⚡ Neural Synthesis Series (2024)

A collection of generative neural network visualizations exploring the intersection of artificial cognition and human perception.

Neural Layer 12 Visualization

Neural Layer 12 Visualization

Generated from a 12-layer GAN network trained on 500,000 quantum field simulations.

Synaptic Web at T=27s

Synaptic Web at T=27s

Time-lapse visualization of neural network weight adjustments during training convergence phase.

Fractal Decision Boundaries

Fractal Decision Boundaries

Interactive simulation of decision trees in adversarial training conditions, rendered into 2D projection space.

Technical Deep Dive

This series was created using a custom hybrid network combining convolutional and transformer architectures. The visualizations represent the activation patterns of 64,000 individual neurons trained on a dataset containing 1.2 million synthetic images generated via fractal noise and quantum wave function simulations.

Generation Process

  • • 8-week training period across 128 AWS P4d instances
  • • 145 epochs of adversarial refinement
  • • Batch normalization with exponential moving averages

Technical Specs

  • • Input dimensions: 4096 x 4096 x 3 x 32
  • • Activation functions: Softshrink + Mish
  • • Optimization: RAdam with warmup phase
Network Architecture

Explore More

Quantum Series Preview

2025 Quantum Series

Entanglement patterns in simulated spacetime curvatures.

About the Creator

About me, the artist, and my journey through algorithmic art.