Search

The integration of multimodal Electronic Health Records (EHR) data has notably advanced clinical predictive capabilities. However, current models that utilize clinical notes and multivariate time-series EHR data often lack the necessary medical context for precise clinical tasks. Previous methods using knowledge graphs (KGs) primarily focus on structured knowledge extraction. To address this, we propose EMERGE, a Retrieval-Augmented Generation (RAG) driven framework aimed at enhancing multimodal EHR predictive modeling. Our approach extracts entities from both time-series data and clinical notes by prompting Large Language Models (LLMs) and aligns them with professional PrimeKG to ensure consistency. Beyond triplet relationships, we include entities' definitions and descriptions to provide richer semantics. The extracted knowledge is then used to generate task-relevant summaries of patients' health statuses. These summaries are fused with other modalities utilizing an adaptive multimodal fusion network with cross-attention. Extensive experiments on the MIMIC-III and MIMIC-IV datasets for in-hospital mortality and 30-day readmission tasks demonstrate the superior performance of the EMERGE framework compared to baseline models. Comprehensive ablation studies and analyses underscore the efficacy of each designed module and the framework's robustness to data sparsity. EMERGE significantly enhances the use of multimodal EHR data in healthcare, bridging the gap with nuanced medical contexts crucial for informed clinical predictions., Comment: arXiv admin note: text overlap with arXiv:2402.07016

New Intent Discovery (NID) strives to identify known and reasonably deduce novel intent groups in the open-world scenario. But current methods face issues with inaccurate pseudo-labels and poor representation learning, creating a negative feedback loop that degrades overall model performance, including accuracy and the adjusted rand index. To address the aforementioned challenges, we propose a Robust New Intent Discovery (RoNID) framework optimized by an EM-style method, which focuses on constructing reliable pseudo-labels and obtaining cluster-friendly discriminative representations. RoNID comprises two main modules: reliable pseudo-label generation module and cluster-friendly representation learning module. Specifically, the pseudo-label generation module assigns reliable synthetic labels by solving an optimal transport problem in the E-step, which effectively provides high-quality supervised signals for the input of the cluster-friendly representation learning module. To learn cluster-friendly representation with strong intra-cluster compactness and large inter-cluster separation, the representation learning module combines intra-cluster and inter-cluster contrastive learning in the M-step to feed more discriminative features into the generation module. RoNID can be performed iteratively to ultimately yield a robust model with reliable pseudo-labels and cluster-friendly representations. Experimental results on multiple benchmarks demonstrate our method brings substantial improvements over previous state-of-the-art methods by a large margin of +1~+4 points., Comment: DASFAA 2024

Log anomaly detection is a key component in the field of artificial intelligence for IT operations (AIOps). Considering log data of variant domains, retraining the whole network for unknown domains is inefficient in real industrial scenarios. However, previous deep models merely focused on extracting the semantics of log sequences in the same domain, leading to poor generalization on multi-domain logs. To alleviate this issue, we propose a unified Transformer-based framework for Log anomaly detection (LogFormer) to improve the generalization ability across different domains, where we establish a two-stage process including the pre-training and adapter-based tuning stage. Specifically, our model is first pre-trained on the source domain to obtain shared semantic knowledge of log data. Then, we transfer such knowledge to the target domain via shared parameters. Besides, the Log-Attention module is proposed to supplement the information ignored by the log-paring. The proposed method is evaluated on three public and one real-world datasets. Experimental results on multiple benchmarks demonstrate the effectiveness of our LogFormer with fewer trainable parameters and lower training costs., Comment: arXiv admin note: text overlap with arXiv:2201.00016

Fully supervised log anomaly detection methods suffer the heavy burden of annotating massive unlabeled log data. Recently, many semi-supervised methods have been proposed to reduce annotation costs with the help of parsed templates. However, these methods consider each keyword independently, which disregards the correlation between keywords and the contextual relationships among log sequences. In this paper, we propose a novel weakly supervised log anomaly detection framework, named LogLG, to explore the semantic connections among keywords from sequences. Specifically, we design an end-to-end iterative process, where the keywords of unlabeled logs are first extracted to construct a log-event graph. Then, we build a subgraph annotator to generate pseudo labels for unlabeled log sequences. To ameliorate the annotation quality, we adopt a self-supervised task to pre-train a subgraph annotator. After that, a detection model is trained with the generated pseudo labels. Conditioned on the classification results, we re-extract the keywords from the log sequences and update the log-event graph for the next iteration. Experiments on five benchmarks validate the effectiveness of LogLG for detecting anomalies on unlabeled log data and demonstrate that LogLG, as the state-of-the-art weakly supervised method, achieves significant performance improvements compared to existing methods., Comment: 12 pages

Transformer structure, stacked by a sequence of encoder and decoder network layers, achieves significant development in neural machine translation. However, vanilla Transformer mainly exploits the top-layer representation, assuming the lower layers provide trivial or redundant information and thus ignoring the bottom-layer feature that is potentially valuable. In this work, we propose the Group-Transformer model (GTrans) that flexibly divides multi-layer representations of both encoder and decoder into different groups and then fuses these group features to generate target words. To corroborate the effectiveness of the proposed method, extensive experiments and analytic experiments are conducted on three bilingual translation benchmarks and two multilingual translation tasks, including the IWLST-14, IWLST-17, LDC, WMT-14 and OPUS-100 benchmark. Experimental and analytical results demonstrate that our model outperforms its Transformer counterparts by a consistent gain. Furthermore, it can be successfully scaled up to 60 encoder layers and 36 decoder layers., Comment: Accepted in IEEE TASLP

Multilingual neural machine translation (MNMT) trained in multiple language pairs has attracted considerable attention due to fewer model parameters and lower training costs by sharing knowledge among multiple languages. Nonetheless, multilingual training is plagued by language interference degeneration in shared parameters because of the negative interference among different translation directions, especially on high-resource languages. In this paper, we propose the multilingual translation model with the high-resource language-specific training (HLT-MT) to alleviate the negative interference, which adopts the two-stage training with the language-specific selection mechanism. Specifically, we first train the multilingual model only with the high-resource pairs and select the language-specific modules at the top of the decoder to enhance the translation quality of high-resource directions. Next, the model is further trained on all available corpora to transfer knowledge from high-resource languages (HRLs) to low-resource languages (LRLs). Experimental results show that HLT-MT outperforms various strong baselines on WMT-10 and OPUS-100 benchmarks. Furthermore, the analytic experiments validate the effectiveness of our method in mitigating the negative interference in multilingual training., Comment: 7 pages, 7 figures, IJCAI-ECAI 2022

Most translation tasks among languages belong to the zero-resource translation problem where parallel corpora are unavailable. Multilingual neural machine translation (MNMT) enables one-pass translation using shared semantic space for all languages compared to the two-pass pivot translation but often underperforms the pivot-based method. In this paper, we propose a novel method, named as Unified Multilingual Multiple teacher-student Model for NMT (UM4). Our method unifies source-teacher, target-teacher, and pivot-teacher models to guide the student model for the zero-resource translation. The source teacher and target teacher force the student to learn the direct source to target translation by the distilled knowledge on both source and target sides. The monolingual corpus is further leveraged by the pivot-teacher model to enhance the student model. Experimental results demonstrate that our model of 72 directions significantly outperforms previous methods on the WMT benchmark., Comment: 7 pages, 5 figures, IJCAI-ECAI 2022

The recently proposed Graph Convolutional Networks (GCNs) have achieved significantly superior performance on various graph-related tasks, such as node classification and recommendation. However, currently researches on GCN models usually recursively aggregate the information from all the neighbors or randomly sampled neighbor subsets, without explicitly identifying whether the aggregated neighbors provide useful information during the graph convolution. In this paper, we theoretically analyze the affection of the neighbor quality over GCN models' performance and propose the Neighbor Enhanced Graph Convolutional Network (NEGCN) framework to boost the performance of existing GCN models. Our contribution is three-fold. First, we at the first time propose the concept of neighbor quality for both node classification and recommendation tasks in a general theoretical framework. Specifically, for node classification, we propose three propositions to theoretically analyze how the neighbor quality affects the node classification performance of GCN models. Second, based on the three proposed propositions, we introduce the graph refinement process including specially designed neighbor evaluation methods to increase the neighbor quality so as to boost both the node classification and recommendation tasks. Third, we conduct extensive node classification and recommendation experiments on several benchmark datasets. The experimental results verify that our proposed NEGCN framework can significantly enhance the performance for various typical GCN models on both node classification and recommendation tasks., Comment: 29 pages, 3 figures, 7 tables. Accepted to Knowledge-Based Systems

Log anomaly detection is a key component in the field of artificial intelligence for IT operations (AIOps). Considering log data of variant domains, retraining the whole network for unknown domains is inefficient in real industrial scenarios especially for low-resource domains. However, previous deep models merely focused on extracting the semantics of log sequence in the same domain, leading to poor generalization on multi-domain logs. Therefore, we propose a unified Transformer-based framework for log anomaly detection (\ourmethod{}), which is comprised of the pretraining and adapter-based tuning stage. Our model is first pretrained on the source domain to obtain shared semantic knowledge of log data. Then, we transfer the pretrained model to the target domain via the adapter-based tuning. The proposed method is evaluated on three public datasets including one source domain and two target domains. The experimental results demonstrate that our simple yet efficient approach, with fewer trainable parameters and lower training costs in the target domain, achieves state-of-the-art performance on three benchmarks., Comment: 6 pages

Graph Convolutional Networks (GCNs) are powerful models for node representation learning tasks. However, the node representation in existing GCN models is usually generated by performing recursive neighborhood aggregation across multiple graph convolutional layers with certain sampling methods, which may lead to redundant feature mixing, needless information loss, and extensive computations. Therefore, in this paper, we propose a novel architecture named Non-Recursive Graph Convolutional Network (NRGCN) to improve both the training efficiency and the learning performance of GCNs in the context of node classification. Specifically, NRGCN proposes to represent different hops of neighbors for each node based on inner-layer aggregation and layer-independent sampling. In this way, each node can be directly represented by concatenating the information extracted independently from each hop of its neighbors thereby avoiding the recursive neighborhood expansion across layers. Moreover, the layer-independent sampling and aggregation can be precomputed before the model training, thus the training process can be accelerated considerably. Extensive experiments on benchmark datasets verify that our NRGCN outperforms the state-of-the-art GCN models, in terms of the node classification performance and reliability., Comment: 5 pages, 2 figures. Accepted to ICASSP 2021

In this paper, we investigate the cause of the high false positive rate in Visual Relationship Detection (VRD). We observe that during training, the relationship proposal distribution is highly imbalanced: most of the negative relationship proposals are easy to identify, e.g., the inaccurate object detection, which leads to the under-fitting of low-frequency difficult proposals. This paper presents Spatially-Aware Balanced negative pRoposal sAmpling (SABRA), a robust VRD framework that alleviates the influence of false positives. To effectively optimize the model under imbalanced distribution, SABRA adopts Balanced Negative Proposal Sampling (BNPS) strategy for mini-batch sampling. BNPS divides proposals into 5 well defined sub-classes and generates a balanced training distribution according to the inverse frequency. BNPS gives an easier optimization landscape and significantly reduces the number of false positives. To further resolve the low-frequency challenging false positive proposals with high spatial ambiguity, we improve the spatial modeling ability of SABRA on two aspects: a simple and efficient multi-head heterogeneous graph attention network (MH-GAT) that models the global spatial interactions of objects, and a spatial mask decoder that learns the local spatial configuration. SABRA outperforms SOTA methods by a large margin on two human-object interaction (HOI) datasets and one general VRD dataset., Comment: 13 pages, 5 figures

We study open domain response generation with limited message-response pairs. The problem exists in real-world applications but is less explored by the existing work. Since the paired data now is no longer enough to train a neural generation model, we consider leveraging the large scale of unpaired data that are much easier to obtain, and propose response generation with both paired and unpaired data. The generation model is defined by an encoder-decoder architecture with templates as prior, where the templates are estimated from the unpaired data as a neural hidden semi-markov model. By this means, response generation learned from the small paired data can be aided by the semantic and syntactic knowledge in the large unpaired data. To balance the effect of the prior and the input message to response generation, we propose learning the whole generation model with an adversarial approach. Empirical studies on question response generation and sentiment response generation indicate that when only a few pairs are available, our model can significantly outperform several state-of-the-art response generation models in terms of both automatic and human evaluation., Comment: Accepted by EMNLP2019

Automatic news comment generation is a new testbed for techniques of natural language generation. In this paper, we propose a "read-attend-comment" procedure for news comment generation and formalize the procedure with a reading network and a generation network. The reading network comprehends a news article and distills some important points from it, then the generation network creates a comment by attending to the extracted discrete points and the news title. We optimize the model in an end-to-end manner by maximizing a variational lower bound of the true objective using the back-propagation algorithm. Experimental results on two datasets indicate that our model can significantly outperform existing methods in terms of both automatic evaluation and human judgment., Comment: Accepted by EMNLP2019

Recent advances in Graph Convolutional Networks (GCNs) have led to state-of-the-art performance on various graph-related tasks. However, most existing GCN models do not explicitly identify whether all the aggregated neighbors are valuable to the learning tasks, which may harm the learning performance. In this paper, we consider the problem of node classification and propose the Label-Aware Graph Convolutional Network (LAGCN) framework which can directly identify valuable neighbors to enhance the performance of existing GCN models. Our contribution is three-fold. First, we propose a label-aware edge classifier that can filter distracting neighbors and add valuable neighbors for each node to refine the original graph into a label-aware~(LA) graph. Existing GCN models can directly learn from the LA graph to improve the performance without changing their model architectures. Second, we introduce the concept of positive ratio to evaluate the density of valuable neighbors in the LA graph. Theoretical analysis reveals that using the edge classifier to increase the positive ratio can improve the learning performance of existing GCN models. Third, we conduct extensive node classification experiments on benchmark datasets. The results verify that LAGCN can improve the performance of existing GCN models considerably, in terms of node classification., Comment: Accepted by CIKM 2020

Network embedding, as a promising way of the network representation learning, is capable of supporting various subsequent network mining and analysis tasks, and has attracted growing research interests recently. Traditional approaches assign each node with an independent continuous vector, which will cause huge memory overhead for large networks. In this paper we propose a novel multi-hot compact embedding strategy to effectively reduce memory cost by learning partially shared embeddings. The insight is that a node embedding vector is composed of several basis vectors, which can significantly reduce the number of continuous vectors while maintain similar data representation ability. Specifically, we propose a MCNE model to learn compact embeddings from pre-learned node features. A novel component named compressor is integrated into MCNE to tackle the challenge that popular back-propagation optimization cannot propagate through discrete samples. We further propose an end-to-end model MCNE$_{t}$ to learn compact embeddings from the input network directly. Empirically, we evaluate the proposed models over three real network datasets, and the results demonstrate that our proposals can save about 90\% of memory cost of network embeddings without significantly performance decline.