Graph Encoding Research Articles

Reusability report: exploring the utility of variational graph encoders for predicting molecular toxicity in drug design

Reusability report: exploring the utility of variational graph encoders for predicting molecular toxicity in drug design

Fully exploring object relation interaction and hidden state attention for video captioning

Video Captioning (VC) is a challenging task of automatically generating natural language sentences for describing video contents. As a video often contains multiple objects, it is comprehensively crucial to identify multiple objects and model relationships between them. Previous models usually adopt Graph Convolutional Networks (GCN) to infer relational information via object nodes, but there exist uncertainty and over-smoothing issues of relational reasoning. To tackle these issues, we propose a Knowledge Graph based Video Captioning Network (KG-VCN) by fully exploring object relation interaction, hidden state and attention enhancement. In encoding stages, we present a Graph and Convolution Hybrid Encoder (GCHE), which uses an object detector to find visual objects with bounding boxes for Knowledge Graph (KG) and Convolutional Neural Network (CNN). To model intrinsic relations between detected objects, we propose a knowledge graph based Object Relation Graph Interaction (ORGI) module. In ORGI, we design triplets (head, relation, tail) to efficiently mine object relations, and create a global node to enable adequate information flow among all graph nodes for avoiding possibly missed relations. To produce accurate and rich captions, we propose a hidden State and Attention Enhanced Decoder (SAED) by integrating hidden states and dynamically updated attention features. Our SAED accepts both relational and visual features, adopts Long Short-Term Memory (LSTM) to produce hidden states, and dynamically update attention features. Unlike existing methods, we concatenate state and attention features to predict next word sequentially. To demonstrate the effectiveness of our model, we conduct experiments on three well-known datasets (MSVD, MSR-VTT, VaTeX), and our model achieves impressive results significantly outperforming existing state-of-the-art models.

Encoder embedding for general graph and node classification

Graph encoder embedding, a recent technique for graph data, offers speed and scalability in producing vertex-level representations from binary graphs. In this paper, we extend the applicability of this method to a general graph model, which includes weighted graphs, distance matrices, and kernel matrices. We prove that the encoder embedding satisfies the law of large numbers and the central limit theorem on a per-observation basis. Under certain condition, it achieves asymptotic normality on a per-class basis, enabling optimal classification through discriminant analysis. These theoretical findings are validated through a series of experiments involving weighted graphs, as well as text and image data transformed into general graph representations using appropriate distance metrics.

Event Camera Denoising Using Asynchronous Spatio-Temporal Event Denoising Neural Network

Abstract. Compared to conventional cameras, event cameras represent a noteworthy advancement in neuromorphic imaging technology, garnering considerable attention from researchers due to their distinct advantages. However, event cameras are susceptible to significant levels of measurement noise, which can detrimentally affect the performance of algorithms reliant on event stream for tasks such as perception and navigation. In this study, we introduce a novel method for denoising event stream, aiming to filter out events that do not accurately reflect genuine logarithmic intensity changes within the observed scene. Our approach focuses on the asynchronous nature and spatiotemporal properties of events, culminating in the development of a novel Asynchronous Spatio- Temporal Event Denoising neural Network(ASTEDNet). This network operates directly on event streams, circumventing the need to convert event stream into denser formats like image frames, thereby preserving their inherent asynchronous nature. Drawing upon principles from graph encoding and temporal convolutional networks, we incorporate spatiotemporal feature attention mechanisms to capture the temporal and spatial correlations between events. This enables the classification of each active event pixel in the original stream as either representing a genuine intensity change or noise. Comparative evaluations conducted on multiple datasets against state-of-the-art methods demonstrate the remarkable efficacy and robustness of our proposed algorithm in noise removal while retaining meaningful event information within the scene.

Remote Sensing LiDAR and Hyperspectral Classification with Multi-Scale Graph Encoder–Decoder Network

The rapid development of sensor technology has made multi-modal remote sensing data valuable for land cover classification due to its diverse and complementary information. Many feature extraction methods for multi-modal data, combining light detection and ranging (LiDAR) and hyperspectral imaging (HSI), have recognized the importance of incorporating multiple spatial scales. However, effectively capturing both long-range global correlations and short-range local features simultaneously on different scales remains a challenge, particularly in large-scale, complex ground scenes. To address this limitation, we propose a multi-scale graph encoder–decoder network (MGEN) for multi-modal data classification. The MGEN adopts a graph model that maintains global sample correlations to fuse multi-scale features, enabling simultaneous extraction of local and global information. The graph encoder maps multi-modal data from different scales to the graph space and completes feature extraction in the graph space. The graph decoder maps the features of multiple scales back to the original data space and completes multi-scale feature fusion and classification. Experimental results on three HSI-LiDAR datasets demonstrate that the proposed MGEN achieves considerable classification accuracies and outperforms state-of-the-art methods.

Predicting functional connectivity network from routinely acquired T1-weighted imaging-based brain network by generative U-GCNet

Predicting the function magnetic resonance imaging (fMRI)-based brain network (fMRI-BN) from structure MRI-based brain network is imperative in clinical practice because fMRIs are not routinely acquired in a vast majority of hospitals. In this study, a generative U-GCNet (U-shaped graph convolutional Network) was proposed to predict fMRI-BN from radiomics-based morphological brain network (radMBN) derived from routinely acquired T1-WI image. Specifically, the U-GCNet consisted of a graph convolutional network (GCN) encoder module (En-GCN), a deep feature connectivity construction module (DF2C), and a GCN decoder module (De-GCN). Both En-GCN and De-GCN employed mixed local-and-long distance node feature aggregation strategy to enhance the graph encoding and decoding ability, and the DF2C reshaped the deep feature matrix into the connectivity matrix for outputting the brain network prediction. Additionally, a multi-scale network similarity loss function was conducted on full values, upper triangular values, and each row values of connectivity matrix. Experiments on 3169 subjects from three publicly available databases demonstrated that the U-GCNet could predict the fMRI-BN from radMBN with a promising performance (MSE [0.0002 0.0025], PCC [0.956 0.991]) over eight alternative methods under comparison. The results exhibited a significant correlation (PCC [0.796, 0.897], P<0.05) between the estimated and real radiomics-function coupling values. The individual-level and group-level brain network visualization was displayed with high consistency. The TOP brain regions identified by four graph-based metrics also exhibited with consistency. These results demonstrated that the proposed U-GCNet could achieve promising prediction of fMRI-BN from radMBN which could alleviate the limited availability of fMRI and boost its usage in clinical practice.

A novel hierarchical network-based approach to unveil the complexity of functional microbial genome

Biological networks serve a crucial role in elucidating intricate biological processes. While interspecies environmental interactions have been extensively studied, the exploration of gene interactions within species, particularly among individual microorganisms, is less developed. The increasing amount of microbiome genomic data necessitates a more nuanced analysis of microbial genome structures and functions. In this context, we introduce a complex structure using higher-order network theory, “Solid Motif Structures (SMS)”, via a hierarchical biological network analysis of genomes within the same genus, effectively linking microbial genome structure with its function. Leveraging 162 high-quality genomes of Microcystis, a key freshwater cyanobacterium within microbial ecosystems, we established a genome structure network. Employing deep learning techniques, such as adaptive graph encoder, we uncovered 27 critical functional subnetworks and their associated SMSs. Incorporating metagenomic data from seven geographically distinct lakes, we conducted an investigation into Microcystis’ functional stability under varying environmental conditions, unveiling unique functional interaction models for each lake. Our work compiles these insights into an extensive resource repository, providing novel perspectives on the functional dynamics within Microcystis. This research offers a hierarchical network analysis framework for understanding interactions between microbial genome structures and functions within the same genus.

An attribution graph-based interpretable method for CNNs

Convolutional Neural Networks (CNNs) have demonstrated outstanding performance in various domains, such as face recognition, object detection, and image segmentation. However, the lack of transparency and limited interpretability inherent in CNNs pose challenges in fields such as medical diagnosis, autonomous driving, finance, and military applications. Several studies have explored the interpretability of CNNs and proposed various post-hoc interpretable methods. The majority of these methods are feature-based, focusing on the influence of input variables on outputs. Few methods undertake the analysis of parameters in CNNs and their overall structure. To explore the structure of CNNs and intuitively comprehend the role of their internal parameters, we propose an Attribution Graph-based Interpretable method for CNNs (AGIC) which models the overall structure of CNNs as graphs and provides interpretability from global and local perspectives. The runtime parameters of CNNs and feature maps of each image sample are applied to construct attribution graphs (At-GCs), where the convolutional kernels are represented as nodes and the SHAP values between kernel outputs are assigned as edges. These At-GCs are then employed to pretrain a newly designed heterogeneous graph encoder based on Deep Graph Infomax (DGI). To comprehensively delve into the overall structure of CNNs, the pretrained encoder is used for two types of interpretable tasks: (1) a classifier is attached to the pretrained encoder for the classification of At-GCs, revealing the dependency of At-GC’s topological characteristics on the image sample categories, and (2) a scoring aggregation (SA) network is constructed to assess the importance of each node in At-GCs, thus reflecting the relative importance of kernels in CNNs. The experimental results indicate that the topological characteristics of At-GC exhibit a dependency on the sample category used in its construction, which reveals that kernels in CNNs show distinct combined activation patterns for processing different image categories, meanwhile, the kernels that receive high scores from SA network are crucial for feature extraction, whereas low-scoring kernels can be pruned without affecting model performance, thereby enhancing the interpretability of CNNs.

Self-adaptive selection graph pooling based fault diagnosis method under few samples and noisy environment

Neural network (NN)-based methods are extensively used for intelligent fault diagnosis in industrial systems. Nevertheless, due to the limited availability of faulty samples and the presence of noise interference, most existing NN-based methods perform limited diagnosis performance. In response to these challenges, a self-adaptive selection graph pooling method is proposed. Firstly, graph encoders with sharing parameters are designed to extract local structure-feature information (SFI) of multiple sensor-wise sub-graphs. Then, the temporal continuity of the SFI is maintained through time-by-time concatenation, resulting in a global sensor graph and reducing the dependency on data volume from the perspective of adding prior knowledge. Subsequently, leveraging a self-adaptive node selection mechanism, the noise interference of redundant and noisy sensor-wise nodes in the graph is alleviated, allowing the networks to concentrate on the fault-attention nodes. Finally, the local max pooling and global mean pooling of the node-selection graph are incorporated in the readout module to get the multi-scale graph features, which serve as input to a multi-layer perceptron for fault diagnosis. Two experimental studies involving different mechanical and electrical systems demonstrate that the proposed method not only achieves superior diagnosis performance with limited data, but also maintains strong anti-interference ability in noisy environments. Additionally, it exhibits good interpretability through the proposed self-adaptive node selection mechanism and visualization methods.

Synthetic lethal connectivity and graph transformer improve synthetic lethality prediction.

Synthetic lethality (SL) has shown great promise for the discovery of novel targets in cancer. CRISPR double-knockout (CDKO) technologies can only screen several hundred genes and their combinations, but not genome-wide. Therefore, good SL prediction models are highly needed for genes and gene pairs selection in CDKO experiments. However, lack of scalable SL properties prevents generalizability of SL interactions to out-of-sample data, thereby hindering modeling efforts. In this paper, we recognize that SL connectivity is a scalable and generalizable SL property. We develop a novel two-step multilayer encoder for individual sample-specific SL prediction model (MLEC-iSL), which predicts SL connectivity first and SL interactions subsequently. MLEC-iSL has three encoders, namely, gene, graph, and transformer encoders. MLEC-iSL achieves high SL prediction performance in K562 (AUPR, 0.73; AUC, 0.72) and Jurkat (AUPR, 0.73; AUC, 0.71) cells, while no existing methods exceed 0.62 AUPR and AUC. The prediction performance of MLEC-iSL is validated in a CDKO experiment in 22Rv1 cells, yielding a 46.8% SL rate among 987 selected gene pairs. The screen also reveals SL dependency between apoptosis and mitosis cell death pathways.

Learning from Feature and Global Topologies: Adaptive Multi-View Parallel Graph Contrastive Learning

To address the limitations of existing graph contrastive learning methods, which fail to adaptively integrate feature and topological information and struggle to efficiently capture multi-hop information, we propose an adaptive multi-view parallel graph contrastive learning framework (AMPGCL). It is an unsupervised graph representation learning method designed to generate task-agnostic node embeddings. AMPGCL constructs and encodes feature and topological views to mine feature and global topological information. To encode global topological information, we introduce an H-Transformer to decouple multi-hop neighbor aggregations, capturing global topology from node subgraphs. AMPGCL learns embedding consistency among feature, topology, and original graph encodings through a multi-view contrastive loss, generating semantically rich embeddings while avoiding information redundancy. Experiments on nine real datasets demonstrate that AMPGCL consistently outperforms thirteen state-of-the-art graph representation learning models in classification accuracy, whether in homophilous or non-homophilous graphs.

Contrastive multi-interest graph attention network for knowledge-aware recommendation

Acquiring high-quality representations for both users and items is essential, facilitating a wide range of recommendation scenarios. Utilizing graph neural networks for knowledge-aware recommendation is a recent trend. However, there are two deficiencies in existing GNN-based knowledge-aware models: (1) They are coarse-grained in user representation, failing to capture the multi-interest distribution of users. (2) The supervised signals come only from historical interactions, failing to provide high-quality representations due to sparse data. In this paper, we propose a novel model, CMGAN with Contrastive Multi-interest Graph Attention Network, tailored for personalized knowledge-aware recommendations. Specifically, CMGAN employs a collaborative knowledge graph encoder, enhancing node representations through relational-aware embedding aggregation. Then a dynamic multi-interest generator crafts fine-grained multi-interest representations, adeptly extracting varied interests for each user based on their historical interactions. Furthermore, CMGAN innovates by integrating multi-level contrastive learning to refine representations at both node and multi-interest granularity. It consists of collaborative knowledge graph contrastive learning and multi-interest contrastive learning. The former pursues the acquisition of node representations that are more uniformly distributed, while the latter aims to obtain interest representations that are more distinct. A series of experiments on three benchmark datasets indicate that our model surpasses current state-of-the-art models. The code can be obtainable at https://github.com/liujianfang2021/CMGAN.

Efficient Workflow Scheduling in Edge Cloud-Enabled Space-Air-Ground-Integrated Information Systems

To address the challenges posed by the dynamism, high latency, and resource scarcity in integrated air-space-ground hybrid edge cloud environments on task completion times and node load, we designed a task scheduling system for scenarios involving the transmission and processing of interdependent tasks. This system integrates a graph neural network with attention mechanism and deep reinforcement learning. Specifically, we employ a graph encoder to extract features from DAG tasks and resources. Task scheduling solutions for dynamic environments are then generated using attention mechanism-equipped graph decoder, which are subsequently optimized based on performance metrics through the use of an Advantage Actor-Critic algorithm. Experimental results indicate that this algorithm performs well in terms of completion time and node load balance across tasks with different workflow structures, demonstrating its adaptability to highly dynamic edge cloud environments.

Privacy Preserving Human Mobility Generation Using Grid-Based Data and Graph Autoencoders

This paper proposes a one-to-one trajectory synthetization method with stable long-term individual mobility behavior based on a generalizable area embedding. Previous methods concentrate on producing highly detailed data on short-term and restricted areas for, e.g., autonomous driving scenarios. Another possibility consists of city-wide and beyond scales that can be used to predict general traffic flows. The now-presented approach takes the tracked mobility behavior of individuals and creates coherent synthetic mobility data. These generated data reflect the person’s long-term mobility behavior, guaranteeing location persistency and sound embedding within the point-of-interest structure of the observed area. After an analysis and clustering step of the original data, the area is distributed into a geospatial grid structure (H3 is used here). The neighborhood relationships between the grids are interpreted as a graph. A feed-forward autoencoder and a graph encoding–decoding network generate a latent space representation of the area. The original clustered data are associated with their respective H3 grids. With a greedy algorithm approach and concerning privacy strategies, new combinations of grids are generated as top-level patterns for individual mobility behavior. Based on the original data, concrete locations within the new grids are found and connected to ways. The goal is to generate a dataset that shows equivalence in aggregated characteristics and distances in comparison with the original data. The described method is applied to a sample of 120 from a study with 1000 participants whose mobility data were generated in the city of Munich in Germany. The results show the applicability of the approach in generating synthetic data, enabling further research on individual mobility behavior and patterns. The result comprises a sharable dataset on the same abstraction level as the input data, which can be beneficial for different applications, particularly for machine learning.

Molecular set representation learning

Computational representation of molecules can take many forms, including graphs, string encodings of graphs, binary vectors or learned embeddings in the form of real-valued vectors. These representations are then used in downstream classification and regression tasks using a wide range of machine learning models. However, existing models come with limitations, such as the requirement for clearly defined chemical bonds, which often do not represent the true underlying nature of a molecule. Here we propose a framework for molecular machine learning tasks based on set representation learning. We show that learning on sets of atom invariants alone reaches the performance of state-of-the-art graph-based models on the most-used chemical benchmark datasets and that introducing a set representation layer into graph neural networks can surpass the performance of established methods in the domains of chemistry, biology and material science. We introduce specialized set representation-based neural network architectures for reaction-yield and protein–ligand binding-affinity prediction. Overall, we show that the technique we denote molecular set representation learning is both an alternative and an extension to graph neural network architectures for machine learning tasks on molecules, molecule complexes and chemical reactions.

Contrastive cross-domain sequential recommendation via emphasized intention features

The objective of cross-domain sequential recommendation is to forecast upcoming interactions by leveraging past interactions across diverse domains. Most methods aim to utilize single-domain and cross-domain information as much as possible for personalized preference extraction and effective integration. However, on one hand, most models ignore that cross-domain information is composed of multiple single-domains when generating representations. They still treat cross-domain information the same way as single-domain information, resulting in noisy representation generation. Only by imposing certain constraints on cross-domain information during representation generation can subsequent models minimize interference when considering user preferences. On the other hand, some methods neglect the joint consideration of users’ long-term and short-term preferences and reduce the weight of cross-domain user preferences to minimize noise interference. To better consider the mutual promotion of cross-domain and single-domains factors, we propose a novel model (C2DREIF) that utilizes Gaussian graph encoders to handle information, effectively constraining the correlation of information and capturing useful contextual information more accurately. It also employs a Top-down transformer to accurately extract user intents within each domain, taking into account the user’s long-term and short-term preferences. Additionally, entropy regularized is applied to enhance contrastive learning and mitigate the impact of randomness caused by negative sample composition.

BP-MoE: Behavior Pattern-aware Mixture-of-Experts for Temporal Graph Representation Learning

Temporal graph representation learning aims to develop low-dimensional embeddings for nodes in a graph that can effectively capture their structural and temporal properties. Prior approaches primarily focus on devising sophisticated neural architectures to encode the historical interactions of nodes, enabling capturing the specific behavior patterns. However, such methods often overlook the fact that different nodes in a graph may exhibit distinct evolutionary preferences, resulting in a failure to adequately model the variations in the node evolution behavior. To address these issues, we propose a novel temporal graph learning framework called Behavior Pattern-aware Mixture-of-Experts (BP-MoE), which leverages multiple expert encoder networks with diverse architectures to comprehensively capture the behavior patterns for nodes. In detail, we first design a multi-perspective graph encoder that employs experts based on the gated recurrent units, graph attention units, and multilayer perceptron units to encode a node’s long-term memory, high-order neighborhood features, and short-term behavior preference, respectively. Subsequently, to achieve the personalized evolution preference learning for each individual node, we incorporate a behavior pattern-aware MoE mechanism to activate the pattern-specific experts for each node via a gating network. Finally, to optimize the training process of expert networks, we introduce two auxiliary losses that prevent the imbalanced training across experts. Extensive experiments conducted on multiple benchmark datasets verify the superiority of our approach over the state-of-the-art baselines. Specifically, BP-MoE consistently outperforms the baseline models for the link prediction and node classification tasks under various experimental settings.

GTFN

With the development of smart education, gaining insights into students' understanding during the learning process is crucial in teaching. However, traditional knowledge tracking methods face challenges in capturing the intricate relationships between problems and knowledge points, as well as students' temporal learning changes. Therefore, we design a knowledge tracking model based on a graph temporal fusion network. Firstly, we construct the structure of the question and knowledge skill graph. Then, we design a knowledge graph encoder layer to capture the complex relationships between questions and knowledge skills. Next, we apply a sequential information extraction layer to dynamically model the outputs of each layer in the upper network over time, capturing students' knowledge changes at different time steps. Finally, we use a dynamic attention aggregation network to learn node information at different levels and time sequences. Experimental results on three datasets demonstrate the effectiveness of our method.

Dismantling complex networks with graph contrastive learning and multi-hop aggregation

Network dismantling is a process of identifying influential nodes that can decompose a network into disconnected sub-networks. This provides a novel approach to understanding and analyzing complex networks abstracted from the real world. State-of-the-art solutions for this task exploit graph encoders to capture the structural features of the network, which are then sent to the multi-layer perceptron for predicting the node importance. This process, however, fails to exploit the interactions among the graph representations learned from different views and neglects the neighboring information when evaluating node importance. In this work, we address these issues with a graph contrastive learning framework with multi-hop aggregation, resulting in the identification of influential nodes. Firstly, we construct role graphs to provide a holistic view of the original graphs. Secondly, graph representations are obtained in the individual views, and enhanced expressiveness is achieved through contrastive learning. Finally, based on the representations, the multi-hop neighbor information of the nodes is aggregated to rank the node importance, and thus aid in the identification of important nodes. We evaluate our proposal on real and synthetic networks, and the results show that our method outperforms the baseline with fewer nodes required to disassemble a network.

AIGO-DTI: Predicting Drug-Target Interactions Based on Improved Drug Properties Combined with Adaptive Iterative Algorithms.

Artificial intelligence-based methods for predicting drug-target interactions (DTIs) aim to explore reliable drug candidate targets rapidly and cost-effectively to accelerate the drug development process. However, current methods are often limited by the topological regularities of drug molecules, making them difficult to generalize to a broader chemical space. Additionally, the use of similarity to measure DTI network links often introduces noise, leading to false DTI relationships and affecting the prediction accuracy. To address these issues, this study proposes an Adaptive Iterative Graph Optimization (AIGO)-DTI prediction framework. This framework integrates atomic cluster information and enhances molecular features through the design of functional group prompts and graph encoders, optimizing the construction of DTI association networks. Furthermore, the optimization of graph structure is transformed into a node similarity learning problem, utilizing multihead similarity metric functions to iteratively update the network structure to improve the quality of DTI information. Experimental results demonstrate the outstanding performance of AIGO-DTI on multiple public data sets and label reversal data sets. Case studies, molecular docking, and existing research validate its effectiveness and reliability. Overall, the method proposed in this study can construct comprehensive and reliable DTI association network information, providing new graphing and optimization strategies for DTI prediction, which contribute to efficient drug development and reduce target discovery costs.