State Space Models for Bioacoustics: A comparative Evaluation with Transformers
This work addresses efficiency challenges in bioacoustic analysis for researchers and practitioners, though it is incremental as it adapts an existing model to a new domain.
The study tackled the problem of applying state space models to bioacoustics by evaluating the Mamba model on the BEANS benchmark, showing that BioMamba achieved comparable performance to a state-of-the-art Transformer model while using significantly less VRAM.
In this study, we evaluate the efficacy of the Mamba model in the field of bioacoustics. We first pretrain a Mamba-based audio large language model (LLM) on a large corpus of audio data using self-supervised learning. We fine-tune and evaluate BioMamba on the BEANS benchmark, a collection of diverse bioacoustic tasks including classification and detection, and compare its performance and efficiency with multiple baseline models, including AVES, a state-of-the-art Transformer-based model. The results show that BioMamba achieves comparable performance with AVES while consumption significantly less VRAM, demonstrating its potential in this domain.