Adaptive model improves long-range sequence learning with efficient parallelism
ADPTNet: Adaptive with Prescriptive Timescales Non-Linear SSM for Sequence Modelling
Neural and Evolutionary ComputingArtificial IntelligenceDistributed, Parallel, and Cluster ComputingMachine Learning
Summary
Many AI systems struggle to remember information over long sequences without using lots of energy. The authors propose ADPTNet, an AI model that balances being adaptable, handling long-range info, running fast on modern hardware, and reasoning in complex ways. They base their design on how the brain processes sounds at fixed timescales, enabling predictable yet flexible behavior. Their model shows better performance on memory tasks and speech recognition while using fewer parameters and allowing new efficient ways to run in parallel.
What this means in practice
- •For speech recognition engineers: Improve spiking neural network accuracy for low-power speech command recognition with ADPTNet-based models.
- •For machine learning framework developers: Integrate Jacobian-free parallel simulation methods to speed up non-linear RNN training without extra computation.
Authors
Matei-Ioan Stan, Oliver Rhodes
Abstract
A central aim of neuromorphic computing is to provide a viable alternative to highly energy-intensive Transformer-based AI. However, efficient alternatives struggle to capture the set of qualities that have secured the Transformer's status as the de facto standard in sequence modelling. Any realistic contender must be data-adaptive, able to capture long-range dependencies, and GPU-parallelisable, but also non-linearly recurrent to enable complex reasoning. Based on evidence suggesting the auditory cortex operates on fixed timescales, this work proposes the ADaptive with Prescriptive Timescales Network (ADPTNet) as a potential solution to achieving all four properties simultaneously. ADPTNet is built around local topological conjugates, obtained by a novel combination of linear attention and Riemannian optimisation, applied to static global dynamics. This enables non-linear yet predictable long-term behaviour. Dynamical systems theory proofs provide theoretical guarantees for the parametric control of ADPTNet's timescales (its Lyapunov spectrum). ADPTNet improves performance on Selective Copying over Hawk, the existing method balancing long-range memory and adaptability, while also improving state tracking over linear SSMs like Mamba. On sequential CIFAR-10, ADPTNet matches linear SSM accuracy and outperforms existing selective models (incl. the Transformer), using fewer parameters. We also introduce a neuromorphic SpikingADPTNet, which achieves a new state-of-the-art accuracy on the Spiking Speech Commands dataset ($83.56\%\pm0.15$). Finally, ADPTNet's constant timescales enable two efficient, Jacobian-free extensions to the DEER parallel simulation algorithm (Conv and Forward DEER) that retain the same average convergence. Conv DEER adds no computational overhead beyond the network's forward pass and enables non-linear RNN parallelisation via iterated convolutions for the first time.