Dynamical learning of dynamics

Abstract

The ability of humans and animals to quickly adapt to novel tasks is difficult to reconcile with the standard paradigm of learning by slow synaptic weight modification. Here we show that fixed-weight neural networks can learn to generate required dynamics by imitation. After appropriate weight pretraining, the networks quickly and dynamically adapt to learn new tasks and thereafter continue to achieve them without further teacher feedback. We explain this ability and illustrate it with a variety of target dynamics, ranging from oscillatory trajectories to driven and chaotic dynamical systems.

References in corpus (10)

• Adam: A Method for Stochastic Optimization
• Neuromorphic Silicon Photonic Networks
• Neuromorphic electronic circuits for building autonomous cognitive systems
• RL${}^2$: Fast Reinforcement Learning via Slow Reinforcement Learning
• full-FORCE: A Target-Based Method for Training Recurrent Networks
• Learning universal computations with spikes
• Controlling Recurrent Neural Networks by Conceptors
• Brain-inspired photonic signal processor for periodic pattern generation and chaotic system emulation
• CaMKII activation supports reward-based neural network optimization through Hamiltonian sampling
• One step back, two steps forward: interference and learning in recurrent neural networks

Cited by in corpus (8)

• Adaptable Hamiltonian neural networks
• Dynamic learning of synchronization in coupled nonlinear systems
• Learning Hamiltonian dynamics by reservoir computer
• Optimized ensemble deep learning framework for scalable forecasting of dynamics containing extreme events
• Criticality in Reservoir Computer of Coupled Phase Oscillators
• Symmetry-Aware Reservoir Computing
• Target spiking patterns enable efficient and biologically plausible learning for complex temporal tasks
• 2022 Roadmap on Neuromorphic Computing and Engineering