A living fungal network as a reservoir computer. NODES are fungal hubs; growing HYPHAE
wire them together; electrical SPIKES propagate along the web with conduction delays. Co-active nodes
strengthen and grow new hyphae (Hebbian, use-it-or-lose-it), so the living topology adapts. A trained
linear read-out taps the reservoir state to learn a simple input→output mapping.
TASK · XOR (2-in → 1-out)stylized bio-computing · DishBrain / Adamatzky spirit — not a real culture
fungal node (idle)
node firing (spike)
hypha (thickness = strength)
spike travelling
INPUT node
READ-OUT node
Living fungal reservoir · grow & learn
⊞ XOR
& AND
⌛ MEMORY
0.55
0.30
0.60
3×
The web self-organises: firing nodes grow hyphae toward recently-active
neighbours; unused hyphae decay and get pruned. Train read-out fits a linear tap over the reservoir
to the current task; Stimulate injects the next test pattern. Drag = pan · scroll = zoom.
HUD · reservoir telemetry
Nodes—
Hyphae (live)—
Spikes travelling—
Firing now—
Total conductance—
Read-out accuracynot trained
Last test—
Reservoir: living hyphae web Read-out: trained linear tap