3/x: Generator continuing to learn an anatomical grammar of cortical fibre architecture.

4/x: Reusable Fibre-Architecture Motifs
Network disentangles and recombines stereotyped micro-patterns (U-fibres, radial glia scaffolds, long tracts) not memorizing

___
@r3rt0.bsky.social @k4tj4.bsky.social

Does the learned latent space: capturing diverse neural microstructures, function as a 'micro-morphospace' allow us to navigate variations in fiber architecture?

Simulation of fiber patterns might change under different hypothetical conditions a data-driven approach for mechanistic simulations?

| [G]⇆[D] 2 M k-img | 24 M latent frames ψ≤0.5 → fibre-bundles braid

W-space sweep ▶︎ 3-vol ensemble reveals cross-slice links.

Next: Working hypothesis why these emergent patterns might reflect fundamental constraints in cortical organization.

@r3rt0.bsky.social @k4tj4.bsky.social

Working on micro-scale connectome down-sampling: pruning full HCP tractograms to ≈ 100 k fibres so a WebGPU viewer can hit 120 fps on everyday laptops.

Prototype render @ 500k ↘︎

Diffusion tensors dancing, every voxel a footstep in the brain’s silent choreography.

Multi-shell diffusion MRI from the Human Connectome Project during my time working with UCLA Semel Neuro.

Experiments with new diffusion-field renderer : poisson-disc jitter + gamma-mapped palette.

PANEL:Technoetics — Media — Art and everything in between
AT: Château du Feÿ

Glitch Art Residency
Part I — What is (ART)ificial Intelligence?

Dr. Iannis Bardakos x Rish Chakrabarty

Moderated by — Gordon Berger

|Château du Feÿ, Castle in Villecien, France