A new research paper introduces LETHE, a self-referential system designed for sonic oblivion and implemented in SuperCollider. This architecture draws inspiration from Generative Adversarial Networks (GANs) but operates in a closed loop without external datasets or supervision after its initial setup. LETHE processes audio through a mixing matrix and delay lines, with its parameters evolving via a linear discriminator and a random-perturbation optimizer. The system's effectiveness in driving parametric evolution was demonstrated through ablation studies. AI
RANK_REASON The cluster describes a new research paper detailing a novel architecture. [lever_c_demoted from research: ic=1 ai=0.7]
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- Francesco Scagliola
- generative adversarial network
- Gotit.pub
- Hugging Face
- LETHE
- reinforcement learning
- ScienceCast
- SuperCollider
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →