Dynamical Learning of Dynamics

2019/02/28 by Christian Klos, Yaroslav Felipe Kalle Kossio, Sven Goedeke +2 · 58 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · Neuroscience · Psychology · #Artificial intelligence #Chaotic #Computer science #Dynamical systems theory #Dynamics (music) #Imitation #Neural Networks and Applications #Neural Networks and Reservoir Computing #Neural dynamics and brain function #Neuroscience #Physics #Psychology #Statistical physics #Variety (cybernetics) #q-bio.NC

paper · pdf · doi:10.1103/physrevlett.125.088103

published in Physical Review Letters 125(8), 088103 (American Physical Society)

openalex publication_date 2020/08/19 · arxiv created 2020/08/25 · arxiv updated 2020/08/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

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.

Citations

• Adam: A Method for Stochastic Optimization
• Controlling Recurrent Neural Networks by Conceptors
• Opening the Black Box: Low-Dimensional Dynamics in High-Dimensional Recurrent Neural Networks
• Neuromorphic electronic circuits for building autonomous cognitive systems
• CaMKII activation supports reward-based neural network optimization through Hamiltonian sampling
• RL²: Fast Reinforcement Learning via Slow Reinforcement Learning
• Neuromorphic Silicon Photonic Networks
• One step back, two steps forward: interference and learning in recurrent neural networks
• Long short-term memory and learning-to-learn in networks of spiking neurons

Cited by

• Inferring bifurcation diagrams of two distinct chaotic systems by a single machine
• Sustaining the dynamics of Kuramoto model by adaptable reservoir computer

Related

• Brain-like neural dynamics for behavioral control develop through reinforcement learning
• Network Dynamics-Based Framework for Understanding Deep Neural Networks
• One step back, two steps forward: interference and learning in recurrent neural networks
• Dynamical stability and chaos in artificial neural network trajectories along training
• Invertible generalized synchronization: A putative mechanism for implicit learning in biological and artificial neural systems