Paper 1
Neural DNA (NDNA)
A compact genome for growing neural network architecture
machine-learningneural-architecturesparsitypytorch
research
the Neural DNA series — three papers exploring how a tiny learned “genome” can encode the wiring of neural and biological networks.
each paper builds on the previous one. start with Paper 1 if you are new to the idea.
Paper 2
Scaling Neural DNA to GPT-2
354 parameters wire a language model. 99,970:1 compression. Beats GPT-2 on 3 benchmarks.
machine-learningneural-architecturesparsitygpt2transformerpytorch
Paper 3
NDNA on Protein Interaction Networks
A 290-parameter genome learns which protein interactions matter for cancer, and the same genome works across breast, lung, colon, and prostate cancer.
machine-learningneural-architecturebiologycancergraph-neural-networkspytorch