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