The goal of the Image Text Multimodal Named Entity Recognition (MNER) task is to identify and classify entities such as individuals, organisations, and places across text and image modalities. Most of the current work focuses on improving the cross-modal alignment model, which has certain limitations: 1. In Chinese social media domain datasets, the mix of Chinese and English language leads to the difficulty of achieving accurate segmentation in the current named entity recognition research. 2. Only image features are regarded as auxiliary information, thus ignoring the impact of image feature processing on the MNER task. To address these issues, we propose a hierarchical subnetwork (HSN) based approach in this paper. Specifically, the visually enhanced MNER model is implemented by constructing subnetworks with flexible different levels. Textual features are extracted through a linguistic model, while visual features are derived from the VGG19 model embedded in SMFE. Features from both text and visual models are fused through two subnetworks dynamically constructed based on the data features to enhance the consistency of the representations. Finally, we use Conditional Random Fields (CRF) to sequence tag the output of the multi-modal interaction module. The proposed method achieves F1 scores of 88.05% and 80.24% on the CMNER dataset and the Wukong-Cmner dataset, respectively, demonstrating the competitiveness of our method.

Objective-space decomposition algorithms (ODAs) are widely studied for solving multi-objective integer programs. However, they often encounter difficulties in handling scalarized problems, which could cause infeasibility or repetitive nondominated points and thus induce redundant runtime. To mitigate the issue, we present a graph neural network (GNN) based method to learn the reduction rule in the ODA. We formulate the algorithmic procedure of generic ODAs as a Markov decision process, and parameterize the policy (reduction rule) with a novel two-stage GNN to fuse information from variables, constraints and especially objectives for better state representation. We train our model with imitation learning and deploy it on a state-of-the-art ODA. Results show that our method significantly improves the solving efficiency of the ODA. The learned policy generalizes fairly well to larger problems or more objectives, and the proposed GNN outperforms existing ones for integer programming in terms of test and generalization accuracy.

Recent studies have shown that the transcriptional functions of REST are much broader than repressing neuronal genes in non-neuronal systems. Whether REST occupies similar chromatin regions in different cell types and how it interacts with other transcriptional regulators to execute its functions in a context-dependent manner has not been adequately investigated. We have applied ChIP-seq analysis to identify the REST cistrome in human CD4+ T cells and compared it with published data from 15 other cell types. We found that REST cistromes were distinct among cell types, with REST binding to several tumor suppressors specifically in cancer cells, whereas 7% of the REST peaks in non-neuronal cells were ubiquitously called and <25% were identified for ≥ 5 cell types. Nevertheless, using a quantitative metric directly comparing raw ChIP-seq signals, we found the majority (∼80%) was shared by ≥ 2 cell types. Integration with RNA-seq data showed that REST binding was generally correlated with low gene expression. Close examination revealed that multiple contexts were correlated with reduced expression of REST targets, e.g., the presence of a cognate RE1 motif and cellular specificity of REST binding. These contexts were shown to play a role in differential corepressor recruitment. Furthermore, transcriptional outcome was highly influenced by REST cofactors, e.g., SIN3 and EZH2 co-occupancy marked higher and lower expression of REST targets, respectively. Unexpectedly, the REST cistrome in differentiated neurons exhibited unique features not observed in non-neuronal cells, e.g., the lack of RE1 motifs and an association with active gene expression. Finally, our analysis demonstrated how REST could differentially regulate a transcription network constituted of miRNAs, REST complex and neuronal factors. Overall, our findings of contexts playing critical roles in REST occupancy and regulatory outcome provide insights into the molecular interactions underlying REST's diverse functions, and point to novel roles of REST in differentiated neurons.