← All models

Many-to-many Potts Network

Explore memory associations between frontal and posterior networks, with equal adaptation times by default.

Ready

Frontal

00.250.50.75101Time (network sweeps)Memory overlapStart a simulation to see the memory overlaps.

Posterior

00.250.50.75101Time (network sweeps)Memory overlapStart a simulation to see the memory overlaps.

256 units and 49 fresh random memories per region. Colours identify memory indices; read the matrix to see their associations.

Model and run details

Adaptation tracks state activity σ. Fast and slow activity-dependent inhibition use T₃A = 10, T₃B = 100000 and γA = 0.5, as in the heterogeneous model.

λ scales intra-area connections by (1 + λ) and incoming inter-area connections by (1 − λ). At λ = 1 the region receives no inter-area input. Each region receives its own transient cue: a selected stored memory or a fresh, unstored random pattern. Cue strength is 5 and decays with a 40-sweep time scale; it is removed after 360 sweeps.

Every Start regenerates both memory sets, unit connection dilution and update tables, then uses a fresh runtime seed. The selected association graph and its scores remain fixed.