Reference library
Intrinsic adaptation in autonomous recurrent neural networks
Paper summary
A discrete-time logistic rate-neuron model adapts intrinsic gain and bias to reduce output-distribution KL divergence from a bounded exponential target with specified mean. The derivation differentiates the transfer function for an input distribution, and uses a small fixed learning rate to track changing statistics rather than ensure exact convergence. A self-excitatory single-neuron example moves into a limiting cycle. Random recurrent networks with fixed signed synapses exhibit synchronized oscillatory, chaotic or intermittent bursting regimes depending on excitation fraction, connectivity and target mean. The authors acknowledge the two-parameter transfer family cannot generally fit the target distribution exactly. Their discussion proposes an interplay with synaptic plasticity; that combined system and an embodied environment are not demonstrated in the reviewed model.
Relevance to the project
Closest reviewed comparison for locally regulating gain/bias to sustain autonomous activity. It supports measuring fixed points, repetition, distribution mismatch and feedback correlations separately, rather than treating ongoing motion or high marginal entropy as sufficient for varied useful dynamics. Its logistic, fixed-synapse network differs from adaptive F/H and the larger changing graph; no gain, learning rate or design choice is adopted.
Computational evidence under specific model assumptions, not biological verification or a guarantee for every seed, every target, tanh, online changing weights, or the project graph. A limit cycle is a counterexample to equating intrinsic adaptation with irregular activity. The claimed general synaptic freezing trend is a cited background claim, not an independent result established by this read.
Source record
- Reference ID
- intrinsic-adaptation-2012
- Archived version
- arXiv:1110.3161v1, 14 October 2011; preprint of the 2012 journal work, not the published journal PDF
- Summary provenance
- reference-library/papers/intrinsic-adaptation-2012/metadata.json · summary
- Review status
- Targeted full-text methods review; neither exhaustive full-paper reading nor independent reproduction
- Reading scope
- Targeted full-text reading of PDF pages 2–9, 12–13, 15–16 and 20–21, including stochastic adaptation, single-neuron stability, network setup, regime comparisons and Discussion. Front matter checked. Pages 10–11, 14 and 17–19, most figure pixels and bibliography not substantively reviewed; no independent simulation or Lyapunov calculation. Extract may lose equation glyphs or order.