Audio Research Platform
Neural representation learning for audio.
Overview
A personal research project on learning representations of audio, using neural encoders and a few coordinated agents to find and characterize structure in sound.
Approach
A set of neural audio encoders producing embeddings, a custom contrastive-learning pipeline with pluggable losses, projection heads, and samplers, and a small set of coordinated agents for analysis, training, and architecture. Fully local model stack.
Scope
A personal-interest research project.