Layer-to-Layer Knowledge Mixing in Graph Neural Network for Chemical
Property Prediction
Daokun Zhang, David K. Chalmers et al.
Paper
Details
#GraphNeuralNetworks #ChemicalPropertyPrediction #LayerWiseKnowledgeMixing
Property Prediction
Daokun Zhang, David K. Chalmers et al.
Paper
Details
#GraphNeuralNetworks #ChemicalPropertyPrediction #LayerWiseKnowledgeMixing
November 2, 2025 at 5:01 PM
Layer-to-Layer Knowledge Mixing in Graph Neural Network for Chemical
Property Prediction
Daokun Zhang, David K. Chalmers et al.
Paper
Details
#GraphNeuralNetworks #ChemicalPropertyPrediction #LayerWiseKnowledgeMixing
Property Prediction
Daokun Zhang, David K. Chalmers et al.
Paper
Details
#GraphNeuralNetworks #ChemicalPropertyPrediction #LayerWiseKnowledgeMixing
A new taxonomy for higher-order graph neural networks has been released, clarifying their classification. Read more: https://getnews.me/new-taxonomy-clarifies-higher-order-graph-neural-networks/ #graphneuralnetworks #taxonomy #machinelearning
October 8, 2025 at 12:48 PM
A new taxonomy for higher-order graph neural networks has been released, clarifying their classification. Read more: https://getnews.me/new-taxonomy-clarifies-higher-order-graph-neural-networks/ #graphneuralnetworks #taxonomy #machinelearning
NatGVD achieves up to 53.04% natural adversarial evasion against GNN‑based and transformer vulnerability detectors, per a study submitted on 6 Oct 2025. https://getnews.me/natgvd-natural-adversarial-attack-on-graph-based-vulnerability-detection/ #graphneuralnetworks #adversarialattack
October 8, 2025 at 6:21 AM
NatGVD achieves up to 53.04% natural adversarial evasion against GNN‑based and transformer vulnerability detectors, per a study submitted on 6 Oct 2025. https://getnews.me/natgvd-natural-adversarial-attack-on-graph-based-vulnerability-detection/ #graphneuralnetworks #adversarialattack
Position paper (6 Oct 2025) shows graph‑based tabular deep learning often misses true feature‑interaction graphs; enforcing the correct graph boosts performance. Read more: https://getnews.me/graph-based-tabular-deep-learning-needs-to-model-feature-interactions/ #graphneuralnetworks #neurips
October 8, 2025 at 2:31 AM
Position paper (6 Oct 2025) shows graph‑based tabular deep learning often misses true feature‑interaction graphs; enforcing the correct graph boosts performance. Read more: https://getnews.me/graph-based-tabular-deep-learning-needs-to-model-feature-interactions/ #graphneuralnetworks #neurips
The authors have formally withdrawn their study on the DiMuST model for point‑of‑interest recommendation, citing a need for major manuscript restructuring. Read more: https://getnews.me/disentangled-multiplex-graph-paper-for-poi-recommendation-withdrawn/ #pointofinterest #graphneuralnetworks
October 6, 2025 at 3:15 PM
The authors have formally withdrawn their study on the DiMuST model for point‑of‑interest recommendation, citing a need for major manuscript restructuring. Read more: https://getnews.me/disentangled-multiplex-graph-paper-for-poi-recommendation-withdrawn/ #pointofinterest #graphneuralnetworks
A study finds that a heterogeneous graph neural network outperforms homogeneous GNNs and a fully‑connected NN in predicting optimal topology‑control actions on unseen grid configurations. https://getnews.me/graph-neural-networks-enhance-power-grid-topology-control/ #graphneuralnetworks #powergrid
October 6, 2025 at 12:53 PM
A study finds that a heterogeneous graph neural network outperforms homogeneous GNNs and a fully‑connected NN in predicting optimal topology‑control actions on unseen grid configurations. https://getnews.me/graph-neural-networks-enhance-power-grid-topology-control/ #graphneuralnetworks #powergrid
Bootstrapped graph neural networks refine node‑pair similarity, achieving three times higher alignment accuracy and solving cases where prior methods failed. Read more: https://getnews.me/sequential-graph-neural-networks-boost-combinatorial-graph-alignment/ #graphneuralnetworks #graphalignment
October 6, 2025 at 10:04 AM
Bootstrapped graph neural networks refine node‑pair similarity, achieving three times higher alignment accuracy and solving cases where prior methods failed. Read more: https://getnews.me/sequential-graph-neural-networks-boost-combinatorial-graph-alignment/ #graphneuralnetworks #graphalignment
VisitHGNN, a heterogeneous graph neural network, predicts POI visit probabilities with a Top‑1 accuracy of 0.853 and R‑square of 0.892 in Fulton County, GA. Read more: https://getnews.me/visithgnn-graph-neural-network-predicts-urban-poi-visit-patterns/ #visithgnn #graphneuralnetworks #urbanmobility
October 6, 2025 at 7:11 AM
VisitHGNN, a heterogeneous graph neural network, predicts POI visit probabilities with a Top‑1 accuracy of 0.853 and R‑square of 0.892 in Fulton County, GA. Read more: https://getnews.me/visithgnn-graph-neural-network-predicts-urban-poi-visit-patterns/ #visithgnn #graphneuralnetworks #urbanmobility
A graph neural network detects asymptomatic carriers in epidemic models, achieving precision and recall. Tested on Erdős‑Rényi, scale‑free and small‑world networks. Read more: https://getnews.me/graph-neural-networks-identify-asymptomatic-spreaders-in-epidemics/ #graphneuralnetworks #epidemics
October 6, 2025 at 6:05 AM
A graph neural network detects asymptomatic carriers in epidemic models, achieving precision and recall. Tested on Erdős‑Rényi, scale‑free and small‑world networks. Read more: https://getnews.me/graph-neural-networks-identify-asymptomatic-spreaders-in-epidemics/ #graphneuralnetworks #epidemics
NC‑Iso improves subgraph matching accuracy while keeping fast inference, outperforming prior GNN models on nine benchmark datasets. Code released on GitHub. Read more: https://getnews.me/hierarchy-aware-neural-subgraph-matching-boosts-accuracy-and-speed/ #nciso #graphneuralnetworks
October 2, 2025 at 9:53 PM
NC‑Iso improves subgraph matching accuracy while keeping fast inference, outperforming prior GNN models on nine benchmark datasets. Code released on GitHub. Read more: https://getnews.me/hierarchy-aware-neural-subgraph-matching-boosts-accuracy-and-speed/ #nciso #graphneuralnetworks
Researchers propose a proactive O‑RAN handover framework using graph neural network link prediction, outperforming statistical methods. The GNNs were tested on data. https://getnews.me/graph-neural-networks-enable-proactive-mobility-management-in-o-ran/ #orannetwork #graphneuralnetworks
October 1, 2025 at 10:42 AM
Researchers propose a proactive O‑RAN handover framework using graph neural network link prediction, outperforming statistical methods. The GNNs were tested on data. https://getnews.me/graph-neural-networks-enable-proactive-mobility-management-in-o-ran/ #orannetwork #graphneuralnetworks
A differential encoding improves GNN embeddings by contrasting node and neighbor features, boosting performance on 7 datasets. In IEEE Transactions on Big Data (Sept 2025). Read more: https://getnews.me/differential-encoding-boosts-graph-neural-network-representations/ #graphneuralnetworks #bigdata
October 1, 2025 at 8:10 AM
A differential encoding improves GNN embeddings by contrasting node and neighbor features, boosting performance on 7 datasets. In IEEE Transactions on Big Data (Sept 2025). Read more: https://getnews.me/differential-encoding-boosts-graph-neural-network-representations/ #graphneuralnetworks #bigdata
Researchers unveiled graph networks that enforce IRC safety and E(2)/O(2) equivariance, matching quark‑gluon classification accuracy; the study was submitted on 26 September 2025. https://getnews.me/interpretable-jet-physics-with-irc-safe-equivariant-neural-networks/ #ircsafety #graphneuralnetworks
October 1, 2025 at 4:33 AM
Researchers unveiled graph networks that enforce IRC safety and E(2)/O(2) equivariance, matching quark‑gluon classification accuracy; the study was submitted on 26 September 2025. https://getnews.me/interpretable-jet-physics-with-irc-safe-equivariant-neural-networks/ #ircsafety #graphneuralnetworks
New AI model ELPG‑DTFS reaches 97.63% accuracy and 97.33% F1 on a 53‑person, 128‑channel EEG MODMA dataset, improving depression screening with adaptive graph learning. Read more: https://getnews.me/new-graph-neural-network-boosts-eeg-based-depression-diagnosis/ #depression #eeg #graphneuralnetworks
October 1, 2025 at 12:36 AM
New AI model ELPG‑DTFS reaches 97.63% accuracy and 97.33% F1 on a 53‑person, 128‑channel EEG MODMA dataset, improving depression screening with adaptive graph learning. Read more: https://getnews.me/new-graph-neural-network-boosts-eeg-based-depression-diagnosis/ #depression #eeg #graphneuralnetworks
DIMIGNN, a new graph neural network, picks diverse neighbors and fuses multiple temporal scales, delivering lower forecast errors on electricity demand and traffic flow data. https://getnews.me/new-graph-neural-network-improves-multivariate-time-series-forecasting/ #graphneuralnetworks #forecasting
September 30, 2025 at 11:39 AM
DIMIGNN, a new graph neural network, picks diverse neighbors and fuses multiple temporal scales, delivering lower forecast errors on electricity demand and traffic flow data. https://getnews.me/new-graph-neural-network-improves-multivariate-time-series-forecasting/ #graphneuralnetworks #forecasting
Pure Node Sampling (PNS) module balances class ratios and topology in GNNs, significantly reducing variance. The arXiv preprint was posted on 28 September 2025. Read more: https://getnews.me/pure-node-sampling-tackles-class-imbalance-in-graph-neural-networks/ #graphneuralnetworks #classimbalance
September 30, 2025 at 11:32 AM
Pure Node Sampling (PNS) module balances class ratios and topology in GNNs, significantly reducing variance. The arXiv preprint was posted on 28 September 2025. Read more: https://getnews.me/pure-node-sampling-tackles-class-imbalance-in-graph-neural-networks/ #graphneuralnetworks #classimbalance
Caterpillar GNN replaces message‑passing with walk‑based aggregation, using far fewer hidden nodes while matching the predictive accuracy of leading MPGNNs on real‑world data. https://getnews.me/caterpillar-gnn-offers-efficient-walk-based-aggregation/ #caterpillargnn #graphneuralnetworks
September 29, 2025 at 10:35 PM
Caterpillar GNN replaces message‑passing with walk‑based aggregation, using far fewer hidden nodes while matching the predictive accuracy of leading MPGNNs on real‑world data. https://getnews.me/caterpillar-gnn-offers-efficient-walk-based-aggregation/ #caterpillargnn #graphneuralnetworks
RsGCN, a graph convolutional network, was trained on TSP instances of up to 100 nodes and, after just three epochs, can generalize to 10 000‑node problems without fine‑tuning. Read more: https://getnews.me/rsgcn-improves-generalization-of-gcn-solvers-for-tsp/ #rsgcn #tsp #graphneuralnetworks
September 29, 2025 at 10:19 PM
RsGCN, a graph convolutional network, was trained on TSP instances of up to 100 nodes and, after just three epochs, can generalize to 10 000‑node problems without fine‑tuning. Read more: https://getnews.me/rsgcn-improves-generalization-of-gcn-solvers-for-tsp/ #rsgcn #tsp #graphneuralnetworks
LGKDE merges graph neural networks with learnable kernel density estimation, beating state‑of‑the‑art baselines on several graph anomaly‑detection benchmarks. https://getnews.me/learnable-kernel-density-estimation-advances-graph-neural-networks/ #graphneuralnetworks #densityestimation
September 29, 2025 at 10:03 PM
LGKDE merges graph neural networks with learnable kernel density estimation, beating state‑of‑the‑art baselines on several graph anomaly‑detection benchmarks. https://getnews.me/learnable-kernel-density-estimation-advances-graph-neural-networks/ #graphneuralnetworks #densityestimation
ATLAS, a spatio‑temporal graph framework for account takeover detection, raised AUC by 6.38% over XGBoost and cut user friction by over 50% in Capital One’s live testing. Read more: https://getnews.me/graph-based-model-boosts-account-takeover-detection-by-6/ #graphneuralnetworks #frauddetection
September 26, 2025 at 9:53 PM
ATLAS, a spatio‑temporal graph framework for account takeover detection, raised AUC by 6.38% over XGBoost and cut user friction by over 50% in Capital One’s live testing. Read more: https://getnews.me/graph-based-model-boosts-account-takeover-detection-by-6/ #graphneuralnetworks #frauddetection
GNNE merges GCN and GAT with entropy to spot vital nodes, outperforming eight topology‑based and four graph‑ML methods on a synthetic Barabási–Albert network and six real‑world datasets. https://getnews.me/graph-neural-network-entropy-model-enhances-vital-node-detection/ #gnne #graphneuralnetworks
September 25, 2025 at 8:20 PM
GNNE merges GCN and GAT with entropy to spot vital nodes, outperforming eight topology‑based and four graph‑ML methods on a synthetic Barabási–Albert network and six real‑world datasets. https://getnews.me/graph-neural-network-entropy-model-enhances-vital-node-detection/ #gnne #graphneuralnetworks
FedIA keeps only the top 5 % of gradient coordinates, raising accuracy. Tested on Twitch gamers and multilingual Wikipedia graphs, it outperformed baselines. Read more: https://getnews.me/fedia-enhances-federated-graph-learning-with-gradient-pruning/ #federatedlearning #graphneuralnetworks
September 25, 2025 at 6:32 PM
FedIA keeps only the top 5 % of gradient coordinates, raising accuracy. Tested on Twitch gamers and multilingual Wikipedia graphs, it outperformed baselines. Read more: https://getnews.me/fedia-enhances-federated-graph-learning-with-gradient-pruning/ #federatedlearning #graphneuralnetworks
Self‑supervised graph‑neural model infers synaptic connections, validated on synthetic ring attractors and mouse head‑direction data. Submitted 21 Sep 2025, NeurIPS 2025. https://getnews.me/graph-neural-networks-power-self-supervised-neural-circuit-discovery/ #graphneuralnetworks #neurips
September 25, 2025 at 9:40 AM
Self‑supervised graph‑neural model infers synaptic connections, validated on synthetic ring attractors and mouse head‑direction data. Submitted 21 Sep 2025, NeurIPS 2025. https://getnews.me/graph-neural-networks-power-self-supervised-neural-circuit-discovery/ #graphneuralnetworks #neurips
TF-DWGNet builds directed weighted graphs per omics layer and merges them via low‑rank tensor fusion, achieving high accuracy on cancer subtype datasets. Read more: https://getnews.me/tf-dwgnet-directed-weighted-graph-ai-boosts-cancer-subtype-classification/ #cancersubtype #graphneuralnetworks
September 25, 2025 at 7:54 AM
TF-DWGNet builds directed weighted graphs per omics layer and merges them via low‑rank tensor fusion, achieving high accuracy on cancer subtype datasets. Read more: https://getnews.me/tf-dwgnet-directed-weighted-graph-ai-boosts-cancer-subtype-classification/ #cancersubtype #graphneuralnetworks
A new method unrolls the dual ascent algorithm into paired graph neural networks, delivering near‑optimal and near‑feasible results; the paper was posted on 21 September 2025. Read more: https://getnews.me/unrolled-graph-neural-networks-for-constrained-optimization/ #graphneuralnetworks #dualascent
September 24, 2025 at 6:20 PM
A new method unrolls the dual ascent algorithm into paired graph neural networks, delivering near‑optimal and near‑feasible results; the paper was posted on 21 September 2025. Read more: https://getnews.me/unrolled-graph-neural-networks-for-constrained-optimization/ #graphneuralnetworks #dualascent