Emergent structures in noisy channel message-passing
08-31, 13:55–14:10 (Europe/Zurich), HS 120

We will explain a mechanism for generating neural network glyphs, like the glyphs we use in human languages. Glyphs are purposeful marks, images with 2D structures used to communicate information. We will use neural networks to generate those structured images, by optimizing for robustness.

Colab Notebook | Slides | Blog Post | Github Repo

Emergent structures in noisy channel message-passing is a blog post I wrote explaining a few experiments I did with neural network image generation. The idea is that we want to generate images that are robust under some noise. So we generate images from random representations. We perturb them and we try to decode the initial representation. This leads to the generator learning to create images with 2D structures.

Public link to supporting material


Abstract as a tweet

Something out of nothing. Neural networks generate visual language for communication.

Project Homepage / Git



Discovery from Data, Data Visualisation, Image Processing, Machine Learning, Open Source Library, Vector and array manipulation

Expected audience expertise: Domain


Expected audience expertise: Python


I finished a masters in AI from Sofia University a few years ago. In my free time, I like doing neural network art, reading about emergence and cellular automata, or playing guitar. When I'm not home you can find me skiing, climbing or hiking.
I also sometimes do youtube - Ил Ай.