Excitatory Neurons (80%)
Inhibitory Neurons (20%)
Excitatory Rate
24.5 Hz
Inhibitory Rate
32.1 Hz
Saturation Ratio
0.28
Network State
Stable Dynamic Equilibrium
80%
-2.5x
30 Hz
2 ms
Biological Gain Control vs Artificial Activations
In standard artificial neural networks (ANNs), non-linearities like ReLU propagate continuous positive outputs without self-regulating feedback, risking vanishing/exploding gradients or saturation.
Biological E/I Balance (Dale's Law)
Inhibitory interneurons exert rapid negative feedback to hold the population firing rate in a high-sensitivity regime without hitting extreme ceiling saturation.
Artificial Activations (ReLU / Sigmoid)
Lacks dedicated negative population feedback. Overactive feedforward drives lead to unbounded linear amplification (ReLU) or zero-gradient flatlines (Sigmoid).
Canonical Telemetry Proof Mirror
Network State: Stable Dynamic Equilibrium
Excitatory Rate: 24.5 Hz
Inhibitory Rate: 32.1 Hz
Saturation Ratio: 0.28