Feature Space Particle Inference for Neural Network Ensembles
This work addresses the problem of improving ensemble diversity and performance for machine learning practitioners, offering an incremental advancement over existing methods.
The paper tackled the challenge of efficiently applying particle-based inference to neural network ensembles by proposing optimization in feature space, which significantly outperformed Deep Ensembles on accuracy, calibration, and robustness across real-world datasets.
Ensembles of deep neural networks demonstrate improved performance over single models. For enhancing the diversity of ensemble members while keeping their performance, particle-based inference methods offer a promising approach from a Bayesian perspective. However, the best way to apply these methods to neural networks is still unclear: seeking samples from the weight-space posterior suffers from inefficiency due to the over-parameterization issues, while seeking samples directly from the function-space posterior often results in serious underfitting. In this study, we propose optimizing particles in the feature space where the activation of a specific intermediate layer lies to address the above-mentioned difficulties. Our method encourages each member to capture distinct features, which is expected to improve ensemble prediction robustness. Extensive evaluation on real-world datasets shows that our model significantly outperforms the gold-standard Deep Ensembles on various metrics, including accuracy, calibration, and robustness. Code is available at https://github.com/DensoITLab/featurePI .