The paper presents a novel method called Adaptive Feature Imputation with Latent Graph for Deep Incomplete Multi-View Clustering (AGDIMC) to address the challenges of incomplete multi-view clustering (IMVC). AGDIMC integrates deep learning and graph-based techniques to improve the accuracy and robustness of clustering on incomplete data. The key contributions of AGDIMC include: 1. **View-Specific Feature Learning**: It employs view-specific deep encoders to capture embedded features of each view, enhancing the representation learning capabilities. 2. **Latent Graph Construction**: Partial latent graphs are constructed from complete data to preserve intrinsic relationships within each view and capture topological information. 3. **Adaptive Imputation Layer**: This layer uses cross-view soft cluster assignments and global cluster centroids to impute missing data, improving the quality of imputed features. 4. **Graph Fusion**: The method fuses weight graphs and latent graphs to enhance the discriminative information in the imputed features. 5. **Graph Embedding Constraint**: This constraint reinforces the neighbor relationships in the latent graphs, preventing the distortion of discriminative information during fusion. Experimental results on multiple real-world datasets demonstrate that AGDIMC outperforms existing state-of-the-art methods in terms of clustering performance, even with high missing rates. The effectiveness of the proposed method is validated through ablation studies and convergence analysis, showing its robustness and stability.The paper presents a novel method called Adaptive Feature Imputation with Latent Graph for Deep Incomplete Multi-View Clustering (AGDIMC) to address the challenges of incomplete multi-view clustering (IMVC). AGDIMC integrates deep learning and graph-based techniques to improve the accuracy and robustness of clustering on incomplete data. The key contributions of AGDIMC include: 1. **View-Specific Feature Learning**: It employs view-specific deep encoders to capture embedded features of each view, enhancing the representation learning capabilities. 2. **Latent Graph Construction**: Partial latent graphs are constructed from complete data to preserve intrinsic relationships within each view and capture topological information. 3. **Adaptive Imputation Layer**: This layer uses cross-view soft cluster assignments and global cluster centroids to impute missing data, improving the quality of imputed features. 4. **Graph Fusion**: The method fuses weight graphs and latent graphs to enhance the discriminative information in the imputed features. 5. **Graph Embedding Constraint**: This constraint reinforces the neighbor relationships in the latent graphs, preventing the distortion of discriminative information during fusion. Experimental results on multiple real-world datasets demonstrate that AGDIMC outperforms existing state-of-the-art methods in terms of clustering performance, even with high missing rates. The effectiveness of the proposed method is validated through ablation studies and convergence analysis, showing its robustness and stability.