Seq103: A Unified Neuroevolution Framework for Compact Sequence Architecture Discovery
Researchers have developed Seq103, a novel neuroevolution framework designed to discover compact sequence architectures. This unified system utilizes a shared evolutionary backbone with an optional recurrent extension to handle both step-wise recurrent and sample-wise feedforward sequence classification tasks. Seq103 demonstrates significant parameter reduction, retaining a high percentage of baseline accuracy across various text classification and time-series datasets. AI
IMPACT This framework could enable more efficient development of sequence models by reducing parameter count while maintaining performance.