natural language phenomena observed in real conversations. Compared to previous dialogue datasets, the conversations collected with our approach are more representative of real human-to-human conversations along multiple metrics. Finally, we demonstrate potential uses of NatCS, including dialogue act classification and intent induction from conversations as potential applications, showing that dialogue act annotations in NatCS provide more effective training data for modeling real conversations compared to existing synthetic written datasets. We publicly release NatCS to facilitate research in natural dialog systems

cs.CL↗
arXiv2023

Intent Induction from Conversations for Task-Oriented Dialogue Track at DSTC 11

With increasing demand for and adoption of virtual assistants, recent work has investigated ways to accelerate bot schema design through the automatic induction of intents or the induction of slots and dialogue states. However, a lack of dedicated benchmarks and standardized evaluation has made progress difficult to track and comparisons between systems difficult to make. This challenge track, held as part of the Eleventh Dialog Systems Technology Challenge, introduces a benchmark that aims to evaluate methods for the automatic induction of customer intents in a realistic setting of customer service interactions between human agents and customers. We propose two subtasks for progressively tackling the automatic induction of intents and corresponding evaluation methodologies. We then present three datasets suitable for evaluating the tasks and propose simple baselines. Finally, we summarize the submissions and results of the challenge track, for which we received submissions from 34 teams.

cs.CL↗
arXiv2023

NeRF applied to satellite imagery for surface reconstruction

We present Surf-NeRF, a modified implementation of the recently introduced Shadow Neural Radiance Field (S-NeRF) model. This method is able to synthesize novel views from a sparse set of satellite images of a scene, while accounting for the variation in lighting present in the pictures. The trained model can also be used to accurately estimate the surface elevation of the scene, which is often a desirable quantity for satellite observation applications. S-NeRF improves on the standard Neural Radiance Field (NeRF) method by considering the radiance as a function of the albedo and the irradiance. Both these quantities are output by fully connected neural network branches of the model, and the latter is considered as a function of the direct light from the sun and the diffuse color from the sky. The implementations were run on a dataset of satellite images, augmented using a zoom-and-crop technique. A hyperparameter study for NeRF was carried out, leading to intriguing observations on the model's convergence. Finally, both NeRF and S-NeRF were run until 100k epochs in order to fully fit the data and produce their best possible predictions. The code related to this article can be found at https://github.com/fsemerar/surfnerf.

cs.CV↗
arXiv2023

Sparks of Artificial General Intelligence: Early experiments with GPT-4

Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. The latest model developed by OpenAI, GPT-4, was trained using an unprecedented scale of compute and data. In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT-4 is part of a new cohort of LLMs (along with ChatGPT and Google's PaLM for example) that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4's capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system. In our exploration of GPT-4, we put special emphasis on discovering its limitations, and we discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI, including the possible need for pursuing a new paradigm that moves beyond next-word prediction. We conclude with reflections on societal influences of the recent technological leap and future research directions.

cs.CL↗
arXiv2023

Single-shot quantum measurements sketch quantum many-body states

Quantum measurements are our eyes to the quantum systems consisting of a multitude of microscopic degrees of freedom. However, the intrinsic uncertainty of quantum measurements and the exponentially large Hilbert space pose natural barriers to simple interpretations of the measurement outcomes. We propose a nonlinear "measurement energy" based upon the measurement outcomes and an iterative effective-Hamiltonian approach to extract the most probable states (maximum likelihood estimates) in an efficient and general fashion, thus reconciling the non-commuting observables and getting more out of the quantum measurements. We showcase the versatility and accuracy of our perspective on random long-range fermion models and Kitaev quantum spin liquid models, where smoking-gun signatures were lacking. Our study also paves the way towards concepts such as nonlinear-operator Hamiltonian and applications such as parent Hamiltonian reconstruction.

cond-mat.str-el↗
arXiv2023

Prospects for detection rate of very-high-energy γ-ray emissions from short γ-ray bursts with the HADAR experiment

The observation of short gamma ray bursts (SGRBs) in the TeV energy range plays an important role in understanding the radiation mechanism and probing new areas of physics such as Lorentz invariance violation. However, no SGRB has been observed in this energy range due to the short duration of SGRBs and the weakness of current experiments. New experiments with new technology are required to detect sub-TeV SGRBs. In this work, we observe the very high energy (VHE) $γ$-ray emissions from SGRBs and calculate the annual detection rate with the High Altitude Detection of Astronomical Radiation HADAR (HADAR) experiment. First, a set of pseudo-SGRB samples is generated and checked using the observations of Fermi-GBM, Fermi-LAT, and SWIFT measurements. The annual detection rate is calculated from these SGRB samples based on the performance of the HADAR instrument. As a result, the HADAR experiment can detect 0.5 SGRB per year if the spectral break-off of $γ$-rays caused by the internal absorption is larger than 100 GeV. For a GRB09010-like GRB in HADAR's view, it should be possible to detect approximately 2000 photons considering the internal absorption. With a time delay assumption due to the Lorentz invariance violation effects, a simulated light curve of GRB090510 has evident energy dependence. We hope that the HADAR experiment can perform the SGRB observations and test our calculations in the future.

astro-ph.HE↗
arXiv2023

Physics-/Model-Based and Data-Driven Methods for Low-Dose Computed Tomography: A survey

Since 2016, deep learning (DL) has advanced tomographic imaging with remarkable successes, especially in low-dose computed tomography (LDCT) imaging. Despite being driven by big data, the LDCT denoising and pure end-to-end reconstruction networks often suffer from the black box nature and major issues such as instabilities, which is a major barrier to apply deep learning methods in low-dose CT applications. An emerging trend is to integrate imaging physics and model into deep networks, enabling a hybridization of physics/model-based and data-driven elements. %This type of hybrid methods has become increasingly influential. In this paper, we systematically review the physics/model-based data-driven methods for LDCT, summarize the loss functions and training strategies, evaluate the performance of different methods, and discuss relevant issues and future directions.

eess.IV↗
arXiv2023

Frequency-Modulated Point Cloud Rendering with Easy Editing

We develop an effective point cloud rendering pipeline for novel view synthesis, which enables high fidelity local detail reconstruction, real-time rendering and user-friendly editing. In the heart of our pipeline is an adaptive frequency modulation module called Adaptive Frequency Net (AFNet), which utilizes a hypernetwork to learn the local texture frequency encoding that is consecutively injected into adaptive frequency activation layers to modulate the implicit radiance signal. This mechanism improves the frequency expressive ability of the network with richer frequency basis support, only at a small computational budget. To further boost performance, a preprocessing module is also proposed for point cloud geometry optimization via point opacity estimation. In contrast to implicit rendering, our pipeline supports high-fidelity interactive editing based on point cloud manipulation. Extensive experimental results on NeRF-Synthetic, ScanNet, DTU and Tanks and Temples datasets demonstrate the superior performances achieved by our method in terms of PSNR, SSIM and LPIPS, in comparison to the state-of-the-art.

cs.CV↗
arXiv2023

Self-driven photo-polarized water molecule triggered graphene-based photodetector

Flowing water can be used as an energy source for generators, providing a major part of the energy for daily life. However, water is rarely used for information or electronic devices. Herein, we present the feasibility of a polarized liquid-triggered photodetector in which polarized water is sandwiched between graphene and semiconductor. Due to the polarization and depolarization processes of water molecular under the driven of photogenerated carriers, a photo-sensitive current can be repeatably produced, resulting in a high-performance photodetector. The response wavelength of the photodetector can be finely tuned as a result of the free choice of semiconductors as there is no requirement of lattice match between graphene and the semiconductors. Under zero voltage bias, the responsivity and specific detectivity of Gr/NaCl (0.5 M)/N-GaN reaches values of 130.7 mA/W and 2.8*10^12 Jones under 350 nm illumination. Meanwhile, using a polar liquid photodetector can successfully read the photoplethysmography signals to produce accurate oxygen blood saturation and heart rate. Compared with the commercial pulse oximetry sensor, the average errors of oxygen saturation and heart rate in the designed photoplethysmography sensor are~1.9% and ~2.1%, respectively. This study transfers the message that water can be used as high-performance photodetector in informative industries.

physics.app-ph↗
arXiv2023

Anatomy of open-boundary bulk in multiband non-Hermitian systems

Although the non-Bloch band theory is a milestone in elaborating bulk energy bands of non-Hermitian systems under the open-boundary condition (OBC), vital issues related to multivalued functions of non-Hermitian energy bands remain unsolved. In this paper, we anatomize the bulk properties of one-dimensional multiband non-Hermitian systems under OBC. We put forward the energy-band branches (EBBs) to settle the multivalued functions of non-Hermitian energy bands, which become gapped or gapless corresponding to disconnected or connected EBBs in the complex energy plane, where the branch points and branch cuts play a crucial role. We clarify the precise significance of the non-Hermitian skin effect, which illustrates the asymptotic behavior of EBB eigenstates (bulk eigenstates) in the deep bulk and compensates previous non-Bloch band theory. We also obtain a general expression of open-boundary Green's functions based on such EBBs and generalized Brillouin zones, useful for studies on non-Hermitian dynamical evolution.

cond-mat.mes-hall↗
arXiv2023

Discontinuous Galerkin Methods for an Elliptic Optimal Control Problem with a General State Equation and Pointwise State Constraints

We investigate discontinuous Galerkin methods for an elliptic optimal control problem with a general state equation and pointwise state constraints on general polygonal domains. We show that discontinuous Galerkin methods for general second-order elliptic boundary value problems can be used to solve the elliptic optimal control problems with pointwise state constraints. We establish concrete error estimates and numerical experiments are shown to support the theoretical results.

math.NA↗
arXiv2023

Prediction of transonic flow over supercritical airfoils using geometric-encoding and deep-learning strategies

The Reynolds-averaged Navier-Stokes equation for compressible flow over supercritical airfoils under various flow conditions must be rapidly and accurately solved to shorten design cycles for such airfoils. Although deep-learning methods can effectively predict flow fields, the accuracy of these predictions near sensitive regions and their generalizability to large-scale datasets in engineering applications must be enhanced. In this study, a modified vision transformer-based encoder-decoder network is designed for the prediction of transonic flow over supercritical airfoils. In addition, four methods are designed to encode the geometric input with various information points and the performances of these methods are compared. The statistical results show that these methods generate accurate predictions over the complete flow field, with a mean absolute error on the order of 1e-4. To increase accuracy near the shock area, multilevel wavelet transformation and gradient distribution losses are introduced into the loss function. This results in the maximum error that is typically observed near the shock area decreasing by 50%. Furthermore, the models are pretrained through transfer learning on large-scale datasets and finetuned on small datasets to improve their generalizability in engineering applications. The results generated by various pretrained models demonstrate that transfer learning yields a comparable accuracy from a reduced training time.

physics.flu-dyn↗
arXiv2023

KBNet: Kernel Basis Network for Image Restoration

How to aggregate spatial information plays an essential role in learning-based image restoration. Most existing CNN-based networks adopt static convolutional kernels to encode spatial information, which cannot aggregate spatial information adaptively. Recent transformer-based architectures achieve adaptive spatial aggregation. But they lack desirable inductive biases of convolutions and require heavy computational costs. In this paper, we propose a kernel basis attention (KBA) module, which introduces learnable kernel bases to model representative image patterns for spatial information aggregation. Different kernel bases are trained to model different local structures. At each spatial location, they are linearly and adaptively fused by predicted pixel-wise coefficients to obtain aggregation weights. Based on the KBA module, we further design a multi-axis feature fusion (MFF) block to encode and fuse channel-wise, spatial-invariant, and pixel-adaptive features for image restoration. Our model, named kernel basis network (KBNet), achieves state-of-the-art performances on more than ten benchmarks over image denoising, deraining, and deblurring tasks while requiring less computational cost than previous SOTA methods.

cs.CV↗
arXiv2023

Augmented smartphone bilirubinometer enabled by a mobile app that turns smartphone into multispectral imager

We present the development of SpeCamX, a mobile application that transforms any unmodified smartphone into a powerful multispectral imager capable of capturing multispectral information. Our application includes an augmented bilirubinometer, enabling accurate prediction of blood bilirubin levels (BBL). In a clinical study involving 320 patients with liver diseases, we used SpeCamX to image the bulbar conjunctiva region, and we employed a hybrid machine learning prediction model to predict BBL. We observed a high correlation with blood test results, demonstrating the efficacy of our approach. Furthermore, we compared our method, which uses spectrally augmented learning (SAL), with traditional learning based on RGB photographs (RGBL), and our results clearly indicate that SpeCamX outperforms RGBL in terms of prediction accuracy, efficiency, and stability. This study highlights the potential of SpeCamX to improve the prediction of bio-chromophores, and its ability to transform an ordinary smartphone into a powerful medical tool without the need for additional investments or expertise. This makes it suitable for widespread use, particularly in areas where medical resources are scarce.

eess.IV↗
arXiv2023

Independence-Encouraging Subsampling for Nonparametric Additive Models

The additive model is a popular nonparametric regression method due to its ability to retain modeling flexibility while avoiding the curse of dimensionality. The backfitting algorithm is an intuitive and widely used numerical approach for fitting additive models. However, its application to large datasets may incur a high computational cost and is thus infeasible in practice. To address this problem, we propose a novel approach called independence-encouraging subsampling (IES) to select a subsample from big data for training additive models. Inspired by the minimax optimality of an orthogonal array (OA) due to its pairwise independent predictors and uniform coverage for the range of each predictor, the IES approach selects a subsample that approximates an OA to achieve the minimax optimality. Our asymptotic analyses demonstrate that an IES subsample converges to an OA and that the backfitting algorithm over the subsample converges to a unique solution even if the predictors are highly dependent in the original big data. The proposed IES method is also shown to be numerically appealing via simulations and a real data application.

stat.ME↗
arXiv2023

Rethinking Noise Synthesis and Modeling in Raw Denoising

The lack of large-scale real raw image denoising dataset gives rise to challenges on synthesizing realistic raw image noise for training denoising models. However, the real raw image noise is contributed by many noise sources and varies greatly among different sensors. Existing methods are unable to model all noise sources accurately, and building a noise model for each sensor is also laborious. In this paper, we introduce a new perspective to synthesize noise by directly sampling from the sensor's real noise. It inherently generates accurate raw image noise for different camera sensors. Two efficient and generic techniques: pattern-aligned patch sampling and high-bit reconstruction help accurate synthesis of spatial-correlated noise and high-bit noise respectively. We conduct systematic experiments on SIDD and ELD datasets. The results show that (1) our method outperforms existing methods and demonstrates wide generalization on different sensors and lighting conditions. (2) Recent conclusions derived from DNN-based noise modeling methods are actually based on inaccurate noise parameters. The DNN-based methods still cannot outperform physics-based statistical methods.

eess.IV↗
arXiv2023

Collaboration-Aware Graph Convolutional Network for Recommender Systems

Graph Neural Networks (GNNs) have been successfully adopted in recommender systems by virtue of the message-passing that implicitly captures collaborative effect. Nevertheless, most of the existing message-passing mechanisms for recommendation are directly inherited from GNNs without scrutinizing whether the captured collaborative effect would benefit the prediction of user preferences. In this paper, we first analyze how message-passing captures the collaborative effect and propose a recommendation-oriented topological metric, Common Interacted Ratio (CIR), which measures the level of interaction between a specific neighbor of a node with the rest of its neighbors. After demonstrating the benefits of leveraging collaborations from neighbors with higher CIR, we propose a recommendation-tailored GNN, Collaboration-Aware Graph Convolutional Network (CAGCN), that goes beyond 1-Weisfeiler-Lehman(1-WL) test in distinguishing non-bipartite-subgraph-isomorphic graphs. Experiments on six benchmark datasets show that the best CAGCN variant outperforms the most representative GNN-based recommendation model, LightGCN, by nearly 10% in Recall@20 and also achieves around 80% speedup. Our code is publicly available at https://github.com/YuWVandy/CAGCN.

cs.IR↗
arXiv2023

Unveiling Transformers with LEGO: a synthetic reasoning task

We propose a synthetic reasoning task, LEGO (Learning Equality and Group Operations), that encapsulates the problem of following a chain of reasoning, and we study how the Transformer architectures learn this task. We pay special attention to data effects such as pretraining (on seemingly unrelated NLP tasks) and dataset composition (e.g., differing chain length at training and test time), as well as architectural variants such as weight-tied layers or adding convolutional components. We study how the trained models eventually succeed at the task, and in particular, we manage to understand some of the attention heads as well as how the information flows in the network. In particular, we have identified a novel \emph{association} pattern that globally attends only to identical tokens. Based on these observations we propose a hypothesis that here pretraining helps for LEGO tasks due to certain structured attention patterns, and we experimentally verify this hypothesis. We also observe that in some data regime the trained transformer finds ``shortcut" solutions to follow the chain of reasoning, which impedes the model's robustness, and moreover we propose ways to prevent it. Motivated by our findings on structured attention patterns, we propose the LEGO attention module, a drop-in replacement for vanilla attention heads. This architectural change significantly reduces Flops and maintains or even \emph{improves} the model's performance at large-scale pretraining.

cs.LG↗