Semantic-Aware Gaussian Process Calibration with Structured Layerwise Kernels for Deep Neural Networks
This addresses the issue of unreliable confidence estimates in deep neural networks for applications requiring robust predictive reliability, representing an incremental improvement over existing GP calibration methods.
The paper tackled the problem of calibrating confidence in neural network classifiers by proposing a Semantic-Aware Layer-wise Gaussian Process (SAL-GP) framework that mirrors the layered architecture of deep neural networks, resulting in enhanced interpretability and principled uncertainty quantification.
Calibrating the confidence of neural network classifiers is essential for quantifying the reliability of their predictions during inference. However, conventional Gaussian Process (GP) calibration methods often fail to capture the internal hierarchical structure of deep neural networks, limiting both interpretability and effectiveness for assessing predictive reliability. We propose a Semantic-Aware Layer-wise Gaussian Process (SAL-GP) framework that mirrors the layered architecture of the target neural network. Instead of applying a single global GP correction, SAL-GP employs a multi-layer GP model, where each layer's feature representation is mapped to a local calibration correction. These layerwise GPs are coupled through a structured multi-layer kernel, enabling joint marginalization across all layers. This design allows SAL-GP to capture both local semantic dependencies and global calibration coherence, while consistently propagating predictive uncertainty through the network. The resulting framework enhances interpretability aligned with the network architecture and enables principled evaluation of confidence consistency and uncertainty quantification in deep models.