Author name cluster

Jing Bai

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

12 papers

2 author rows

EAAI Journal 2026 Journal Article

A cross-scale interaction framework combining Mamba and Convolutional Neural Networks for Arbitrary-Scale Super-Resolution of infrared images

Feiwei Qin
Xinyu Cao
Changmiao Wang
Kai Zhang
Yong Peng
Jing Bai

Infrared image super-resolution (SR) is a critical technique in aerospace and remote sensing. However, current methods frequently ignore the continuity of real-world signals in favor of discrete representations of infrared images. As a result, these methods necessitate training a separate model for each scale, leading to inefficient training and increased storage requirements. To overcome this limitation, we propose a Mamba-Convolutional Neural Network Cross-Scale Interaction Arbitrary-Scale Super-Resolution (MCASSR) framework, the first arbitrary-scale super-resolution method for infrared images capable of achieving SR across multiple continuous scales with a single model. Specifically, we design a Mamba-Convolutional Neural Network Mutual Learning Backbone (MCMLB) for deep feature extraction and a Compound Implicit Attention Upsampler (CIAU) for continuous upsampling. The MCMLB effectively integrates the long-range dependency modeling of Mamba and the local detail capturing of Convolutional Neural Network through bidirectional feature interaction. The CIAU enables adaptive continuous feature interpolation based on relative position and feature similarity while effectively leveraging cross-scale non-local texture priors in infrared images. Furthermore, we introduce a gradual residual decoder to capture high-frequency details in infrared images efficiently by alleviating the inherent spectral bias of multilayer perceptrons. Extensive experimental results demonstrate that MCASSR provides superior visual improvements across multiple infrared image datasets compared to current state-of-the-art methods.

Details DOI

KER Journal 2025 Journal Article

A comprehensive survey on advertising click-through rate prediction algorithm

Jing Bai
Xinyu Geng
Jiaqi Deng
Zhen Xia
Hongxia Jiang
Guoqiang Yan
Jing Liang

Abstract Advertising click-through rate (CTR) prediction is a fundamental task in recommender systems, aimed at estimating the likelihood of users interacting with advertisements based on their historical behavior. This prediction process has evolved through two main stages: from traditional shallow interaction models to more advanced deep learning approaches. Shallow models typically operate at the level of individual features, failing to fully leverage the rich, multilevel information available across different feature sets, leading to less accurate predictions. In contrast, deep learning models exhibit superior feature representation and learning capabilities, enabling a more realistic simulation of user interactions and improving the accuracy of CTR prediction. This paper provides a comprehensive overview of CTR prediction algorithms in the context of recommender systems. The algorithms are categorized into two groups: shallow interactive models and deep learning-based prediction models, including deep neural networks, convolutional neural networks, recurrent neural networks, and graph neural networks. Additionally, this paper also discusses the advantages and disadvantages of the aforementioned algorithms, as well as the benchmark datasets and model evaluation methods used for CTR prediction. Finally, it identifies potential future research directions in this rapidly advancing field.

Details DOI

ICLR Conference 2025 Conference Paper

DreamBench++: A Human-Aligned Benchmark for Personalized Image Generation

Yuang Peng
Yuxin Cui
Haomiao Tang
Zekun Qi
Runpei Dong
Jing Bai
Chunrui Han
Zheng Ge

Personalized image generation holds great promise in assisting humans in everyday work and life due to its impressive function in creatively generating personalized content. However, current evaluations either are automated but misalign with humans or require human evaluations that are time-consuming and expensive. In this work, we present DreamBench++, a human-aligned benchmark that advanced multimodal GPT models automate. Specifically, we systematically design the prompts to let GPT be both human-aligned and self-aligned, empowered with task reinforcement. Further, we construct a comprehensive dataset comprising diverse images and prompts. By benchmarking 7 modern generative models, we demonstrate that \dreambench results in significantly more human-aligned evaluation, helping boost the community with innovative findings.

Details

JBHI Journal 2025 Journal Article

LGG-NeXt: A Next Generation CNN and Transformer Hybrid Model for the Diagnosis of Alzheimer's Disease Using 2D Structural MRI

Jing Bai
Zhengyang Zhang
Yue Yin
Weikang Jin
Talal Ahmed Ali Ali
Yong Xiong
Zhu Xiao

Incurable Alzheimer's disease (AD) plagues many elderly people and families. It is important to accurately diagnose and predict it at an early stage. However, the existing methods have shortcomings, such as inability to learn local and global information and the inability to extract effective features. In this paper, we propose a lightweight classification network Local and Global Graph ConvNeXt. This model has a hybrid architecture of convolutional neural network and Transformers. We build the Global NeXt Block and the Local NeXt Block to extract the local and global features of the structural magnetic resonance imaging (sMRI). These two blocks are optimized by adding global multilayer perceptron and locally grouped attention, respectively. Then, the features are fed into the pixel graph neural network to aggregate the valid pixel features using mask attention. In addition, we decoupled the loss by category to optimize the calculation of the loss. This method was tested on slices of the processed sMRI datasets from ADNI and achieved excellent performance. Our model achieves 95. 81% accuracy with fewer parameters and floating point operations per second (FLOPS) than other classical efficient models in the diagnosis of AD.

Details DOI

ICML Conference 2025 Conference Paper

Vulnerability-Aware Alignment: Mitigating Uneven Forgetting in Harmful Fine-Tuning

Liang Chen 0001
Xueting Han
Li Shen 0008
Jing Bai
Kam-Fai Wong

Harmful fine-tuning (HFT), performed directly on open-source LLMs or through Fine-tuning-as-a-Service, breaks safety alignment and poses significant threats. Existing methods aim to mitigate HFT risks by learning robust representation on alignment data or making harmful data unlearnable, but they treat each data sample equally, leaving data vulnerability patterns understudied. In this work, we reveal that certain subsets of alignment data are consistently more prone to forgetting during HFT across different fine-tuning tasks and exhibit lower robustness compared to other subsets. Inspired by these findings, we propose Vulnerability-Aware Alignment (VAA), which calculates data vulnerability, partitions data into "vulnerable" and "invulnerable" groups, and encourages balanced learning using a group distributionally robust optimization (Group DRO) framework. Specifically, VAA learns an adversarial sampler that samples examples from the currently underperforming group and then applies group-dependent adversarial perturbations to the data during training, aiming to encourage a balanced learning process across groups. Experiments across four fine-tuning tasks demonstrate that VAA significantly reduces harmful scores while preserving downstream task performance, outperforming state-of-the-art baselines.

Details

AAAI Conference 2024 Conference Paper

AdapterGNN: Parameter-Efficient Fine-Tuning Improves Generalization in GNNs

Shengrui Li
Xueting Han
Jing Bai

Fine-tuning pre-trained models has recently yielded remarkable performance gains in graph neural networks (GNNs). In addition to pre-training techniques, inspired by the latest work in the natural language fields, more recent work has shifted towards applying effective fine-tuning approaches, such as parameter-efficient fine-tuning (PEFT). However, given the substantial differences between GNNs and transformer-based models, applying such approaches directly to GNNs proved to be less effective. In this paper, we present a comprehensive comparison of PEFT techniques for GNNs and propose a novel PEFT method specifically designed for GNNs, called AdapterGNN. AdapterGNN preserves the knowledge of the large pre-trained model and leverages highly expressive adapters for GNNs, which can adapt to downstream tasks effectively with only a few parameters, while also improving the model's generalization ability. Extensive experiments show that AdapterGNN achieves higher performance than other PEFT methods and is the only one consistently surpassing full fine-tuning (outperforming it by 1.6% and 5.7% in the chemistry and biology domains respectively, with only 5% and 4% of its parameters tuned) with lower generalization gaps. Moreover, we empirically show that a larger GNN model can have a worse generalization ability, which differs from the trend observed in large transformer-based models. Building upon this, we provide a theoretical justification for PEFT can improve generalization of GNNs by applying generalization bounds. Our code is available at https://github.com/Lucius-lsr/AdapterGNN.

PDF Details DOI

AAAI Conference 2022 Conference Paper

Graph Pointer Neural Networks

Tianmeng Yang
Yujing Wang
Zhihan Yue
Yaming Yang
Yunhai Tong
Jing Bai

Graph Neural Networks (GNNs) have shown advantages in various graph-based applications. Most existing GNNs assume strong homophily of graph structure and apply permutation-invariant local aggregation of neighbors to learn a representation for each node. However, they fail to generalize to heterophilic graphs, where most neighboring nodes have different labels or features, and the relevant nodes are distant. Few recent studies attempt to address this problem by combining multiple hops of hidden representations of central nodes (i. e. , multi-hop-based approaches) or sorting the neighboring nodes based on attention scores (i. e. , rankingbased approaches). As a result, these approaches have some apparent limitations. On the one hand, multi-hop-based approaches do not explicitly distinguish relevant nodes from a large number of multi-hop neighborhoods, leading to a severe over-smoothing problem. On the other hand, ranking-based models do not joint-optimize node ranking with end tasks and result in sub-optimal solutions. In this work, we present Graph Pointer Neural Networks (GPNN) to tackle the challenges mentioned above. We leverage a pointer network to select the most relevant nodes from a large amount of multihop neighborhoods, which constructs an ordered sequence according to the relationship with the central node. 1D convolution is then applied to extract high-level features from the node sequence. The pointer-network-based ranker in GPNN is joint-optimized with other parts in an end-to-end manner. Extensive experiments are conducted on six public node classification datasets with heterophilic graphs. The results show that GPNN significantly improves the classification performance of state-of-the-art methods. In addition, analyses also reveal the privilege of the proposed GPNN in filtering out irrelevant neighbors and reducing over-smoothing.

PDF Details

NeurIPS Conference 2020 Conference Paper

Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting

Defu Cao
Yujing Wang
Juanyong Duan
Ce Zhang
Xia Zhu
Congrui Huang
Yunhai Tong
Bixiong Xu

Multivariate time-series forecasting plays a crucial role in many real-world applications. It is a challenging problem as one needs to consider both intra-series temporal correlations and inter-series correlations simultaneously. Recently, there have been multiple works trying to capture both correlations, but most, if not all of them only capture temporal correlations in the time domain and resort to pre-defined priors as inter-series relationships. In this paper, we propose Spectral Temporal Graph Neural Network (StemGNN) to further improve the accuracy of multivariate time-series forecasting. StemGNN captures inter-series correlations and temporal dependencies jointly in the spectral domain. It combines Graph Fourier Transform (GFT) which models inter-series correlations and Discrete Fourier Transform (DFT) which models temporal dependencies in an end-to-end framework. After passing through GFT and DFT, the spectral representations hold clear patterns and can be predicted effectively by convolution and sequential learning modules. Moreover, StemGNN learns inter-series correlations automatically from the data without using pre-defined priors. We conduct extensive experiments on ten real-world datasets to demonstrate the effectiveness of StemGNN.

PDF Details

AAAI Conference 2020 Conference Paper

TextNAS: A Neural Architecture Search Space Tailored for Text Representation

Yujing Wang
Yaming Yang
Yiren Chen
Jing Bai
Ce Zhang
Guinan Su
Xiaoyu Kou
Yunhai Tong

Learning text representation is crucial for text classiﬁcation and other language related tasks. There are a diverse set of text representation networks in the literature, and how to ﬁnd the optimal one is a non-trivial problem. Recently, the emerging Neural Architecture Search (NAS) techniques have demonstrated good potential to solve the problem. Nevertheless, most of the existing works of NAS focus on the search algorithms and pay little attention to the search space. In this paper, we argue that the search space is also an important human prior to the success of NAS in different applications. Thus, we propose a novel search space tailored for text representation. Through automatic search, the discovered network architecture outperforms state-of-the-art models on various public datasets on text classiﬁcation and natural language inference tasks. Furthermore, some of the design principles found in the automatic network agree well with human intuition.

PDF Details

JBHI Journal 2013 Journal Article

360° Fourier Transform Profilometry in Surface Reconstruction for Fluorescence Molecular Tomography

Bi'er Shi
Bin Zhang
Fei Liu
Jianwen Luo
Jing Bai

Fluorescence molecular tomography (FMT) is an emerging tool in the observation of diseases. A fast and accurate surface reconstruction of the experimental object is needed as a boundary constraint for FMT reconstruction. In this paper, an automatic, noncontact, and 3-D surface reconstruction method named 360° Fourier transform profilometry (FTP) is proposed to reconstruct 3-D surface profiles for FMT system. This method can reconstruct 360° integrated surface profiles utilizing the single-frame FTP at different angles. Results show that the relative mean error of the surface reconstruction of this method is less than 1. 4% in phantom experiments, and is no more than 2. 9% in mouse experiments in vivo. Compared with the Radon transform method, the proposed method reduces the computation time by more than 90% with a minimal error increase. At last, a combined 360° FTP/FMT experiment is conducted on a nude mouse. Not only can the 360° FTP system operate with the FMT system simultaneously, but it can also help to monitor the status of animals. Moreover, the 360° FTP system is independent of FMT system and can be performed to reconstruct the surface by itself.

Details DOI

TIST Journal 2012 Journal Article

Leveraging Auxiliary Data for Learning to Rank

Ke Zhou
Jing Bai
Hongyuan Zha
Gui-Rong Xue

In learning to rank, both the quality and quantity of the training data have significant impacts on the performance of the learned ranking functions. However, in many applications, there are usually not sufficient labeled training data for the construction of an accurate ranking model. It is therefore desirable to leverage existing training data from other tasks when learning the ranking function for a particular task, an important problem which we tackle in this article utilizing a boosting framework with transfer learning. In particular, we propose to adaptively learn transferable representations called super-features from the training data of both the target task and the auxiliary task. Those super-features and the coefficients for combining them are learned in an iterative stage-wise fashion. Unlike previous transfer learning methods, the super-features can be adaptively learned by weak learners from the data. Therefore, the proposed framework is sufficiently flexible to deal with complicated common structures among different learning tasks. We evaluate the performance of the proposed transfer learning method for two datasets from the Letor collection and one dataset collected from a commercial search engine, and we also compare our methods with several existing transfer learning methods. Our results demonstrate that the proposed method can enhance the ranking functions of the target tasks utilizing the training data from the auxiliary tasks.

Details DOI

IROS Conference 2006 Conference Paper

The Speech Recognition Based on the Bark Wavelet and CZCPA Features

Xueying Zhang
Jing Bai

The paper uses a new wavelet-Bark wavelet to meet critical frequency band division demand that is consistent with the perception of the human ear to the speech frequency. It is used in front-end processing of speech recognition system as filter bank instead of original FIR filter bank for improving the system performance. The paper gave the concept and parameter setting method of Bark wavelet. At the same time, the paper also presents an improved feature: CZCPA (combining zero-crossings with peak amplitudes) based on ZCPA feature. The new feature includes the information of speech signals and its difference signal. It can improve system performance to some extent. The recognition network uses HMM. The experiment results show that the results of using CZCPA feature with Bark wavelet filters as front-end processor are superior to the results of using ZCPA feature with FIR filter as front-end in speech recognition system

Details