<ul><li>Graph Neural Networks (GNNs) face model bias issues, especially in the presence of class imbalance.</li><li>NeuBM (Neutral Bias Mitigation) is introduced to mitigate model bias in GNNs using neutral input calibration.</li><li>NeuBM leverages a dynamically updated neutral graph to estimate and correct biases, improving predictions and reducing bias across different classes.</li><li>Extensive experiments show that NeuBM significantly enhances balanced accuracy and recall of minority classes, particularly effective in scenarios with severe class imbalance and limited labeled data.</li></ul>

NeuBM: Mitigating Model Bias in Graph Neural Networks through Neutral Input Calibration

Discover more