arXiv subjects

Jingyi Zhou

Publications and source records attributed to Jingyi Zhou.

25 records · Page 2Linked to original sources

arXiv2024

Exploring Multi-Timestep Multi-Stage Diffusion Features for Hyperspectral Image Classification

The effectiveness of spectral-spatial feature learning is crucial for the hyperspectral image (HSI) classification task. Diffusion models, as a new class of groundbreaking generative models, have the ability to learn both contextual semantics and textual details from the distinct timestep dimension, enabling the modeling of complex spectral-spatial relations in HSIs. However, existing diffusion-based HSI classification methods only utilize manually selected single-timestep single-stage features, limiting the full exploration and exploitation of rich contextual semantics and textual information hidden in the diffusion model. To address this issue, we propose a novel diffusion-based feature learning framework that explores Multi-Timestep Multi-Stage Diffusion features for HSI classification for the first time, called MTMSD. Specifically, the diffusion model is first pretrained with unlabeled HSI patches to mine the connotation of unlabeled data, and then is used to extract the multi-timestep multi-stage diffusion features. To effectively and efficiently leverage multi-timestep multi-stage features,two strategies are further developed. One strategy is class & timestep-oriented multi-stage feature purification module with the inter-class and inter-timestep prior for reducing the redundancy of multi-stage features and alleviating memory constraints. The other one is selective timestep feature fusion module with the guidance of global features to adaptively select different timestep features for integrating texture and semantics. Both strategies facilitate the generality and adaptability of the MTMSD framework for diverse patterns of different HSI data. Extensive experiments are conducted on four public HSI datasets, and the results demonstrate that our method outperforms state-of-the-art methods for HSI classification, especially on the challenging Houston 2018 dataset.

cs.CV↗

arXiv2023

A Soft Contrastive Learning-based Prompt Model for Few-shot Sentiment Analysis

Few-shot text classification has attracted great interest in both academia and industry due to the lack of labeled data in many fields. Different from general text classification (e.g., topic classification), few-shot sentiment classification is more challenging because the semantic distances among the classes are more subtle. For instance, the semantic distances between the sentiment labels in a positive or negative polarity (e.g., ``love" and ``joy", ``remorse" and ``sadness") are close, while the distances are large for the sentiment labels in two opposite polarities (e.g., ``love" and ``sadness"). To address this problem, we propose a Soft Contrastive learning-based Prompt (\texttt{SCP}) model for few-shot sentiment analysis. First, we design a sentiment-aware chain of thought prompt module to guide the model to predict the sentiment from coarse grain to fine grain via a series of intermediate reasoning steps. Then, we propose a soft contrastive learning algorithm to take the correlation of the labels into account. A series of experiments on several sentiment analysis datasets show the great advantages of \texttt{SCP} by comparing it with SOTA baselines (e.g., ChatGPT).

cs.CL↗

arXiv2023

Lowest-order QED radiative corrections in unpolarized elastic electron-deuteron scattering beyond the ultra-relativistic limit for the proposed deuteron charge radius measurement at Jefferson Laboratory

Analogous to the well-known proton charge radius puzzle, a similar puzzle exists for the deuteron charge radius, $r_{d}$. There are discrepancies observed in the results of $r_{d}$, measured from electron-deuteron ($e-d$) scattering experiments, as well as from atomic spectroscopy. In order to help resolve the charge radius puzzle of the deuteron, the PRad collaboration at Jefferson Lab has proposed an experiment for measuring $r_{d}$, named DRad. This experiment is designed to measure the unpolarized elastic $e-d$ scattering cross section in a low-$Q^{2}$ region. To extract the cross section with a high precision, having reliable knowledge of QED radiative corrections is important. In this paper, we present complete numerical calculations of the lowest-order radiative corrections in $e-d$ scattering for the DRad kinematics. The calculations have been performed within a covariant formalism and beyond the ultra-relativistic approximation ($m_{e}^{2} \ll Q^{2}$). Besides, we present a systematic uncertainty on $r_{d}$ arising from higher-order radiative corrections, estimated based on our cross-section results.

nucl-th↗

arXiv2022

Understanding the systematic differences in extractions of the proton electric form factors at low-$Q^2$

Systematic differences exist between values of the proton's electric form factors in the low-$Q^2$ region extracted by different experimental and theoretical groups, though they are all making use of basically the same electron-proton scattering data. To try understand the source of these differences, we make use of the analytically well-behaved rational (N=1, M=1) function, a predictive function that can be reasonably used for extrapolations at $Q^{2} \rightarrow 0$. First, we test how well this deceptively simple two-parameter function describes the extremely complex and state-of-the-art dispersively improved chiral effective field theory calculations. Second, we carry out a complete re-analysis of the 34 sets of eletron-proton elastic scattering cross-section data of the Mainz A1 Collaboration with its unconstrained 31 normalization parameters up to $Q^{2} = 0.5~{\rm (GeV/c)^{2}}$. We find that subtle shifts in the normalization parameters can result in relatively large changes in the extracted physical qualities. In conclusion, we show that by simply using a well-behaved analytic function, the apparent discrepancy between recent form-factor extractions can be resolved.

hep-ph↗

arXiv2022

Union-net: A deep neural network model adapted to small data sets

In real applications, generally small data sets can be obtained. At present, most of the practical applications of machine learning use classic models based on big data to solve the problem of small data sets. However, the deep neural network model has complex structure, huge model parameters, and training requires more advanced equipment, which brings certain difficulties to the application. Therefore, this paper proposes the concept of union convolution, designing a light deep network model union-net with a shallow network structure and adapting to small data sets. This model combines convolutional network units with different combinations of the same input to form a union module. Each union module is equivalent to a convolutional layer. The serial input and output between the 3 modules constitute a "3-layer" neural network. The output of each union module is fused and added as the input of the last convolutional layer to form a complex network with a 4-layer network structure. It solves the problem that the deep network model network is too deep and the transmission path is too long, which causes the loss of the underlying information transmission. Because the model has fewer model parameters and fewer channels, it can better adapt to small data sets. It solves the problem that the deep network model is prone to overfitting in training small data sets. Use the public data sets cifar10 and 17flowers to conduct multi-classification experiments. Experiments show that the Union-net model can perform well in classification of large data sets and small data sets. It has high practical value in daily application scenarios. The model code is published at https://github.com/yeaso/union-net

cs.CV↗

arXiv2021

Elastic Positron-Proton Scattering at Low Q$^2$

Systematic differences in the the proton's charge radius, as determined by ordinary atoms and muonic atoms, have caused a resurgence of interest in elastic lepton scattering measurements. The proton's charge radius, defined as the slope of the charge form factor at Q$^2$=0, does not depend on the probe. Any difference in the apparent size of the proton, when determined from ordinary versus muonic hydrogen, could point to new physics or need for the higher order corrections. While recent measurements seem to now be in agreement, there is to date no high precision elastic scattering data with both electrons and positrons. A high precision proton radius measurement could be performed in Hall B at Jefferson Lab with a positron beam and the calorimeter based setup of the PRad experiment. This measurement could also be extended to deuterons where a similar discrepancy has been observed between the muonic and electronic determination of deuteron charge radius. A new, high precision measurement with positrons, when viewed alongside electron scattering measurements and the forthcoming MUSE muon scattering measurement, could help provide new insights into the origins of the proton radius puzzle, and also provide new experimental constraints on radiative correction calculations.

nucl-ex↗

arXiv2021

Advanced extraction of the deuteron charge radius from electron-deuteron scattering data

To extract the charge radius of the proton, $r_{p}$, from the electron scattering data, the PRad collaboration at Jefferson Lab has developed a rigorous framework for finding the best functional forms - the fitters - for a robust extraction of $r_{p}$ from a wide variety of sample functions for the range and uncertainties of the PRad data. In this paper we utilize and further develop this framework. Herein we discuss methods for searching for the best fitter candidates as well as a procedure for testing the robustness of extraction of the deuteron charge radius, $r_{d}$, from parametrizations based on elastic electron-deuteron scattering data. The ansatz proposed in this paper for the robust extraction of $r_{d}$, for the proposed low-$Q^{2}$ DRad experiment at Jefferson Lab, can be further improved once there are more data.

nucl-ex↗