Fig. 1From: Explicit-memory multiresolution adaptive framework for speech and music separationA brief view of the proposed framework incorporating the following bio-inspired design elements: (1) multi-view feature extraction with parallel streams trained independently to yield multiple read-outs, (2) a distributed network of memories for targets of interest used at each local stream to gate embeddings of interest, (3) feedback from the integrated output is propagated to re-tune local memories during inference (specialist system), (4) the architecture can be extended across multiple levels of a hierarchyBack to article page