Semi-Supervised Learning (SSL) aims to improve the learning performance of supervised learning with a large number of unlabeled samples. The existing SSL methods such as FixMatch and FlexMatch select unlabeled samples with high-confident pseudo-labels and make consistency constraints between their weak and strong augmentations. Unfortunately, they cannot be applied Semi-Supervised Regression (SSR) because regression predictions can not reflect the confidence of pseudo-labels. To solve this, a recent SSR method RankUp incorporates an auxiliary ranking task by leveraging sample pairs with high-confident pseudo-ranks. In this paper, we upgrade Rankup to a novel SSR method, namely Semi-Supervised Regression by Ranking Close Unlabeled Samples (SSR-RCUS). Its basic idea is reconstructing closed mixup augmented samples with high-confident pseudo-ranks under a monotonicity assumption, and then applying them to the auxiliary ranking task to improve regression performance. We conduct extensive experiments to evaluate the performance of SSR-RCUS on benchmark datasets, and empirical results demonstrate that SSR-RCUS can outperform the existing baselines in various settings, especially when labeled data are scarce.

The task of stochastic human motion prediction has attracted significant attention in recent years due to its wide-ranging applications in robotics, animation, and human-computer interaction. While diffusion models have demonstrated promising progress in this domain, they remain hindered by two critical limitations: (1) slow inference speeds due to their reliance on iterative sampling, and (2) performance degradation resulting from suboptimal sample allocation during generation. To overcome these challenges, we propose SPARD (Single-step Inference with Adaptive Sampling in Residual Diffusion for Human Motion Prediction), a novel framework that achieves efficient single-step inference while maintaining high predictive accuracy. Furthermore, we introduce a novel adaptive noise predictor module that dynamically samples latent representations based on observed motion sequences, ensuring both accuracy and plausibility in generated motions. Extensive experiments on benchmark datasets demonstrate that SPARD significantly outperforms state-of-the-art methods in both inference efficiency and motion quality, achieving a 15× to 18× speedup in sampling time compared to conventional diffusion-based baselines while preserving generation quality.

Text classification is a crucial and fundamental task in web content mining. Compared with the previous learning paradigm of pre-training and fine-tuning by cross entropy loss, the recently proposed supervised contrastive learning approach has received tremendous attention due to its powerful feature learning capability and robustness. Although several studies have incorporated this technique for text classification, some limitations remain. First, many text datasets are imbalanced, and the learning mechanism of supervised contrastive learning is sensitive to data imbalance, which may harm the model's performance. Moreover, these models leverage separate classification branches with cross entropy and supervised contrastive learning branches without explicit mutual guidance. To this end, we propose a novel model named SharpReCL for imbalanced text classification tasks. First, we obtain the prototype vector of each class in the balanced classification branch to act as a representation of each class. Then, by further explicitly leveraging the prototype vectors, we construct a proper and sufficient target sample set with the same size for each class to perform the supervised contrastive learning procedure. The empirical results show the effectiveness of our model, which even outperforms popular large language models across several datasets. Our code is available here.

MOTIVATION: Spatial transcriptomics (ST) enables the measurement of gene expression while preserving the spatial context of tissues. However, the sparsity of ST data leads to poor usage of gene expression and spatial information, resulting in the embeddings that are not well represented and challenging for downstream analyses. RESULTS: Here, we introduced GSG, a generative self-supervised representation learning framework for ST data that leverages a masking mechanism to learn informative representations. For spatial domain identification, GSG consistently outperformed state-of-the-art methods across benchmarking datasets, regardless of sequencing platforms. In addition, we applied GSG to an in-house human fetal heart dataset, revealing anatomically coherent spatial domains and identifying APCDD1 as an endocardial-specific marker potentially involved in congenital heart disease. Our results showcase GSG's superiority and underscore its valuable contributions to advancing ST analysis. AVAILABILITY AND IMPLEMENTATION: Our software package is available at https://github.com/keaml-Guan/GSG.

Competitive endogenous RNA (ceRNA) networks are pivotal for uncovering disease molecular mechanisms. Graph representation learning is a cornerstone for modeling biological regulatory networks and predicting disease-related biomarkers. However, current methods face challenges: traditional graph neural network (GNN) rely on low-order graph structures, which struggle to capture high-order molecular interactions, resulting in topological information loss; shallow GNN fail to model long-range dependencies, while deep architectures suffer from over-smoothing, limiting complex regulatory expression; static embeddings overlook dynamic molecular interactions, reducing biomarker accuracy. These limitations highlight the need for advanced graph learning frameworks. To address these challenges, we propose DMHLF, a Dynamic Multi-scale Hypergraph Learning Framework for predicting disease-associated ceRNA biomarkers. The framework first integrates multiple regulatory relationships among miRNAs, lncRNAs, circRNAs, mRNAs, and diseases to construct disease-specific ceRNA regulatory networks, capturing local and global regulatory patterns through multi-Hop hyperedges. Subsequently, we devise a Hypergraph-Weighted Dynamic Random Walk (HEDRW) method to dynamically extract node meta-embeddings that encode high-order regulatory information. Concurrently, we extend Eigen-GNN spectral analysis to hypergraph structures, incorporating a residual-enhanced hypergraph neural network to preserve the global topological properties of shallow hypergraphs. Finally, a cross-scale attention mechanism aligns and fuses multi-scale features to generate high-quality node embeddings for disease-ceRNA association prediction. Experiments on diverse datasets demonstrate that DMHLF significantly outperforms existing methods. Case study further validates the framework's efficacy in identifying disease-related ceRNA biomarkers, providing a reliable predictive tool for biomedical research.

The advancement of Internet technology has spurred a rise in the dissemination of misinformation, which has had profoundly negative impacts across a wide array of fields. To address this issue, the field of Misinformation Detection (MD), which focuses on the automated identification of online misinformation, has gained significant traction among researchers. In our study, we introduce an innovative plugand- play augmentation technique for MD, termed DEtecting Misinformation by Uncovering Commonsense Conflict (DEMUC). Our approach is grounded in previous psychological research that suggests that fake content often contains commonsense. Accordingly, we develop commonsense expressions for articles to highlight potential conflicts between the inferred commonsense triplets and the established ones derived from reliable commonsense reasoning tools. According to the used tools, we induce two variants DEMUC-KLM using the knowledge language model COMET and DEMUC-LLM using the large language models. These generated expressions are then applied as augmentations to each article, enabling any MD method to be trained on these augmented datasets. Additionally, we have manually compiled a new dataset CoMis, which consists exclusively of fake articles characterized by commonsense conflicts. By integrating DEMUC with various existing MD frameworks and evaluating them on four public benchmark datasets and CoMis, our empirical findings show that both DEMUC-KLM and DEMUC-LLM consistently and significantly outperform current MD baselines, while also generating precise commonsense expressions.

Short text classification has gained significant attention in the information age due to its prevalence and real-world applications. Recent advancements in graph learning combined with contrastive learning have shown promising results in addressing the challenges of semantic sparsity and limited labeled data in short text classification. However, existing models have certain limitations. They rely on explicit data augmentation techniques to generate contrastive views, resulting in semantic corruption and noise. Additionally, these models only focus on learning the intrinsic consistency between the generated views, neglecting valuable discriminative information from other potential views. To address these issues, we propose a Simple graph contrastive learning framework for Short Text Classification (SimSTC). Our approach involves performing graph learning on multiple text-related component graphs to obtain multi-view text embeddings. Subsequently, we directly apply contrastive learning on these embeddings. Notably, our method eliminates the need for data augmentation operations to generate contrastive views while still leveraging the benefits of multi-view contrastive learning. Despite its simplicity, our model achieves outstanding performance, surpassing large language models on various datasets.

Short text classification, as a research subtopic in natural language processing, is more challenging due to its semantic sparsity and insufficient labeled samples in practical scenarios. We propose a novel model named MI-DELIGHT for short text classification in this work. Specifically, it first performs multi-source information (i.e., statistical information, linguistic information, and factual information) exploration to alleviate the sparsity issues. Then, the graph learning approach is adopted to learn the representation of short texts, which are presented in graph forms. Moreover, we introduce a dual-level (i.e., instance-level and cluster-level) contrastive learning auxiliary task to effectively capture different-grained contrastive information within massive unlabeled data. Meanwhile, previous models merely perform the main task and auxiliary tasks in parallel, without considering the relationship among tasks. Therefore, we introduce a hierarchical architecture to explicitly model the correlations between tasks. We conduct extensive experiments across various benchmark datasets, demonstrating that MI-DELIGHT significantly surpasses previous competitive models. It even outperforms popular large language models on several datasets.

Drug repositioning is a promising approach to discover new therapeutic applications with existing drugs. However, existing Graph Neural Networks (GNNs)-based methods suffer from two limitations: (1) the sparsity of labeled drug-disease associations and high-quality node representations, despite the continuous emergence of biomedical knowledge from multiple modalities; and (2) the over-smoothing issue in GNNs, which causes representations to become indistinguishable, thereby degrading model performance. To address these issues, we propose Panacea, a graph learning framework that leverages multimodal information to enhance drug repositioning performance. Panacea introduces an automated knowledge acquisition and refinement pipeline that collects and encodes multimodal information, including drug molecular structures (SMILES strings), clinical symptom descriptions of diseases, and both homogeneous and heterogeneous biomedical graphs. The resulting multimodal embeddings are aligned via a learnable projection layer, providing strong initial node representations for graph learning. In the subsequent hierarchical graph learning stage, we improve the Graph Isomorphism Network (GIN) with gated mechanisms and residual connections to enhance the network's representation capacity and avoid over-smoothing. The multimodal embeddings and the graph learning module are jointly optimized in an end-toend manner, enabling node representations to fuse graph structure. Experimental results demonstrate that Panacea outperforms state-of-the-art methods, achieving significant improvements in drug repositioning tasks, especially in scenarios with sparse data and insufficient information representation. Code is available at here.

Early screening for prenatal depression is of critical importance to maternal and infant health. However, traditional screening methods are fundamentally limited by a confluence of challenges, including data imbalance, resource constraints, poor model adaptability, and a lack of interpretability. While Large Language Models (LLMs) show potential in medical text analysis, their direct application to class-imbalanced tasks requires significant optimization. To address these limitations, we propose and evaluate an Instruction Fine-Tuning methodology, integrating prompt engineering directly into the training phase by prepending each questionnaire-based case description with a structured, task-specific directive. This directive encodes expert domain knowledge and an explicit instruction for the model to prioritize high recall, thereby directly addressing the clinical need to minimize false negatives. Several language models were adapted using this instruction-augmented dataset via the parameter-efficient Low-Rank Adaptation (LoRA) technique. The experimental results demonstrate that models trained with our Instruction Fine-Tuning method exhibit significantly higher recall for the minority class and achieve a more balanced overall performance. Our best-performing instruction-tuned model achieved an AUC of 0.81 and a Balanced Accuracy of 0.72, validating the effectiveness of our approach. Crucially, our interpretability analysis revealed that the models learned to focus on clinically-relevant risk factors, providing transparent and plausible justifications for their predictions. This research confirms that Instruction Fine-Tuning is a highly effective methodology for adapting LLMs to sensitive, imbalanced clinical screening tasks. The proposed approach presents a robust, interpretable, and adaptable solution well-suited to the demands of low-resource medical environments.

The proliferation of misinformation across diverse social media platforms has drawn significant attention from both academic and industrial communities due to its detrimental effects. Accordingly, automatically distinguishing misinformation, dubbed as Misinformation Detection (MD), has become an increasingly active research topic. The mainstream methods formulate MD as a static learning paradigm, which learns the mapping between the content, links, and propagation of news articles and the corresponding manual veracity labels. However, the static assumption is often violated, since in real-world scenarios, the veracity of news articles may vacillate within the dynamically evolving social environment. To tackle this problem, we propose a novel framework, namely Misinformation detection with Dynamic Environmental Representations (MISDER). The basic idea of MISDER lies in learning a social environmental representation for each period and employing a temporal model to predict the representation for future periods. In this work, we specify the temporal model as the LSTM model, continuous dynamics equation, and pre-trained dynamics system, suggesting three variants of MISDER, namely MISDER-LSTM, MISDER-ODE, and MISDER-PT, respectively. To evaluate the performance of MISDER, we compare it to various MD baselines across 2 prevalent datasets, and the experimental results can indicate the effectiveness of our proposed model.

The development of Internet technology has led to an increased prevalence of misinformation, causing severe negative effects across diverse domains. To mitigate this challenge, Misinformation Detection (MD), aiming to detect online misinformation automatically, emerges as a rapidly growing research topic in the community. In this paper, we propose a novel plug-and-play augmentation method for the MD task, namely Misinformation Detection with Potential Commonsense Conflict (MD-PCC). We take inspiration from the prior studies indicating that fake articles are more likely to involve commonsense conflict. Accordingly, we construct commonsense expressions for articles, serving to express potential commonsense conflicts inferred by the difference between extracted commonsense triplet and golden ones inferred by the well-established commonsense reasoning tool COMET. These expressions are then specified for each article as augmentation. Any specific MD methods can be then trained on those commonsense-augmented articles. Besides, we also collect a novel commonsense-oriented dataset named CoMis, whose all fake articles are caused by commonsense conflict. We integrate MD-PCC with various existing MD backbones and compare them across both 4 public benchmark datasets and CoMis. Empirical results demonstrate that MD-PCC can consistently outperform the existing MD baselines.

Graph few-shot learning has garnered significant attention for its ability to rapidly adapt to downstream tasks with limited labeled data, sparking considerable interest among researchers. Recent advancements in graph few-shot learning models have exhibited superior performance across diverse applications. Despite their successes, several limitations still exist. First, existing models in the meta-training phase predominantly focus on instance-level features within tasks, neglecting crucial set-level features essential for distinguishing between different categories. Second, these models often utilize query sets directly on classifiers trained with support sets containing only a few labeled examples, overlooking potential distribution shifts between these sets and leading to suboptimal performance. Finally, previous models typically require necessitate abundant labeled data from base classes to extract transferable knowledge, which is typically infeasible in real-world scenarios. To address these issues, we propose a novel model named STAR, which leverages Set funcTions and optimAl tRansport for enhancing unsupervised graph few-shot learning. Specifically, STAR utilizes expressive set functions to obtain set-level features in an unsupervised manner and employs optimal transport principles to align the distributions of support and query sets, thereby mitigating distribution shift effects. Theoretical analysis demonstrates that STAR can capture more task-relevant information and enhance generalization capabilities. Empirically, extensive experiments across multiple datasets validate the effectiveness of STAR. Our code can be found here.

Positive and Unlabeled (PU) learning is a special case of binary classification with weak supervision, where only positive labeled and unlabeled data are available. Previous studies suggest several specific risk estimators of PU learning such as non-negative PU (nnPU), which are unbiased and consistent with the expected risk of supervised binary classification. In nnPU, the negative-class empirical risk is estimated by positive labeled and unlabeled data with a non-negativity constraint. However, its negative-class empirical risk estimator approaches 0, so the negative class is over-played, resulting in imbalanced error rates between positive and negative classes. To solve this problem, we suppose that the expected risks of the positive-class and negative-class should be close. Accordingly, we constrain that the negative-class empirical risk estimator is lower bounded by the positive-class empirical risk, instead of 0; and also incorporate an explicit equality constraint between them. we suggest a risk estimator of PU learning that balances positive and negative classification error rates, named $\mathrm{D{\small C-PU} }$, and suggest an efficient training method for $\mathrm{D{\small C-PU} }$ based on the augmented Lagrange multiplier framework. We theoretically analyze the estimation error of $\mathrm{D{\small C-PU} }$ and empirically validate that $\mathrm{D{\small C-PU} }$ achieves higher accuracy and converges more stable than other risk estimators of PU learning. Additionally, $\mathrm{D{\small C-PU} }$ also performs competitive accuracy performance with practical PU learning methods.

Graph few-shot learning has attracted increasing attention due to its ability to rapidly adapt models to new tasks with only limited labeled nodes. Despite the remarkable progress made by existing graph few-shot learning methods, several key limitations remain. First, most current approaches rely on predefined and unified graph filters (e.g., low-pass or high-pass filters) to globally enhance or suppress node frequency signals. Such fixed spectral operations fail to account for the heterogeneity of local topological structures inherent in real-world graphs. Moreover, these methods often assume that the support and query sets are drawn from the same distribution. However, under few-shot conditions, the limited labeled data in the support set may not sufficiently capture the complex distribution of the query set, leading to suboptimal generalization. To address these challenges, we propose GRACE, a novel Graph few-shot leaRning framework that integrates Adaptive spectrum experts with Cross-sEt distribution calibration techniques. Theoretically, the proposed approach enhances model generalization by adapting to both local structural variations and cross-set distribution calibration. Empirically, GRACE consistently outperforms state-of-the-art baselines across a wide range of experimental settings. Our code can be found here.

Graph neural networks have been demonstrated as a powerful paradigm for effectively learning graph-structured data on the web and mining content from it. %the wide web. for downstream task analysis. Current leading graph models require a large number of labeled samples for training, which unavoidably leads to overfitting in few-shot scenarios. Recent research has sought to alleviate this issue by simultaneously leveraging graph learning and meta-learning paradigms. However, these graph meta-learning models assume the availability of numerous meta-training tasks to learn transferable meta-knowledge. Such assumption may not be feasible in the real world due to the difficulty of constructing tasks and the substantial costs involved. Therefore, we propose a SiMple yet effectIve approach for graph few-shot Learning with fEwer tasks, named SMILE. We introduce a dual-level mixup strategy, encompassing both within-task and across-task mixup, to simultaneously enrich the available nodes and tasks in meta-learning. Moreover, we explicitly leverage the prior information provided by the node degrees in the graph to encode expressive node representations. Theoretically, we demonstrate that SMILE can enhance the model generalization ability. Empirically, SMILE consistently outperforms other competitive models by a large margin across all evaluated datasets with in-domain and cross-domain settings. Our anonymous code can be found https://github.com/KEAML-JLU/SMILE.

Craters are the most typical geologic structures and landforms on the surface of Mars. Martian craters are widely distributed in a variety of morphology with multiple types and exhibit significant differences in scale. Many attempts have been made to automatic identification of Martian craters, yet existing methods do not satisfy the need for large-scale identification. In this article, we first integrate a Martian crater dataset related to the mid- and low-latitude regions, which contains different types and various scales of craters. Then, a dual convolutional neural network (CNN)-Transformer-based cross-modality adaptive feature fusion network (DCT-CMAFFNet) is proposed for accurate identification of the multitype and multiscale craters at large scale on the Martian surface. The proposed network takes full advantage of the rich morphological features contained in Martian imagery and the topographical information reflected by the digital elevation model (DEM) data. It contains two modules: one is the dual CNN-Transformer part, which employs a hybrid architecture to extract the local detailed and global deep features of Martian craters from images and DEM and the other is CMAFF module, which exploits self-attention mechanism to learn the relationship between the images and DEM modalities and weigh each position of the deep feature maps to ensure a comprehensive identification of multitype and multiscale Martian craters. By adaptively fusing the rich information from imagery and DEM, the proposed network identified 3166 new Martian impact craters larger than 1 km, achieving a 14%–24% improvement in accuracy compared to methods using either a single data source or data feature fusion modality.

Text classification occupies an important role in natural language processing and has many applications in real life. Short text classification, as one of its subtopics, has attracted increasing interest from researchers since it is more challenging due to its semantic sparsity and insufficient labeled data. Recent studies attempt to combine graph learning and contrastive learning to alleviate the above problems in short text classification. Despite their fruitful success, there are still several inherent limitations. First, the generation of augmented views may disrupt the semantic structure within the text and introduce negative effects due to noise permutation. Second, they ignore the clustering-friendly features in unlabeled data and fail to further utilize the prior information in few valuable labeled data. To this end, we propose a novel model that utilizes improved Graph contrastIve learning for short text classiFicaTion (GIFT). Specifically, we construct a heterogeneous graph containing several component graphs by mining from an internal corpus and introducing an external knowledge graph. Then, we use singular value decomposition to generate augmented views for graph contrastive learning. Moreover, we employ constrained kmeans on labeled texts to learn clustering-friendly features, which facilitate cluster-oriented contrastive learning and assist in obtaining better category boundaries. Extensive experimental results show that GIFT significantly outperforms previous state-of-the-art methods. Our code can be found in https://github.com/KEAML-JLU/GIFT.

Drug-Drug Interaction (DDI) task plays a crucial role in clinical treatment and drug development. Recently, deep learning methods have been successfully applied for DDI prediction. However, training deep learning models always need large amount of data, while known DDIs are scarce. To address this challenge, a graph neural network-based DDI prediction model named PNESR-DDI is proposed, which compensates for the lack of DDIs by enriching drug representations. First, to obtain initial node representations that incorporate rich semantic information from the biomedical knowledge graph (KG), a link prediction pre-training method on external KG is proposed in the node embedding pre-training module. Then, considering the large scale of the KG, subgraph extraction for the target drug pairs is introduced to reduce noise and decrease computational complexity in the subgraph anchoring module. After that, the subgraph is updated, and node similarities are propagated in the subgraph reconstruction module. Based on the node similarity scores, the subgraph is pruned and reconstructed, which adjusts node representations to be more conducive to DDI prediction. Finally, the drug embeddings, subgraph representations, and drug fingerprint features are concatenated to predict DDIs. PNESRDDI is evaluated on two benchmark DDI datasets: DrugBank and TWOSIDES. Experiment results show that PNESR-DDI achieves better performance than baselines. Ablation results validate the effectiveness of the pre-training method and the adaptive subgraph reconstruction strategy.

Named entity recognition (NER) in electronic medical records (EMRs) is critical for identifying medical entities, constructing medical knowledge graphs, and supporting clinical decision-making. However, the scarcity of EMR datasets and the challenges posed by the complexity of Chinese medical texts hinder progress in this area. To address these issues, we introduce CMR-NER, a nested named entity recognition model that enhances entity prediction by integrating multiple features and considering a global view of entity boundaries. Leveraging the robust generalization capabilities of the large language model, CMR-NER involves collaboration with medical experts to ensure accuracy and reduce annotation costs. Additionally, we present HF-NER, a high-quality dataset specifically curated for Chinese EMR entity recognition focusing on heart failure. This dataset, constructed through a methodology combining ChatGPT’s capabilities and expert validation, is a significant contribution to the field. It facilitates a comprehensive evaluation of CMR-NER. Experimental results demonstrate that our approach achieves comparable or superior performance to existing methods, underscoring its effectiveness in this challenging domain.

Nowadays, misinformation is widely spreading over various social media platforms and causes extremely negative impacts on society. To combat this issue, automatically identifying misinformation, especially those containing multimodal content, has attracted growing attention from the academic and industrial communities, and induced an active research topic named Multimodal Misinformation Detection (MMD). Typically, existing MMD methods capture the semantic correlation and inconsistency between multiple modalities, but neglect some potential clues in multimodal content. Recent studies suggest that manipulated traces of the images in articles are non-trivial clues for detecting misinformation. Meanwhile, we find that the underlying intentions behind the manipulation, e.g., harmful and harmless, also matter in MMD. Accordingly, in this work, we propose to detect misinformation by learning manipulation features that indicate whether the image has been manipulated, as well as intention features regarding the harmful and harmless intentions of the manipulation. Unfortunately, the manipulation and intention labels that make these features discriminative are unknown. To overcome the problem, we propose two weakly supervised signals as alternatives by introducing additional datasets on image manipulation detection and formulating two classification tasks as positive and unlabeled learning problems. Based on these ideas, we propose a novel MMD method, namely Harmfully Manipulated Images Matter in MMD (Hami-m3d). Extensive experiments across three benchmark datasets can demonstrate that Hami-m3d can consistently improve the performance of any MMD baselines.

Zero-shot Chain-of-Thought (CoT) prompting emerges as a simple and effective strategy for enhancing the performance of large language models (LLMs) in real-world reasoning tasks. Nonetheless, the efficacy of a singular, task-level prompt uniformly applied across the whole of instances is inherently limited since one prompt cannot be a good partner for all, a more appropriate approach should consider the interaction between the prompt and each instance meticulously. This work introduces an instance-adaptive prompting algorithm as an alternative zero-shot CoT reasoning scheme by adaptively differentiating good and bad prompts. Concretely, we first employ analysis on LLMs through the lens of information flow to detect the mechanism under zero-shot CoT reasoning, in which we discover that information flows from question to prompt and question to rationale jointly influence the reasoning results most. We notice that a better zero-shot CoT reasoning needs the prompt to obtain semantic information from the question then the rationale aggregates sufficient information from the question directly and via the prompt indirectly. On the contrary, lacking any of those would probably lead to a bad one. Stem from that, we further propose an instance-adaptive prompting strategy (IAP) for zero-shot CoT reasoning. Experiments conducted with LLaMA-2, LLaMA-3, and Qwen on math, logic, and commonsense reasoning tasks (e.g., GSM8K, MMLU, Causal Judgement) obtain consistent improvement, demonstrating that the instance-adaptive zero-shot CoT prompting performs better than other task-level methods with some curated prompts or sophisticated procedures, showing the significance of our findings in the zero-shot CoT reasoning mechanism.

Graphs, as a fundamental data structure, have proven efficacy in modeling complex relationships between objects and are therefore found in wide web applications. Graph classification is an essential task in graph data analysis, which can effectively assist in extracting information and mining content from the web. Recently, few-shot graph classification, a more realistic and challenging task, has garnered great research interest. Existing few-shot graph classification models are all supervised, assuming abundant labeled data in base classes for meta-training. However, sufficient annotation is often challenging to obtain in practice due to high costs or demand for expertise. Moreover, they commonly adopt complicated meta-learning algorithms via episodic training to transfer prior knowledge from base classes. To break free from these constraints, in this paper, we propose a simple yet effective approach named SMART for unsupervised few-shot graph classification without using any labeled data. SMART employs transfer learning philosophy instead of the previously prevailing meta-learning paradigm, avoiding the need for sophisticated meta-learning algorithms. Additionally, we adopt a novel mixup strategy to augment the original graph data and leverage unsupervised pretraining on these data to obtain the expressive graph encoder. We also utilize the prompt tuning technique to alleviate the overfitting and low fine-tuning efficiency caused by the limited support samples of novel classes. Extensive experimental results demonstrate the superiority of our proposed approach, significantly surpassing even leading supervised few-shot graph classification models. Our code is available here.

MOTIVATION: Research shows that competing endogenous RNA is widely involved in gene regulation in cells, and identifying the association between circular RNA (circRNA), microRNA (miRNA), and cancer can provide new hope for disease diagnosis, treatment, and prognosis. However, affected by reductionism, previous studies regarded the prediction of circRNA-miRNA interaction, circRNA-cancer association, and miRNA-cancer association as separate studies. Currently, few models are capable of simultaneously predicting these three associations. RESULTS: Inspired by holism, we propose a multi-task prediction method based on neighborhood structure embedding and signed graph representation learning, CMCSG, to infer the relationship between circRNA, miRNA, and cancer. Our method aims to extract feature descriptors of all molecules from the circRNA-miRNA-cancer regulatory network using known types of association information to predict unknown types of molecular associations. Specifically, we first constructed the circRNA-miRNA-cancer association network (CMCN), which is constructed based on the experimentally verified biomedical entity regulatory network; next, we combine topological structure embedding methods to extract feature representations in CMCN from local and global perspectives, and use denoising autoencoder for enhancement; then, combined with balance theory and state theory, molecular features are extracted from the point of social relations through the propagation and aggregation of signed graph attention network; finally, the GBDT classifier is used to predict the association of molecules. The results show that CMCSG can effectively predict the relationship between circRNA, miRNA, and cancer. Additionally, the case studies also demonstrate that CMCSG is capable of accurately identifying biomarkers across various types of cancer. The data and source code can be found at https://github.com/1axin/CMCSG.

The discovery of diagnostic and therapeutic biomarkers for complex diseases, especially cancer, has always been a central and long-term challenge in molecular association prediction research, offering promising avenues for advancing the understanding of complex diseases. To this end, researchers have developed various network-based prediction techniques targeting specific molecular associations. However, limitations imposed by reductionism and network representation learning have led existing studies to narrowly focus on high prediction efficiency within single association type, thereby glossing over the discovery of unknown types of associations. Additionally, effectively utilizing network structure to fit the interaction properties of regulatory networks and combining specific case biomarker validations remains an unresolved issue in cancer biomarker prediction methods. To overcome these limitations, we propose a multi-view learning framework, CeRVE, based on directed graph neural networks (DGNN) for predicting unknown type cancer biomarkers. CeRVE effectively extracts and integrates subgraph information through multi-view feature learning. Subsequently, CeRVE utilizes DGNN to simulate the entire regulatory network, propagating node attribute features and extracting various interaction relationships between molecules. Furthermore, CeRVE constructed a comparative analysis matrix of three cancers and adjacent normal tissues through The Cancer Genome Atlas and identified multiple types of potential cancer biomarkers through differential expression analysis of mRNA, microRNA, and long noncoding RNA. Computational testing of multiple types of biomarkers for 72 cancers demonstrates that CeRVE exhibits superior performance in cancer biomarker prediction, providing a powerful tool and insightful approach for AI-assisted disease biomarker discovery.

Identifying potential cancer biomarkers is a key task in biomedical research, providing a promising avenue for the diagnosis and treatment of human tumors and cancers. In recent years, several machine learning-based RNA-disease association prediction techniques have emerged. However, they primarily focus on modeling relationships of a single type, overlooking the importance of gaining insights into molecular behaviors from a complete regulatory network perspective and discovering biomarkers of unknown types. Furthermore, effectively handling local and global topological structural information of nodes in biological molecular regulatory graphs remains a challenge to improving biomarker prediction performance. To address these limitations, we propose a multichannel graph neural network based on multisimilarity modality hypergraph contrastive learning (MML-MGNN) for predicting unknown types of cancer biomarkers. MML-MGNN leverages multisimilarity modality hypergraph contrastive learning to delve into local associations in the regulatory network, learning diverse insights into the topological structures of multiple types of similarities, and then globally modeling the multisimilarity modalities through a multichannel graph autoencoder. By combining representations obtained from local-level associations and global-level regulatory graphs, MML-MGNN can acquire molecular feature descriptors benefiting from multitype association properties and the complete regulatory network. Experimental results on predicting three different types of cancer biomarkers demonstrate the outstanding performance of MML-MGNN. Furthermore, a case study on gastric cancer underscores the outstanding ability of MML-MGNN to gain deeper insights into molecular mechanisms in regulatory networks and prominent potential in cancer biomarker prediction.

Identifying drug-target interactions (DTIs) holds significant importance in drug discovery and development, playing a crucial role in various areas such as virtual screening, drug repurposing and identification of potential drug side effects. However, existing methods commonly exploit only a single type of feature from drugs and targets, suffering from miscellaneous challenges such as high sparsity and cold-start problems. We propose a novel framework called MSI-DTI (Multi-Source Information-based Drug-Target Interaction Prediction) to enhance prediction performance, which obtains feature representations from different views by integrating biometric features and knowledge graph representations from multi-source information. Our approach involves constructing a Drug-Target Knowledge Graph (DTKG), obtaining multiple feature representations from diverse information sources for SMILES sequences and amino acid sequences, incorporating network features from DTKG and performing an effective multi-source information fusion. Subsequently, we employ a multi-head self-attention mechanism coupled with residual connections to capture higher-order interaction information between sparse features while preserving lower-order information. Experimental results on DTKG and two benchmark datasets demonstrate that our MSI-DTI outperforms several state-of-the-art DTIs prediction methods, yielding more accurate and robust predictions. The source codes and datasets are publicly accessible at https://github.com/KEAML-JLU/MSI-DTI.

Node classification is an essential problem in graph learning. However, many models typically obtain unsatisfactory performance when applied to few-shot scenarios. Some studies have attempted to combine meta-learning with graph neural networks to solve few-shot node classification on graphs. Despite their promising performance, some limitations remain. First, they employ the node encoding mechanism of homophilic graphs to learn node embeddings, even in heterophilic graphs. Second, existing models based on meta-learning ignore the interference of randomness in the learning process. Third, they are trained using only limited labeled nodes within the specific task, without explicitly utilizing numerous unlabeled nodes. Finally, they treat almost all sampled tasks equally without customizing them for their uniqueness. To address these issues, we propose a novel framework for few-shot node classification called Meta-GPS \(++\) . Specifically, we first adopt an efficient method to learn discriminative node representations on homophilic and heterophilic graphs. Then, we leverage a prototype-based approach to initialize parameters and contrastive learning for regularizing the distribution of node embeddings. Moreover, we apply self-training to extract valuable information from unlabeled nodes. Additionally, we adopt S \({}^{2}\) (scaling and shifting) transformation to learn transferable knowledge from diverse tasks. The results on real-world datasets show the superiority of Meta-GPS \(++\) . Our code is available here .

Wasserstein barycenter, built on the theory of Optimal Transport (OT), provides a powerful framework to aggregate probability distributions, and it has increasingly attracted great attention within the machine learning community. However, it is often intractable to precisely compute, especially for high dimensional and continuous settings. To alleviate this problem, we develop a novel regularization by using the fact that c-cyclical monotonicity is often necessary and sufficient conditions for optimality in OT problems, and incorporate it into the dual formulation of Wasserstein barycenters. For efficient computations, we adopt a variational distribution as the approximation of the true continuous barycenter, so as to frame the Wasserstein barycenters problem as an optimization problem with respect to variational parameters. Upon those ideas, we propose a novel end-to-end continuous approximation method, namely Variational Wasserstein Barycenters with c-Cyclical Monotonicity Regularization (VWB-CMR), given sample access to the input distributions. We show theoretical convergence analysis and demonstrate the superior performance of VWB-CMR on synthetic data and real applications of subset posterior aggregation.

Text clustering is a critical step in text data analysis and has been extensively studied by the text mining community. Most existing text clustering algorithms are based on the bag-of-words model, which faces the high-dimensional and sparsity problems and ignores text structural and sequence information. Deep learning-based models such as convolutional neural networks and recurrent neural networks regard texts as sequences but lack supervised signals and explainable results. In this paper, we propose a deep feature-based text clustering (DFTC) framework that incorporates pretrained text encoders into text clustering tasks. This model, which is based on sequence representations, breaks the dependency on supervision. The experimental results show that our model outperforms classic text clustering algorithms on almost all the considered datasets. In addition, the explanation of the clustering results is significant for understanding the principles of the deep learning approach. Our proposed clustering framework includes an explanation module that can help users understand the meaning and quality of the clustering results. Our code is available at https://github.com/KEAML-JLU/DeepTextClustering.

Many models that leverage knowledge graphs (KGs) have recently demonstrated remarkable success in question answering (QA) tasks. In the real world, many facts contained in KGs are time-constrained thus temporal KGQA has received increasing attention. Despite the fruitful efforts of previous models in temporal KGQA, they still have several limitations. (I) They neither emphasize the graph structural information between entities in KGs nor explicitly utilize a multi-hop relation path through graph neural networks to enhance answer prediction. (II) They adopt pre-trained language models (LMs) to obtain question representations, focusing merely on the global information related to the question while not highlighting the local information of the entities in KGs. To address these limitations, we introduce a novel model that simultaneously explores both Local information and Global information for the task of temporal KGQA (LGQA). Specifically, we first introduce an auxiliary task in the temporal KG embedding procedure to make timestamp embeddings time-order aware. Then, we design information fusion layers that effectively incorporate local and global information to deepen question understanding. We conduct extensive experiments on two benchmarks, and LGQA significantly outperforms previous state-of-the-art models, especially in difficult questions. Moreover, LGQA can generate interpretable and trustworthy predictions.

Semantic segmentation is one of the most challenging tasks for very high resolution (VHR) remote sensing applications. Deep convolutional neural networks (CNN) based on the attention mechanism have shown outstanding performance in VHR remote sensing images semantic segmentation. However, existing attention-guided methods require the estimation of a large number of parameters that are affected by the limited number of available labeled samples that results in underperforming segmentation results. In this paper, we propose a multi-scale feature fusion lightweight model (MSFFL) to greatly reduce the number of parameters and improve the accuracy of semantic segmentation. In this model, two parallel enhanced attention modules, i.e., the spatial attention module (SAM) and the channel attention module (CAM) are designed by introducing encoding position information. Then a covariance calculation strategy is adopted to recalibrate the generated attention maps. The integration of enhanced attention modules into the proposed lightweight module results in an efficient lightweight attention network (LiANet). The performance of the proposed LiANet is assessed on two benchmark datasets. Experimental results demonstrate that LiANet can achieve promising performance with a small number of parameters.

The cold-start problem is a major factor that limits the effectiveness of recommendation systems. Having too few available interaction records brings a series of challenges when predicting user preferences. At present, there are two main kinds of strategies for solving this problem from different perspectives. One is cross-domain recommendation (CDR), which introduces additional information by domain knowledge propagation with transfer learning. However, CDR methods follow traditional training processes in machine learning and cannot solve this typical few-shot problem from the perspective of optimization. The other type of methods that has recently emerged is based on meta-learning. Most of these approaches focus only on generating a meta-model to perform better on new tasks and ignore improvements based on cross-domain information. Therefore, it is necessary to design a novel approach to solve this problem with both domain knowledge and meta-optimization. To achieve this goal, a novel cross-domain meta-learner for cold-start recommendation (MetaCDR) is proposed. In MetaCDR, we design a domain knowledge meta-transfer module to connect different domain networks. In addition, we introduce a pretraining strategy to ensure its efficiency. The experimental results show that MetaCDR performs significantly better than state-of-the-art models in a variety of scenarios.

User cold-start is a major challenge in building personalized recommender systems. Due to the lack of sufficient interactions, it is difficult to effectively model new users. One of the main solutions is to obtain an initial model through meta-learning (mainly gradient-based methods) and adapt it to new users with a few steps of gradient descent. Although these methods have achieved remarkable performance, they are still far from being usable in real-world applications due to their high-demand data processing, heavy computational burden, and inability to perform effective user-incremental update. In this paper, we propose a d eployable and c ontinuable m eta-learning-based r ecommendation (DCMR) approach, which can achieve fast user-incremental updating with task replay and first-order gradient descent. Specifically, we introduce a dual-constrained task sampler, distillation-based loss functions, and an adaptive controller in this framework to balance the trade-off between stability and plasticity in updating. In summary, DCMR can be updated while serving new users; in other words, it learns continuously and rapidly from a sequential user stream and is able to make recommendations at any time. The extensive experiments conducted on three benchmark datasets illustrate the superiority of our model.

Attributed networks, as a manifestation of data in non-Euclidean domains, have a wide range of applications in the real world, such as molecular property prediction, social network analysis and anomaly detection. Node classification, as a fundamental research problem in attributed networks, has attracted increasing attention among research communities. However, most existing models cannot be directly applied to the data with limited labeled instances (\textiti.e., the few-shot scenario). Few-shot node classification on attributed networks is gradually becoming a research hotspot. Although several methods aim to integrate meta-learning with graph neural networks to address this problem, some limitations remain. First, they all assume node representation learning using graph neural networks in homophilic graphs. %Hence, suboptimal performance is obtained when these models are applied to heterophilic graphs. Second, existing models based on meta-learning entirely depend on instance-based statistics. %which in few-shot settings are unavoidably degraded by data noise or outliers. Third, most previous models treat all sampled tasks equally and fail to adapt their uniqueness. %which has a significant impact on the overall performance of the model. To solve the above three limitations, we propose a novel graph Meta -learning framework called G raph learning based on P rototype and S caling & shifting transformation (Meta-GPS ). More specifically, we introduce an efficient method for learning expressive node representations even on heterophilic graphs and propose utilizing a prototype-based approach to initialize parameters in meta-learning. Moreover, we also leverage S$^2$ (scaling & shifting) transformation to learn effective transferable knowledge from diverse tasks. Extensive experimental results on six real-world datasets demonstrate the superiority of our proposed framework, which outperforms other state-of-the-art baselines by up to 13% absolute improvement in terms of related metrics.

User cold-start recommendation is a serious problem that limits the performance of recommender systems (RSs). Recent studies have focused on treating this issue as a few-shot problem and seeking solutions with model-agnostic meta-learning (MAML). Such methods regard making recommendations for one user as a task and adapt to new users with a few steps of gradient updates on the meta-model. However, none of those methods consider the limitation of user representation learning imposed by the special task setting of MAML-based RSs. And they learn a common meta-model for all users while ignoring the implicit grouping distribution induced by the correlation differences among users. In response to the above problems, we propose a pretrained network modulation and task adaptation approach (PNMTA) for user cold-start recommendation. In the pretraining stage, a pretrained model is obtained with non-meta-learning methods to achieve better user representation and generalization, which can also transfer the learned knowledge to the meta-learning stage for modulation. During the meta-learning stage, an encoder modulator is utilized to realize the memorization and correction of prior parameters for the meta-learning task, and a predictor modulator is introduced to condition the model initialization on the task identity for adaptation steps. In addition, PNMTA can also make use of the existing non-cold-start users for pretraining. Comprehensive experiments on two benchmark datasets demonstrate that our model can achieve significant and consistent improvements against other state-of-the-art methods.

Coronavirus disease 2019 (COVID-19) has infected hundreds of millions of people and killed millions of them. As an RNA virus, COVID-19 is more susceptible to variation than other viruses. Many problems involved in this epidemic have made biosafety and biosecurity (hereafter collectively referred to as 'biosafety') a popular and timely topic globally. Biosafety research covers a broad and diverse range of topics, and it is important to quickly identify hotspots and trends in biosafety research through big data analysis. However, the data-driven literature on biosafety research discovery is quite scant. We developed a novel topic model based on latent Dirichlet allocation, affinity propagation clustering and the PageRank algorithm (LDAPR) to extract knowledge from biosafety research publications from 2011 to 2020. Then, we conducted hotspot and trend analysis with LDAPR and carried out further studies, including annual hot topic extraction, a 10-year keyword evolution trend analysis, topic map construction, hot region discovery and fine-grained correlation analysis of interdisciplinary research topic trends. These analyses revealed valuable information that can guide epidemic prevention work: (1) the research enthusiasm over a certain infectious disease not only is related to its epidemic characteristics but also is affected by the progress of research on other diseases, and (2) infectious diseases are not only strongly related to their corresponding microorganisms but also potentially related to other specific microorganisms. The detailed experimental results and our code are available at https://github.com/KEAML-JLU/Biosafety-analysis.

Due to the scarcity of labeled samples, clustering in hyperspectral images (HSIs) has a great potential and application value. However, current clustering methods are mainly pixel-level techniques that neglect the large spectral variability of a scene and suffer from massive time and memory consumption when dealing with large HSIs. In this article, we propose a superpixel-level global and local similarity graph-based clustering (SGLSC) algorithm that can classify ground objects exploiting spectral and spatial dimensions with reasonable time and memory consumption on large HSIs. The proposed SGLSC exploits the superpixel concept, which is treated as a homogeneous entity, into the clustering process. For modeling the essential structure of HSIs, a similarity graph combing the global and local information is constructed and inserted into the spectral clustering to partition the superpixel-level graph structure. The proposed method was tested on three benchmark HSIs’ datasets and compared with some advanced literature algorithms. Experiments demonstrate that it can obtain promising results.

Text clustering is a critical step in text data analysis and has been extensively studied by the text mining community. Most existing text clustering algorithms are based on the bag-of-words model, which faces the high-dimensional and sparsity problems and ignores text structural and sequence information. Deep learning-based models such as convolutional neural networks and recurrent neural networks regard texts as sequences but lack supervised signals and explainable results. In this paper, we propose adeepfeature-basedtextclustering (DFTC) framework that incorporates pretrained text encoders into text clustering tasks. This model, which is based on sequence representations, breaks the dependency on supervision. The experimental results show that our model outperforms classic text clustering algorithms and the state-of-the-art pretrained language model, i.e., BERT, on almost all the considered datasets. In addition, the explanation of the clustering results is significant for understanding the principles of the deep learning approach. Our proposed clustering framework includes an explanation module that can help users understand the meaning and quality of the clustering results.

Medical scene graph contributes to cognitive tasks such as question answering. An automatic medical scene graph generator can annotate medical images with scene graphs conveniently. An end-to-end model is proposed, which contains an automatic medical scene graph generator based on medical images and surgery records. The novel model can extract symptom information of the images from surgery records via name entity recognition then convert it into scene graphs. The content and format of the scene graph are pre-defined according to the features of the images. A medical scene graph (MSG) dataset is also produced. The dataset contains 116 scene graphs where each graph includes 6 to 30 entities. Experiments show that our model performs well on making scene graphs of CT coronary angiography. Significantly, the scene graphs have the type, state, location, and index of arteries. To verify the practicability of the MSG dataset, We construct an intelligent model that can generate key diagnostic results. In addition, our medical scene graph generator can be applied to other medical areas to annotate different kinds of medical images intelligently. And the MSG dataset can directly be used in other cognitive visual tasks.

The Traditional Chinese Medicine’s ancient literature recorded the massive medical theories and abundant medical experiences. To better understand and utilize, the knowledge from the literature, the Acupuncture and Tuina Knowledge Graph is proposed in this paper. Meanwhile, a deep learning network is established for acupuncture and tuina-related entity recognition and entity-relationship extraction. Finally, the trained network is able to reach an 82%+ F1-score for NER and 70%+ F1-score for relationship extraction.

The worldwide spread of COVID-19 has made a severe impact on human health and life. It has shown rapid propagation, long in vitro survival, and a long incubation period. More seriously, COVID-19 is more susceptible to variation, as it is an RNA virus. Mutations of COVID-19 have been reported in multiple countries worldwide, which makes drug and vaccine development a significant challenge. To search for potential drugs and vaccines and reveal the atlas of COVID-19 evolution, we extract information from massive unstructured data and construct a COVID-19 knowledge graph using the COVID-19 data. Based on machine learning approaches, we infer and predict novel coronavirus pneumonia-related diseases, drug action targets, etc. to speculate on new and more effective treatment methods. In addition, to study transcriptome of SARS-CoV-2, new ideas can be provided to biomedical experts with flexible responses to viral variation. An in-depth analysis of the COVID-19 pathomechanism at the pharmaceutical, genetic, and protein levels provides effective means and tools for novel coronavirus pneumonia vaccines, drug development, and therapeutic program design.

Paper-publication venue prediction aims to predict candidate publication venues that effectively suit given submissions. This technology is developing rapidly with the popularity of machine learning models. However, most previous methods ignore the structure information of papers, while modeling them with graphs can naturally solve this drawback. Meanwhile, they either use hand-crafted or bag-of-word features to represent the papers, ignoring the ones that involve high-level semantics. Moreover, existing methods assume that the venue where a paper is published as a correct venue for the data annotation, which is unrealistic. One paper can be relevant to many venues. In this paper, we attempt to address these problems above and develop a novel prediction model, namelyVenue Prediction with Abstract-Level Graph (Vpalg xspace), which can serve as an effective decision-making tool for venue selections. Specifically, to achieve more discriminative paper abstract representations, we construct each abstract as a semantic graph and perform a dual attention message passing neural network for representation learning. Then, the proposed model can be trained over the learned abstract representations with their labels and generalized via self-training. Empirically, we employ the PubMed dataset and further collect two new datasets from the top journals and conferences in computer science. Experimental results indicate the superior performance of Vpalg xspace, consistently outperforming the existing baseline methods.

Text classification is a fundamental task with broad applications in natural language processing.Recently, graph neural networks (GNNs) have attracted much attention due to their powerful representation ability.However, most existing methods for text classification based on GNNs consider only one-hop neighborhoods and low-frequency information within texts, which cannot fully utilize the rich context information of documents.Moreover, these models suffer from over-smoothing issues if many graph layers are stacked.In this paper, a Deep Attention Diffusion Graph Neural Network (DADGNN) model is proposed to learn text representations, bridging the chasm of interaction difficulties between a word and its distant neighbors.Experimental results on various standard benchmark datasets demonstrate the superior performance of the present approach.

Band selection (BS) can mitigate the “curse of dimensionality” problem and improve the performance of hyperspectral image (HSI) classification. Genetic algorithms (GAs) have been applied to the task of hyperspectral BS showing significant advantages compared with other literature methods. However, the traditional GAs-based methods often select sets of bands having residual redundancy due to the large search space related to hyperspectral BS and the limitation of premature convergence in GAs. Moreover, existing GAs-based methods often are supervised, and that needs a large number of labeled samples to compute the fitness value for assessing the quality of selected bands. In this article, an unsupervised BS approach based on an improved GA is proposed. A fitness function based on the fisher score combined with superpixel is designed for evaluating the discriminability of band subsets considering both spectral and spatial information. Then, modified genetic operations are constructed to restrain the search space and reduce the redundancy of selected bands. The performance of the proposed spectral-spatial GA-based BS method is evaluated on three HSIs. The experimental results demonstrate that the proposed method is superior to the traditional GA-based method and seven state-of-the-art unsupervised methods.

Large-scale knowledge graphs have currently reached impressive sizes; however, they are still far from complete. In addition, most existing methods for knowledge graph completion only consider the direct links between entities, ignoring the vital impact of the semantics of relation paths. In this paper, we study the problem of how to better embed entities and relations of knowledge graphs into different low-dimensional spaces by taking full advantage of the additional semantics of relation paths and propose a novel relation path embedding model named as RPE. Specifically, with the corresponding relation and path projections, RPE can simultaneously embed each entity into two types of latent spaces. Moreover, type constraints are extended from traditional relation-specific type constraints to the proposed path-specific type constraints and both of the two type constraints can be seamlessly incorporated into RPE. The proposed model is evaluated on the benchmark tasks of link prediction and triple classification. The results of experiments demonstrate our method outperforms all baselines on both tasks. They indicate that our model is capable of catching the semantics of relation paths, which is significant for knowledge representation learning.

Automatically describing contents of an image using natural language has drawn much attention because it not only integrates computer vision and natural language processing but also has practical applications. Using an end-to-end approach, we propose a bidirectional semantic attention-based guiding of long short-term memory (Bag-LSTM) model for image captioning. The proposed model consciously refines image features from previously generated text. By fine-tuning the parameters of convolution neural networks, Bag-LSTM obtains more text-related image features via feedback propagation than other models. As opposed to existing guidance-LSTM methods which directly add image features into each unit of an LSTM block, our fine-tuned model dynamically leverages more text-conditional image features, acquired by the semantic attention mechanism, as guidance information. Moreover, we exploit bidirectional gLSTM as the caption generator, which is capable of learning long term relations between visual features and semantic information by making use of both historical and future contextual information. In addition, variations of the Bag-LSTM model are proposed in an effort to sufficiently describe high-level visual-language interactions. Experiments on the Flickr8k and MSCOCO benchmark datasets demonstrate the effectiveness of the model, as compared with the baseline algorithms, such as it is 51.2% higher than BRNN on CIDEr metric.

As computer science and information technology are making broad and deep impacts on our daily lives, more and more papers are being submitted to computer science journals and conferences. To help authors decide where they should submit their manuscripts, we present the Content-based Journals & Conferences Recommender System on computer science, as well as its web service at http://www.keaml.cn/prs/ . This system recommends suitable journals or conferences with a priority order based on the abstract of a manuscript. To follow the fast development of computer science and technology, a web crawler is employed to continuously update the training set and the learning model. To achieve interactive online response, we propose an efficient hybrid model based on chi-square feature selection and softmax regression. Our test results show that, the system can achieve an accuracy of 61.37% and suggest the best journals or conferences in about 5 s on average.

This paper presents an unsupervised approach to band selection in hyperspectral images that considers both spectral and spatial information in data dimensionality reduction. The approach exploits the concepts of superpixel and chunklets for identifying the spectral channels most suitable to be used in classification for discriminating land-cover classes. The segmented superpixels can be regarded as many small spectral homogeneous and spatial neighboring pixel chunklets. Based on the observation that the superpixel chunklets achieve high homogeneity and consistency within land-cover classes, a series of band criteria (BC) is identified by learning the optimal band transformation that results in low within-class variability and high total variability. Then, the learned BC, which are called band measures, are given in input to an efficient clustering algorithm, i.e., the affinity propagation, for selecting highly separable bands with low redundancy. The effectiveness of proposed approach was assessed on three hyperspectral data sets. The results point out the advantages of the proposed methods over five state-of-the-art unsupervised methods.

Diabetes mellitus and obesity are becoming some of the most serious public health challenges in the world. To help researchers more quickly reveal the complex relationships existing between diabetes mellitus, obesity, and related diseases in the literature, and give them an inspiration to search the effective treatments for these diseases, we propose a novel model named as representative latent Dirichlet allocation topic model (RLDA). We conducted the representation learning model on more than 337,000 pieces of diabetes and obesity related literature published in the recent decade. Then, an explicit analysis of the final result using a series of visualization tools to discover meaningful relations among diabetes mellitus, obesity, and other diseases was performed. In order to show the credibility of our discoveries, we used clinical reports, such as Standards of Medical Care in Diabetes, which were not used in our training data, to verify our results. Fortunately, a sufficient number of the reports were direct matches. With the help of our model, we achieved satisfactory results for diabetes mellitus and obesity. For example, we discovered that 22 other diseases are closely related to diabetes mellitus, 10 with obesity and 8 with both. In addition, the tumor, adolescent/child, inflammation, and hypertension will be the hottest research topics relating to diabetes and obesity in the near future. We believe that the representational learning model we have built can help biomedical researchers direct the focus and adjust the direction of their work.

Pulmonary adenocarcinoma in situ (AIS) is an intermediate subtype of lung adenocarcinoma that exhibits non-invasive growth patterns, but can develop into invasive. Almost 100% of AIS patients can be cured with complete resection. In contrast, the five-year survival rate for those diagnosed with invasive lung adenocarcinoma is only about 4%. In order to get a better understanding of adenocarcinoma and identify marks indicating its evaluation and progression, what needs to be done is a genome-wide evaluation of the disease. In this study, we used RNA-seq data from normal, AIS, and invasive lung cancer samples to identify gene module with differential gene expressions that represents the properties of AIS, distinct from normal and invasive tumor. Utilizing the differential expression patterns of protein-coding genes and long non-coding RNAs (lncRNAs) between AIS and other conditions, we obtained a small group of 72 AIS-specific genes consisting 41 protein-coding genes, 5 annotated lncRNAs, and 26 novel lncRNA transcripts that expressed specifically in AIS samples. We consider that twelve of the protein-coding genes are lung cancer driver genes with located driver somatic mutations. Moreover, these AIS-specific genes show capabilities (98% accuracy in an independent data set) for indicating early stage lung cancer and normal situations. These genes are determined based on cell-adhesion functioning, including angiogenesis and fibronectin, etc., that are highly related to cancer development. The comparison of AIS with normal and invasive tumor provides the gene list revealing the mechanisms that accounts for AIS progression to invasive cancer. Furthermore we identified important signatures contributing to the early diagnosis of lung cancer, providing auxiliary studies to our ongoing precision medicine research (http://americancse.org/events/csce2017/keynotes_lectures/yang_talk).

Thanks to recent advances in the field of genomics, it is now possible to create a comprehensive atlas of the basic units of life-cells. In this paper, we present a frame work for single cell genomics research which employs several new machine learning models such as convolutional neural networks, deep auto-encoder, recurrent neural networks etc. With these effective learning models on multi-source data, such as biomedical literatures and cell images, we can achieve novel cell types and functional gene sets.

Breast cancer is the most common type of invasive cancer in females. It accounts for 18.2% of all cancer deaths worldwide. Although somatic mutations play important roles in cancer development and prognosis, the outcome predictions are largely based on the expression of marker genes. We submit that developing an innovative prognostic model incorporating somatic mutations with gene expression can improve survival prediction of cancer. We studied the whole genome-wide gene expression and somatic mutations of 1091 breast invasive carcinoma cases from The Cancer Genome Atlas (TCGA). We identified expression of 118 genes that could be used to build the predictor for breast cancer survival risks (log rank p <; 0.0001 and c-index=0.63627122). Multiple breast cancer survival-analysis-related genes are found in this gene set, such as FOXR2, FOXD1, MTNR1B, SDC1, PF4, IGF2BP1, ZIC3, OXT2, and PID1. We then selected between different survival-risk groups of 2000 mutated genes with different mutation rates. We applied enrichment analysis to the mutated gene list and identified 25 functional annotations, 15 gene ontology, and 14 gene pathway enriched terms that were related the cancer outcomes. We built the novel predictor and our results showed that combining the different features helps improved performance of survival prediction (c-index = 0. 64033769). Thus our model can be used to facilitate the advancement of our going precision medicine research (http://americancse.org/events/csce2017/keynotes_lectures/yang_talk).

Relevant component analysis has shown effective in metric learning. It finds a transformation matrix of the feature space using equivalence constraints. This paper explores this idea for constructing a feature metric (FM) and develops a novel semisupervised feature-selection technique for hyperspectral image classification. Two feature measures referred to as band correlation metric (BCM) and band separability metric (BSM) are derived for the FM. The BCM can measure the spectral correlation among the bands, while the BSM can assess the class discrimination capability of a single band. The proposed feature-metric-based affinity propagation (AP) (FM-AP) technique utilizes exemplar-based clustering, i.e., AP, to group bands from original spectral channels with the FM. Experimental results are conducted on two hyperspectral images and show the advantages of the proposed technique over traditional feature-selection methods.

Clustering is used for land-cover identification in remote sensing images when training data are not available. However, in many applications, it is often possible to collect a small number of labeled samples. To effectively exploit this small number of labeled samples combined with a multitude of the unlabeled data, we present a novel semisupervised clustering technique [incremental and decremental affinity propagation (ID-AP)] that incorporates labeled exemplars into the AP algorithm. Unlike standard semisupervised clustering methods, the proposed technique improves the performance by using both the labeled samples to adjust the similarity matrix and an ID-learning principle for unlabeled data selection and useless labeled samples rejection, respectively. This avoids both learning-bias and stability-plasticity dilemma. In order to assess the effectiveness of the proposed ID-AP technique, the experimental analysis was carried out on three different kinds of multispectral images including different percentages of labeled samples. In the analysis, we also studied the accuracy and the stability of two semisupervised clustering algorithms [i.e., constrainedk-means and semisupervised AP (SAP)] and one incremental semisupervised clustering algorithm (i.e., incremental SAP). Experimental results demonstrate that the proposed ID-AP technique adequately captures and takes full advantage of the intrinsic relationship between the labeled samples and unlabeled data, and produces better performance than the other considered methods.

Based on an effective clustering algorithm-Affinity Propagation (AP)-we present in this paper a novel semisupervised text clustering algorithm, called Seeds Affinity Propagation (SAP). There are two main contributions in our approach: 1) a new similarity metric that captures the structural information of texts, and 2) a novel seed construction method to improve the semisupervised clustering process. To study the performance of the new algorithm, we applied it to the benchmark data set Reuters-21578 and compared it to two state-of-the-art clustering algorithms, namely, k-means algorithm and the original AP algorithm. Furthermore, we have analyzed the individual impact of the two proposed contributions. Results show that the proposed similarity metric is more effective in text clustering (F-measures ca. 21 percent higher than in the AP algorithm) and the proposed semisupervised strategy achieves both better clustering results and faster convergence (using only 76 percent iterations of the original AP). The complete SAP algorithm obtains higher F-measure (ca. 40 percent improvement over k-means and AP) and lower entropy (ca. 28 percent decrease over k-means and AP), improves significantly clustering execution time (20 times faster) in respect that k-means, and provides enhanced robustness compared with all other methods.

Due to a high number of spectral channels and a large information quantity, multispectral remote-sensing images are difficult to be classified with high accuracy and efficiency by conventional classification methods, particularly when training data are not available and when unsupervised clustering techniques should be considered for data analysis. In this paper, we propose a novel image clustering method [called fuzzy-statistics-based affinity propagation (FS-AP)] which is based on a fuzzy statistical similarity measure (FSS) to extract land-cover information in multispectral imagery. AP is a clustering algorithm proposed recently in the literature, which exhibits a fast execution speed and finds clusters with small error, particularly for large datasets. FSS can get objective estimates of how closely two pixel vectors resemble each other. The proposed method simultaneously considers all data points to be equally suitable as initial exemplars, thus reducing the dependence of the final clustering from the initialization. Results obtained on three kinds of multispectral images (Landsat-7 ETM+, Quickbird, and moderate resolution imaging spectroradiometer) by comparing the proposed technique with K-means, fuzzy K-means, and AP based on Euclidean distance (ED-AP) demonstrate the good efficiency and high accuracy of FS-AP.

Affinity propagation is an impressive clustering algorithm which was published in Science, 2007. However, the original algorithm couldn't cope with part known data directly. Focusing on this issue, a semi-supervised scheme called incremental affinity propagation clustering is proposed in the paper. In the scheme, the pre-known information is represented by adjusting similarity matrix. Moreover, an incremental study is applied to amplify the prior knowledge. To examine the effectiveness of the method, we concentrate it to text clustering problem and describe the specific method accordingly. The method is applied to the benchmark data set Reuters-21578. Numerical results show that the proposed method performs very well on the data set and has most advantages over two other commonly used clustering methods.

Principal component analysis (PCA) is a favorite multivariate statistical method for image enhancement and compression. However, it is well known that the classical PCA is sensitive to outliers and missing data. Fortunately, fuzzy statistics is an effective theory for processing these kinds of data. Fuzziness and randomicity are just the important characteristics of the data of remote-sensing images. Therefore, by introducing fuzzy statistics variables into classical PCA methods, a novel method for multispectral image processing called fuzzy-statistics-based PCA (FS-PCA) is proposed in this paper. To verify our proposed method, both the classical PCA and the FS-PCA are applied to the multispectral Landsat ETM+ data for image enhancement. The experimental results show that the differences among surface characteristics are expanded sufficiently and that the accuracy of surface feature recognition is improved greatly.