Corpus ID: 27199345

Probabilistic Synapses

  title={Probabilistic Synapses},
  author={L. Aitchison and Alex Pouget and P. Latham},
  • L. Aitchison, Alex Pouget, P. Latham
  • Published 2014
  • Biology
  • Learning, especially rapid learning, is critical for survival. However, learning is hard: a large number of synaptic weights must be set based on noisy, often ambiguous, sensory information. In such a high-noise regime, keeping track of probability distributions over weights — not just point estimates — is the optimal strategy. Here we hypothesize that synapses take that optimal strategy: they do not store just the mean weight; they also store their degree of uncertainty — in essence, they put… CONTINUE READING
    3 Citations

    Figures from this paper.

    Paper Mentions

    Rapid Bayesian learning in the mammalian olfactory system
    Noisy Synaptic Conductance: Bug or a Feature?
    • 1
    • PDF
    Ongoing, rational calibration of reward-driven perceptual biases
    • 18


    Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning
    • 1,693
    Highly Nonrandom Features of Synaptic Connectivity in Local Cortical Circuits
    • 1,247
    • Highly Influential
    • PDF
    Applied Optimal Estimation
    • 5,154
    • Highly Influential
    Adaptive switching circuits
    • 2,780
    • PDF