synaptic plasticity

the mechanism by which connection strengths between neurons change. three irreducible types:

type valence function timescale cyber analog
Hebbian learning +1 (excitatory) strengthen co-active connections ms-min $\Delta\pi$ reward for correlated focus
anti-Hebbian learning -1 (inhibitory) weaken co-active connections, decorrelate ms-min market inhibition via inversely coupled bonding surface
homeostatic learning 0 (modulatory) scale all weights to maintain target activity hours-days focus conservation ($\sum \pi_i = 1$), forgetting

these three types are irreducible — remove any one and the system degenerates:

  • without Hebbian: no pattern detection, no structure formation
  • without anti-Hebbian: no decorrelation, representations collapse to redundancy
  • without homeostatic: runaway excitation or complete silencing, no stable operating regime

the triad maps precisely onto the two three paradox: the synapse is binary (connection exists or not), but the learning signal is ternary (strengthen, weaken, regulate). binary topology, ternary economics.

see learning, collective learning, brain

Local Graph