Layer Ensembles (2210.04882v3)
Abstract: Deep Ensembles, as a type of Bayesian Neural Networks, can be used to estimate uncertainty on the prediction of multiple neural networks by collecting votes from each network and computing the difference in those predictions. In this paper, we introduce a method for uncertainty estimation that considers a set of independent categorical distributions for each layer of the network, giving many more possible samples with overlapped layers than in the regular Deep Ensembles. We further introduce an optimized inference procedure that reuses common layer outputs, achieving up to 19x speed up and reducing memory usage quadratically. We also show that the method can be further improved by ranking samples, resulting in models that require less memory and time to run while achieving higher uncertainty quality than Deep Ensembles.
- David J C Mackay, “Probable networks and plausible predictions — a review of practical bayesian methods for supervised neural networks,” Network, vol. 6, no. 3, pp. 469–505, 1995.
- “Bayesian Deep Learning and a Probabilistic Perspective of Generalization,” in NeurIPS, 2020, vol. 33.
- “Bayesian neural networks,” arXiv:2006.01490, 2020.
- “Bayesian learning for neural networks: an algorithmic survey,” arxiv:2211.11865, 2022.
- “Weight Uncertainty in Neural Networks,” in ICML, 2015, vol. 37, pp. 1613–1622.
- “Hypermodels for exploration,” in ICLR, 2020, vol. 8.
- “Dropout as a bayesian approximation: Representing model uncertainty in deep learning,” in ICML, 2016, vol. 48, pp. 1050–1059.
- “Randomized prior functions for deep reinforcement learning,” in NeurIPS, 2018, vol. 31, pp. 8626–8638.
- “Epistemic Neural Networks,” arXiv:2107.08924, 2021.
- “Deep neural networks as gaussian processes,” in ICLR, 2018.
- S. Kullback and R. A. Leibler, “On Information and Sufficiency,” Ann. Math. Stat., vol. 22, no. 1, pp. 79 – 86, 1951.
- “Dropout: A simple way to prevent neural networks from overfitting,” JMLR, vol. 15, no. 56, pp. 1929–1958, 2014.
- “Auto-encoding variational bayes,” in ICLR, 2014, vol. 2.
- “Variational neural networks,” arxiv:2207.01524, 2022.
- Matias Valdenegro-Toro, “Deep sub-ensembles for fast uncertainty estimation in image classification,” arxiv:1910.08168, 2019.
- “Batchensemble: An alternative approach to efficient ensemble and lifelong learning,” in ICLR, 2020, vol. 8.
- “Layer ensembles: A single-pass uncertainty estimation in deep learning for segmentation,” in MICCAI, 2022, vol. 13438, pp. 514–524.
- “Conditional deep learning for energy-efficient and enhanced pattern recognition,” in DATE, 2016, pp. 475–480.
- “Github code for epistemic neural networks,” https://github.com/deepmind/enn, 2021.
Collections
Sign up for free to add this paper to one or more collections.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.