Stochasticity from function - Why the Bayesian brain may need no noise
| Autor: | Walter Senn, Johannes Schemmel, Mihai A. Petrovici, Karlheinz Meier, Ilja Bytschok, Dominik Dold, Oliver Breitwieser, Akos F. Kungl, Andreas Baumbach |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
FOS: Computer and information sciences
0209 industrial biotechnology Cognitive Neuroscience Computation Bayesian probability Models Neurological FOS: Physical sciences 610 Medicine & health Machine Learning (stat.ML) 02 engineering and technology Bayesian inference Rendering (computer graphics) Membrane Potentials 020901 industrial engineering & automation Artificial Intelligence Statistics - Machine Learning 0202 electrical engineering electronic engineering information engineering Computer Simulation Neural and Evolutionary Computing (cs.NE) Physics - Biological Physics Spiking neural network Neurons Emulation Neuronal Plasticity Quantitative Biology::Neurons and Cognition Computer Science - Neural and Evolutionary Computing Brain Bayes Theorem Disordered Systems and Neural Networks (cond-mat.dis-nn) Condensed Matter - Disordered Systems and Neural Networks Neuromorphic engineering Receptive field Biological Physics (physics.bio-ph) FOS: Biological sciences Quantitative Biology - Neurons and Cognition 020201 artificial intelligence & image processing Neurons and Cognition (q-bio.NC) Neural Networks Computer Algorithm |
| Zdroj: | Dold, Dominik; Bytschok, Ilja; Kungl, Akos F.; Baumbach, Andreas; Breitwieser, Oliver; Senn, Walter; Schemmel, Johannes; Meier, Karlheinz; Petrovici, Mihai A. (2019). Stochasticity from function — Why the Bayesian brain may need no noise. Neural networks, 119, pp. 200-213. Elsevier 10.1016/j.neunet.2019.08.002 Neural Networks |
| ISSN: | 1879-2782 |
| Popis: | An increasing body of evidence suggests that the trial-to-trial variability of spiking activity in the brain is not mere noise, but rather the reflection of a sampling-based encoding scheme for probabilistic computing. Since the precise statistical properties of neural activity are important in this context, many models assume an ad-hoc source of well-behaved, explicit noise, either on the input or on the output side of single neuron dynamics, most often assuming an independent Poisson process in either case. However, these assumptions are somewhat problematic: neighboring neurons tend to share receptive fields, rendering both their input and their output correlated; at the same time, neurons are known to behave largely deterministically, as a function of their membrane potential and conductance. We suggest that spiking neural networks may, in fact, have no need for noise to perform sampling-based Bayesian inference. We study analytically the effect of auto- and cross-correlations in functionally Bayesian spiking networks and demonstrate how their effect translates to synaptic interaction strengths, rendering them controllable through synaptic plasticity. This allows even small ensembles of interconnected deterministic spiking networks to simultaneously and co-dependently shape their output activity through learning, enabling them to perform complex Bayesian computation without any need for noise, which we demonstrate in silico, both in classical simulation and in neuromorphic emulation. These results close a gap between the abstract models and the biology of functionally Bayesian spiking networks, effectively reducing the architectural constraints imposed on physical neural substrates required to perform probabilistic computing, be they biological or artificial. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect