CONTINUOUS MEDIUM · DISTRIBUTED INTELLIGENCE

A membrane with a neuron in every cell

INTERACTIVE CONCEPT

Apply a harmonic input and explore how local displacement becomes neural traction. Cells share their nodes: each neuron's force feeds back into the continuous membrane.

Neuro-skin membrane and selected finite neuro-elementA planar mesh shows magnified longitudinal displacement and neural activation. Select a cell with the mouse, or focus the membrane and use arrow keys. The right panel shows four nodal displacements feeding the selected neuron and traction returning to each node.01 / NEURO-MEMBRANEIn-plane displacement · magnified · normalized illustrationHarmonic inputClick a cell to inspect its neuron →02 / INSPECT ONE CELLu₁u₂u₄u₃f(z)Each node receives Fᵢ = τ A / 403 / LOCAL INPUT & ACTIVATIONTraction is applied along the membrane.No out-of-plane bending is represented.Shared nodes transmit local feedback.
Negative activationPositive activationLocal input zActivation f(z)
DISPLACEMENT → ACTIVATION → TRACTION → MOTION

z = Σ wᵢuᵢ   →   τ = wₒ f(z)
Fᵢ = τ A / 4   →   M ü + Fᵢₙₜ = Fₑₓₜ

WHAT THE PAPER FOUND

Neuron type strongly affects the membrane response. For the tested sigmoid cases, refining 200 to 800 cells produced nearly indistinguishable overall responses under comparable total loading. Training was outside the study's scope.

The paper uses nonlinear dynamic finite elements, von Mises plasticity with isotropic hardening, and explicit central differences.

Conceptual animation: prescribed traveling displacement and normalized neuron responses; no finite-element equilibrium or plasticity solver is run here. Stop memory is illustrated with a bounded incremental operator. Mesh changes alter resolution, not a claimed numerical prediction. The paper also investigates step and saw-tooth neurons.

Joghataie & Shafiei Dizaji (2019), Neuro-Skins: Dynamics, Plasticity and Effect of Neuron Type and Cell Size on Their Response, Neural Processing Letters 49, 19–41 (online 2018).