2000 character limit reached
Uncertainty in latent representations of variational autoencoders optimized for visual tasks (2404.15390v2)
Published 23 Apr 2024 in cs.LG and cs.AI
Abstract: Deep Generative Models (DGMs) can learn flexible latent variable representations of images while avoiding intractable computations, common in Bayesian inference. However, investigating the properties of inference in Variational Autoencoders (VAEs), a major class of DGMs, reveals severe problems in their uncertainty representations. Here we draw inspiration from classical computer vision to introduce an inductive bias into the VAE by incorporating a global explaining-away latent variable, which remedies defective inference in VAEs. Unlike standard VAEs, the Explaing-Away VAE (EA-VAE) provides uncertainty estimates that align with normative requirements across a wide spectrum of perceptual tasks, including image corruption, interpolation, and out-of-distribution detection. We find that restored inference capabilities are delivered by developing a motif in the inference network (the encoder) which is widespread in biological neural networks: divisive normalization. Our results establish EA-VAEs as reliable tools to perform inference under deep generative models with appropriate estimates of uncertainty.
- Catalyzing next-generation artificial intelligence through neuroai. Nature communications, 14(1):1597, 2023.
- Performance-optimized hierarchical models predict neural responses in higher visual cortex. Proceedings of the national academy of sciences, 111(23):8619–8624, 2014.
- A task-optimized neural network replicates human auditory behavior, predicts brain responses, and reveals a cortical processing hierarchy. Neuron, 98(3):630–644, 2018.
- A deep learning framework for neuroscience. Nature Neuroscience, 22(11):1761–1770, Nov 2019.
- Simulating a primary visual cortex at the front of cnns improves robustness to image perturbations. In H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. Lin, editors, Advances in Neural Information Processing Systems, volume 33, pages 13073–13087. Curran Associates, Inc., 2020.
- Natural and artificial intelligence: A brief introduction to the interplay between ai and neuroscience research. Neural Networks, 144:603–613, 2021.
- A diverse range of factors affect the nature of neural representations underlying short-term memory. Nature neuroscience, 22(2):275–283, 2019.
- Cortical-like dynamics in recurrent circuits optimized for sampling-based probabilistic inference. Nature neuroscience, 23(9):1138–1149, 2020.
- Daniel L. K. Yamins and James J. DiCarlo. Using goal-driven deep learning models to understand sensory cortex. Nature Neuroscience, 19(3):356–365, Mar 2016.
- Emergence of simple-cell receptive field properties by learning a sparse code for natural images. Nature, 381(6583):607–609, Jun 1996.
- David JC MacKay. Information theory, inference and learning algorithms. Cambridge university press, 2003.
- The bayesian brain: the role of uncertainty in neural coding and computation. Trends in Neurosciences, 27(12):712–719, 2004.
- David J. C. MacKay. Information Theory, Inference, and Learning Algorithms. Copyright Cambridge University Press, 2003.
- Edwin T Jaynes. Probability theory: The logic of science. Cambridge university press, 2003.
- Deep learning. nature, 521(7553):436–444, 2015.
- Deep convolutional models improve predictions of macaque v1 responses to natural images. PLoS computational biology, 15(4):e1006897, 2019.
- On calibration of modern neural networks. In International conference on machine learning, pages 1321–1330. PMLR, 2017.
- Dropout as a bayesian approximation: Representing model uncertainty in deep learning. In international conference on machine learning, pages 1050–1059. PMLR, 2016.
- Auto-encoding variational bayes, 2022.
- Sparse-coding variational auto-encoders. bioRxiv, 2018.
- Top-down inference in an early visual cortex inspired hierarchical variational autoencoder, 2022.
- Spontaneous cortical activity reveals hallmarks of an optimal internal model of the environment. Science, 331(6013):83–87, 2011.
- Scale mixtures of gaussians and the statistics of natural images. In S. Solla, T. Leen, and K. Müller, editors, Advances in Neural Information Processing Systems, volume 12. MIT Press, 1999.
- Neuronal variability reflects probabilistic inference tuned to natural image statistics. Nature Communications, 12(1):3635, Jun 2021.
- Y. LECUN. The mnist database of handwritten digits. http://yann.lecun.com/exdb/mnist/, 1998.
- Chestx-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases. In CVPR, pages 3462–3471, 2017.
- Spatial and temporal jitter distort estimated functional properties of visual sensory neurons. J Comput Neurosci, 27(3):309–319, April 2009.
- M Carandini and DJ Heeger. Normalization as a canonical neural computation. Nature Reviews Neuroscience, 13(1):51, 2012.
- Towards unraveling calibration biases in medical image analysis. In Workshop on Clinical Image-Based Procedures, pages 132–141. Springer, 2023.
- C.M. Bishop. Pattern Recognition and Machine Learning. Information Science and Statistics. Springer, 2006.
- Neural variability and sampling-based probabilistic representations in the visual cortex. Neuron, 92(2):530–543, 2016.
- Ernst Heinrich Weber. De Pulsu, resorptione, auditu et tactu: Annotationes anatomicae et physiologicae… CF Koehler, 1834.
- J H van Hateren and A van der Schaaf. Independent component filters of natural images compared with simple cells in primary visual cortex. Proc Biol Sci, 265(1394):359–366, March 1998.