Local Context Features Research Articles

DeepEnhancerPPO: An Interpretable Deep Learning Approach for Enhancer Classification

Enhancers are short genomic segments located in non-coding regions of the genome that play a critical role in regulating the expression of target genes. Despite their importance in transcriptional regulation, effective methods for classifying enhancer categories and regulatory strengths remain limited. To address this challenge, we propose a novel end-to-end deep learning architecture named DeepEnhancerPPO. The model integrates ResNet and Transformer modules to extract local, hierarchical, and long-range contextual features. Following feature fusion, we employ Proximal Policy Optimization (PPO), a reinforcement learning technique, to reduce the dimensionality of the fused features, retaining the most relevant features for downstream classification tasks. We evaluate the performance of DeepEnhancerPPO from multiple perspectives, including ablation analysis, independent tests, assessment of PPO’s contribution to performance enhancement, and interpretability of the classification results. Each module positively contributes to the overall performance, with ResNet and PPO being the most significant contributors. Overall, DeepEnhancerPPO demonstrates superior performance on independent datasets compared to other models, outperforming the second-best model by 6.7% in accuracy for enhancer category classification. The model consistently ranks among the top five classifiers out of 25 for enhancer strength classification without requiring re-optimization of the hyperparameters and ranks as the second-best when the hyperparameters are refined. This indicates that the DeepEnhancerPPO framework is highly robust for enhancer classification. Additionally, the incorporation of PPO enhances the interpretability of the classification results.

Contextual factors that influence adoption and sustainment of self-management support in cancer survivorship care: a practical application of theory with qualitative interviews

BackgroundSelf-management support (SMS) is a recommended component of cancer survivorship care that improves health-related quality of life and reduces healthcare utilisation. However, widespread implementation has been difficult to achieve, with...

MethylGPT: a foundation model for the DNA methylome.

DNA methylation serves as a powerful biomarker for disease diagnosis and biological age assessment. However, current analytical approaches often rely on linear models that cannot capture the complex, context-dependent nature of methylation regulation. Here we present MethylGPT, a transformer-based foundation model trained on 226,555 (154,063 after QC and deduplication) human methylation profiles spanning diverse tissue types from 5,281 datasets, curated 49,156 CpG sites, and 7.6 billion training tokens. MethylGPT learns biologically meaningful representations of CpG sites, capturing both local genomic context and higher-order chromosomal features without external supervision. The model demonstrates robust methylation value prediction (Pearson R=0.929) and maintains stable performance in downstream tasks with up to 70% missing data. Applied to age prediction across multiple tissue types, MethylGPT achieves superior accuracy compared to existing methods. Analysis of the model's attention patterns reveals distinct methylation signatures between young and old samples, with differential enrichment of developmental and aging-associated pathways. When finetuned to mortality and disease prediction across 60 major conditions using 18,859 samples from Generation Scotland, MethylGPT achieves robust predictive performance and enables systematic evaluation of intervention effects on disease risks, demonstrating potential for clinical applications. Our results demonstrate that transformer architectures can effectively model DNA methylation patterns while preserving biological interpretability, suggesting broad utility for epigenetic analysis and clinical applications.

CTFNet: Long-Sequence Time-Series Forecasting Based on Convolution and Time-Frequency Analysis.

Although current time-series forecasting methods have significantly improved the state-of-the-art (SOTA) results for long-sequence time-series forecasting (LSTF), they still have difficulty in capturing and extracting the features and dependencies of long-term sequences and suffer from information utilization bottlenecks and high-computational complexity. To address these issues, a lightweight single-hidden layer feedforward neural network (SLFN) combining convolution mapping and time-frequency decomposition called CTFNet is proposed with three distinctive characteristics. First, time-domain (TD) feature mining-in this article, a method for extracting the long-term correlation of horizontal TD features based on matrix factorization is proposed, which can effectively capture the interdependence among different sample points of a long time series. Second, multitask frequency-domain (FD) feature mining-this can effectively extract different frequency feature information of time-series data from the FD and minimize the loss of data features. Integrating multiscale dilated convolutions, simultaneously focusing on both global and local context feature dependencies at the sequence level, and mining the long-term dependencies of the multiscale frequency information and the spatial dependencies among the different scale frequency information, break the bottleneck of data utilization, and ensure the integrity of feature extraction. Third, highly efficient-the CTFNet model has a short training time and fast inference speed. Our empirical studies with nine benchmark datasets show that compared with state-of-the-art methods, CTFNet can reduce prediction error by 64.7% and 53.7% for multivariate and univariate time series, respectively.

AMHFN: Aggregation Multi-Hierarchical Feature Network for Hyperspectral Image Classification

Deep learning methods like convolution neural networks (CNNs) and transformers are successfully applied in hyperspectral image (HSI) classification due to their ability to extract local contextual features and explore global dependencies, respectively. However, CNNs struggle in modeling long-term dependencies, and transformers may miss subtle spatial-spectral features. To address these challenges, this paper proposes an innovative hybrid HSI classification method aggregating hierarchical spatial-spectral features from a CNN and long pixel dependencies from a transformer. The proposed aggregation multi-hierarchical feature network (AMHFN) is designed to capture various hierarchical features and long dependencies from HSI, improving classification accuracy and efficiency. The proposed AMHFN consists of three key modules: (a) a Local-Pixel Embedding module (LPEM) for capturing prominent spatial-spectral features; (b) a Multi-Scale Convolutional Extraction (MSCE) module to capture multi-scale local spatial-spectral features and aggregate hierarchical local features; (c) a Multi-Scale Global Extraction (MSGE) module to explore multi-scale global dependencies and integrate multi-scale hierarchical global dependencies. Rigorous experiments on three public hyperspectral image (HSI) datasets demonstrated the superior performance of the proposed AMHFN method.

Dynamic convolutional time series forecasting based on adaptive temporal bilateral filtering

Time series data typically contain complex dynamic patterns, which not only include linear trends and seasonal variations but also significant nonlinear changes and complex dependencies. Currently, feature extraction methods for time series data primarily employ mixed-mode extraction within the temporal domain, neglecting the effective extraction and analysis of nonlinear characteristics between different observation points and the interdependencies between variables. To address these issues, we designed an adaptive temporal bilateral filtering module that effectively preserves and highlights the nonlinear features and patterns in time series while filtering out noise and redundant information. We also designed a nonlinear feature adaptive extraction module that integrates a gating mechanism and deformable convolutions. This design allows the model to adaptively adjust the shape of the convolutional kernels according to different time steps and conditions. This enables accurate capture and extraction of nonlinear features between various time observations and dependencies between different variables. Additionally, we use stacked convolutional layers to extract local contextual features, addressing fluctuations in local features caused by changes in data distribution in real-world scenarios. In summary, we propose DCNet, a dynamic convolutional network based on an adaptive temporal bilateral filter, and evaluated its performance on twelve real-world datasets. The results indicate that DCNet consistently achieves state-of-the-art performance in both short-term and long-term forecasting tasks, with favorable runtime efficiency.

DNA flexibility regulates transcription factor binding to nucleosomes.

Cell fate decisions are controlled by sequence-specific transcription factors (TFs), referred to as 'pioneer' factors, that bind their target sites within nucleosomes ('pioneer binding') and thus initiate chromatin opening. However, pioneers bind just a minority of their recognition sequences present in the genome, suggesting that local sequence context features may regulate pioneer binding. Here, we developed PIONEAR-seq, a highly parallel sequencing-based biochemical assay for high-throughput analysis of TF binding to nucleosomes on nucleosome positioning sequences. Using PIONEAR-seq, we characterized the pioneer binding of 7 human pioneer TFs. Comparison of TF binding to nucleosomes based on the synthetic Widom 601 (W601) model sequence versus three different genomic sequences revealed that the positional preferences of these TFs' binding to nucleosomes (i.e., dyad, periodic and end binding) is determined by the broader sequence context of the nucleosome, rather than being a property intrinsic to the TF. We propose a model where the flexibility and rigidity within nucleosomal DNA regulate where pioneers bind within nucleosomes. Our results suggest that the broader physical properties of nucleosomal DNA represent another layer of cis-regulatory information read out by TFs in eukaryotic genomes.

Interactive Siamese spatial-Spectral cross-layer fusion transformer for hyperspectral image change detection

ABSTRACT Recently, methods based on Transformer have been widely used in the research field of hyperspectral image (HSI) change detection (CD). However, existing transformer-based CD research does not sufficiently utilize the spatial-spectral features of HSIs. In this article, we propose an interactive Siamese spatial-spectral cross-layer fusion Transformer (IS2CF-Former) network to improve the accuracy of HSI-CD. The proposed Siamese interactive module integrates the Siamese network with the Transformer structure, enhancing communication between bi-temporal images. We have made improvements to the cross-layer adaptive fusion (CAF) Transformer, where the cross-layer fusion module enhances the interaction between layers and the ability to capture local contextual features, concurrently reducing the model’s parameter count to mitigate the risk of overfitting. The CAF Transformer is applied to extract spatial and spectral features. Evaluating the detection performance of the proposed model on three bi-temporal HSIs through extensive experiments demonstrates superior accuracy compared to seven excellent CD methods.

ConvTrans-CL: Ocean time series temperature data anomaly detection based context contrast learning

Ocean temperature data anomaly detection is instrumental in monitoring environmental changes and implementing measures to alleviate adverse consequences. This holds immense importance for marine environmental observation and scientific inquiry. However, existing anomaly detection methods encounter significant challenges in extracting features from data, which severely affects the performance of anomaly detection. Existing models have limitations in capturing the local contextual distribution features and high stochastic distribution trends of ocean temperature data. Therefore, this paper introduces the ConvTrans-CL model, which integrates the Transformer encoder with causal convolution and employs a contrastive learning approach. Causal convolution extracts the distributional features of short-time subsequences within a sliding window and integrates them into the self-attention mechanism, enabling the model to focus on the localized features of the data. Contrastive learning efficiently captures the long-range dependencies of time-series data by distinguishing between pairs of subsequences with adjacent time intervals and pairs of non-adjacent subsequences. This enables the model to capture the trend of high stochasticity in the distribution of the temperature data. Finally, we select temperature data from two sea areas that are susceptible to multiple environmental factors for experiments and compare the feature extraction and anomaly detection capabilities of ConvTrans-CL with other methods, confirming the superior performance of our method.

An Infrared Aircraft Detection Algorithm Based on Context Perception Feature Enhancement

To address the issue of insufficient extraction of target features and the resulting impact on detection performance in long-range infrared aircraft target detection caused by small imaging area and weak radiation intensity starting from the idea of perceiving target context to enhance the features extracted by convolutional neural network, this paper proposes a detecting algorithm based on AWFGLC (adaptive weighted fusion of global–local context). Based on the mechanism of AWFGLC, the input feature map is randomly reorganized and partitioned along the channel dimension, resulting in two feature maps. One feature map is utilized by self-attention for global context modeling, establishing the correlation between target features and background features to highlight the salient features of the target, thereby enabling the detecting algorithm to better perceive the global features of the target. The other feature map is subjected to window partitioning, with max pooling and average pooling performed within each window to highlight the local features of the target. Subsequently, self-attention is applied to the pooled feature map for local context modeling, establishing the correlation between the target and its surrounding neighborhood, further enhancing the weaker parts of the target features, and enabling the detecting algorithm to better perceive the local features of the target. Based on the characteristics of the target, an adaptive weighted fusion strategy with learnable parameters is employed to aggregate the global context and local context feature maps. This results in a feature map containing more complete target information, enhancing the ability of the detection algorithm to distinguish between target and background. Finally, this paper integrates the mechanism of AWFGLC into YOLOv7 for the detection of infrared aircraft targets. The experiments indicate that the proposed algorithm achieves mAP50 scores of 97.8% and 88.7% on self-made and publicly available infrared aircraft datasets, respectively. Moreover, the mAP50:95 scores reach 65.7% and 61.2%, respectively. These results outperform those of classical target detection algorithms, indicating the effective realization of infrared aircraft target detection.

LACTNet: A Lightweight Real-Time Semantic Segmentation Network Based on an Aggregated Convolutional Neural Network and Transformer

Transformers have demonstrated a significant advantage over CNNs in modeling long-range dependencies, leading to increasing attention being paid towards their application in semantic segmentation tasks. In the present work, a novel semantic segmentation model, LACTNet, is introduced, which synergistically combines Transformer and CNN architectures for the real-time processing of local and global contextual features. LACTNet is designed with a lightweight Transformer, which integrates a specially designed gated convolutional feedforward network, to establish feature dependencies across distant regions. A Lightweight Average Feature Bottleneck (LAFB) module is designed to effectively capture spatial detail information within the features, thereby enhancing segmentation accuracy. To address the issue of spatial feature loss in the decoder, a long skip-connection approach is employed through the designed Feature Fusion Enhancement Module (FFEM), which enhances the integrity of spatial features and the feature interaction capability in the decoder. LACTNet is evaluated on two datasets, achieving a segmentation accuracy of 74.8% mIoU and a frame rate of 90 FPS on the Cityscapes dataset, and a segmentation accuracy of 71.8% mIoU with a frame rate of 126 FPS on the CamVid dataset.

A novel plant type, leaf disease and severity identification framework using CNN and transformer with multi-label method.

The growth of plants is threatened by numerous diseases. Accurate and timely identification of these diseases is crucial to prevent disease spreading. Many deep learning-based methods have been proposed for identifying leaf diseases. However, these methods often combine plant, leaf disease, and severity into one category or treat them separately, resulting in a large number of categories or complex network structures. Given this, this paper proposes a novel leaf disease identification network (LDI-NET) using a multi-label method. It is quite special because it can identify plant type, leaf disease and severity simultaneously using a single straightforward branch model without increasing the number of categories and avoiding extra branches. It consists of three modules, i.e., a feature tokenizer module, a token encoder module and a multi-label decoder module. The LDI-NET works as follows: Firstly, the feature tokenizer module is designed to enhance the capability of extracting local and long-range global contextual features by leveraging the strengths of convolutional neural networks and transformers. Secondly, the token encoder module is utilized to obtain context-rich tokens that can establish relationships among the plant, leaf disease and severity. Thirdly, the multi-label decoder module combined with a residual structure is utilized to fuse shallow and deep contextual features for better utilization of different-level features. This allows the identification of plant type, leaf disease, and severity simultaneously. Experiments show that the proposed LDI-NET outperforms the prevalent methods using the publicly available AI challenger 2018 dataset.

MR2CPPIS: Accurate prediction of protein–protein interaction sites based on multi-scale Res2Net with coordinate attention mechanism

Proteins play a vital role in various biological processes and achieve their functions through protein–protein interactions (PPIs). Thus, accurate identification of PPI sites is essential. Traditional biological methods for identifying PPIs are costly, labor-intensive, and time-consuming. The development of computational prediction methods for PPI sites offers promising alternatives. Most known deep learning (DL) methods employ layer-wise multi-scale CNNs to extract features from protein sequences. But, these methods usually neglect the spatial positions and hierarchical information embedded within protein sequences, which are actually crucial for PPI site prediction. In this paper, we propose MR2CPPIS, a novel sequence-based DL model that utilizes the multi-scale Res2Net with coordinate attention mechanism to exploit multi-scale features and enhance PPI site prediction capability. We leverage the multi-scale Res2Net to expand the receptive field for each network layer, thus capturing multi-scale information of protein sequences at a granular level. To further explore the local contextual features of each target residue, we employ a coordinate attention block to characterize the precise spatial position information, enabling the network to effectively extract long-range dependencies. We evaluate our MR2CPPIS on three public benchmark datasets (Dset 72, Dset 186, and PDBset 164), achieving state-of-the-art performance. The source codes are available at https://github.com/YyinGong/MR2CPPIS.

Semantic segmentation of large-scale point cloud scenes via dual neighborhood feature and global spatial-aware

As a core task in 3D scene information extraction, point cloud semantic segmentation is crucial for understanding 3D scenes and environmental perception. While extracting local geometric structural features from point clouds, existing research often overlooks the long-range dependencies present in the scene, making it challenging to fully uncover the long-range contextual features hidden within point clouds. On this basis, we propose a segmentation algorithm (DG-Net) that integrates dual neighborhood features with global spatial-aware. Initially, the local structure information encoding module is designed to learn about local geometric shapes by encoding spatial position and directional features, thus supplementing structural information. Subsequently, a dual neighborhood features complementary module is introduced to merge the geometric structural and semantic features within local neighborhoods, learning local dependencies and capturing distinguishable local contextual features. Finally, these features are relayed to a global spatial-aware module equipped with a gated unit, which dynamically adjusts the weights of features at different stages, effectively modeling long-range dependencies between local structures and finely extracting long-range contextual features. We conducted experiments on benchmark datasets of point cloud scenes, and both quantitative and qualitative results demonstrate that our algorithm can accurately identify small-scale objects with complex geometric structures within scenes, surpassing other mainstream networks in segmentation performance. The mIoU on the S3DIS, Toronto3D, and SensatUrban datasets are 71.9 %, 82.1 %, and 59.8 %, respectively.

SRTRP-Net: A multi-task learning network for segmentation and prediction of stereotactic radiosurgery treatment response in brain metastases

Before the Stereotactic Radiosurgery (SRS) treatment, it is of great clinical significance to avoid secondary genetic damage and guide the personalized treatment plans for patients with brain metastases (BM) by predicting the response to SRS treatment of brain metastatic lesions. Thus, we developed a multi-task learning model termed SRTRP-Net to provide prior knowledge of BM ROI and predict the SRS treatment response of the lesion. In dual-encoder tumor segmentation Network (DTS-Net), two parallel encoders encode the original and mirrored multi-modal MRI images. The differences in the dual-encoder features between foreground and background are enhanced by the symmetrical visual difference block (SVDB). In the bottom layer of the encoder, a transformer is used to extract local contextual features in the spatial and depth dimensions of low-resolution images. Then, the decoder of DTS-Net provides the prior knowledge for predicting the response to SRS treatment by performing BM segmentation. SRS response prediction network (SRP-Net) directly utilizes shared multi-modal MRI features weighted by the signed distance map (SDM) of the masks. The bidirectional multi-dimensional feature fusion module (BMDF) fuses the shared features and the clinical text information features to obtain comprehensive tumor information for characterizing tumors and predicting SRS treatment response. Experiments based on internal and external clinical datasets have shown that SRTRP-Net achieves comparable or better results. We believe that SRTRP-Net can help clinicians accurately develop personalized first-time treatment regimens for BM patients and improve their survival.

CMNet: Cascaded context fusion and multi-attention network for multiple lesion segmentation of diabetic retinopathy images

Diabetic retinopathy (DR) is a leading cause of blindness in humans, and regular screening is helpful to early detection and containment of DR. Therefore, automated and accurate lesion segmentation is crucial for DR grading and diagnosis. However, it is full of challenges due to complex structures, inconsistent scales and blurry edges of different kinds of lesions. To address the above issues, this paper proposes a cascaded context fusion and multi-attention network (CMNet) for multiple lesion segmentation of DR images. The CMNet includes triple attention module (TAM), cascaded context fusion module (CFM) and balanced attention module (BAM). Firstly, TAM is proposed to extract channel attention, spatial attention and pixel-point attention features for addition fusion to ensure selectivity and consistency of information representation. Moreover, CFM applies adaptive average pooling and non-local operation to capture local and global contextual features for concatenation fusion to expand the receptive field of lesions. Finally, BAM computes foreground, background and boundary attention maps of lesions, as well as uses Squeeze-and-Excitation block to weigh the feature channels and rebalance attention of three regions to make the network focus on the edge details for fine segmentation. We performed comprehensive comparison experiments and ablation studies on three public datasets of DDR, IDRiD and E-Ophtha, and mAUC of four lesions segmentation reached 0.6765, 0.7466 and 0.6710, respectively. The experimental results show that the proposed model outperforms current state-of-the-arts, which overcomes the adverse interference of background noise, and preserves contour details for realizing precise segmentation of multi-scale lesion.

Unveiling group activity recognition: Leveraging Local–Global Context-Aware Graph Reasoning for enhanced actor–scene interactions

Group activity recognition aims to recognize holistic activity in multi-person scene, necessitating consideration of the interactions between actors and their surroundings. It has various applications, such as public surveillance, video analysis. Nonetheless, existing works merely extract scene features as a supplementary component of activity features, failing to adequately explore the interplay between the scene and actors. To address this limitation, this paper proposes a Local–Global Context-Aware Graph Reasoning Model (LG-CAGR), which leverages and reasons through local and global context features to gain deeper insights into group activity within the scene. In particular, we present an innovative feature extraction strategy to harness local location features and global scene attributes, effectively complementing actor features by capturing spatial group topology and determining relative positions between actors. Subsequently, we delve into these features by devising a local–global group reasoning module that deduces pair-wise interactions between actors and scenes within Graph Convolutional Network, comprehensively elucidating correlations between overall scene and local individuals to construct group-level features. Multi-graphs are constructed considering actor’s features, scene features as nodes, and interactions as edges. A self-attention graph pooling network is introduced to automatically integrate key actor features and form rich group-level features to recognize group activity. The results on Collective Activity Dataset, Collective Activity Extended Dataset, Volleyball Dataset and Public Life in Public Space dataset have reached 94.0%, 97.7%, 92.7% and 56.1%. Compared with existing methods using the same backbone, we exceeded 1%, 2.1%, 0.3%, and 14.9% respectively, affirming the superiority of the proposed method compared with state-of-the-art methods.

Light3DHS: A lightweight 3D hippocampus segmentation method using multiscale convolution attention and vision transformer

The morphological analysis and volume measurement of the hippocampus are crucial to the study of many brain diseases. Therefore, an accurate hippocampal segmentation method is beneficial for the development of clinical research in brain diseases. U-Net and its variants have become prevalent in hippocampus segmentation of Magnetic Resonance Imaging (MRI) due to their effectiveness, and the architecture based on Transformer has also received some attention. However, some existing methods focus too much on the shape and volume of the hippocampus rather than its spatial information, and the extracted information is independent of each other, ignoring the correlation between local and global features. In addition, many methods cannot be effectively applied to practical medical image segmentation due to many parameters and high computational complexity. To this end, we combined the advantages of CNNs and ViTs (Vision Transformer) and proposed a simple and lightweight model: Light3DHS for the segmentation of the 3D hippocampus. In order to obtain richer local contextual features, the encoder first utilizes a multi-scale convolutional attention module (MCA) to learn the spatial information of the hippocampus. Considering the importance of local features and global semantics for 3D segmentation, we used a lightweight ViT to learn high-level features of scale invariance and further fuse local-to-global representation. To evaluate the effectiveness of encoder feature representation, we designed three decoders of different complexity to generate segmentation maps. Experiments on three common hippocampal datasets demonstrate that the network achieves more accurate hippocampus segmentation with fewer parameters. Light3DHS performs better than other state-of-the-art algorithms.

A Multi-Scale Attention Fusion Network for Retinal Vessel Segmentation

The structure and function of retinal vessels play a crucial role in diagnosing and treating various ocular and systemic diseases. Therefore, the accurate segmentation of retinal vessels is of paramount importance to assist a clinical diagnosis. U-Net has been highly praised for its outstanding performance in the field of medical image segmentation. However, with the increase in network depth, multiple pooling operations may lead to the problem of crucial information loss. Additionally, handling the insufficient processing of local context features caused by skip connections can affect the accurate segmentation of retinal vessels. To address these problems, we proposed a novel model for retinal vessel segmentation. The proposed model is implemented based on the U-Net architecture, with the addition of two blocks, namely, an MsFE block and MsAF block, between the encoder and decoder at each layer of the U-Net backbone. The MsFE block extracts low-level features from different scales, while the MsAF block performs feature fusion across various scales. Finally, the output of the MsAF block replaces the skip connection in the U-Net backbone. Experimental evaluations on the DRIVE dataset, CHASE_DB1 dataset, and STARE dataset demonstrated that MsAF-UNet exhibited excellent segmentation performance compared with the state-of-the-art methods.

MGCoT: Multi-Grained Contextual Transformer for table-based text generation

Recent advances in Transformer have led to the revolution of table-based text generation. However, most existing Transformer-based architectures ignore the rich contexts among input tokens distributed in multi-level units (e.g., cell, row, or column), leading to sometimes unfaithful text generation that fails to establish accurate association relationships and misses vital information. In this paper, we propose Multi-Grained Contextual Transformer (MGCoT), a novel architecture that fully capitalizes on the multi-grained contexts among input tokens and thus strengthens the capacity of table-based text generation. The key primitive, Multi-Grained Contexts (MGCo) module, involves two components: a local context sub-module that adaptively gathers neighboring tokens to form the token-wise local context features, and a global context sub-module that consistently aggregates tokens from a broader range to form the shared global context feature. The former aims at modeling the short-range dependencies that reflect the salience of tokens within similar fine-grained units (e.g., cell and row) attending to the query token, while the latter aims at capturing the long-range dependencies that reflect the significance of each token within similar coarse-grained units (e.g., multiple rows or columns). Based on the fused multi-grained contexts, MGCoT can flexibly and holistically model the content of a table across multi-level structures. On three benchmark datasets, ToTTo, FeTaQA, and Tablesum, MGCoT outperforms strong baselines by a large margin on the quality of the generated texts, demonstrating the effectiveness of multi-grained context modeling. Our source codes are available at https://github.com/Cedric-Mo/MGCoT.