full-FORCE: A Target-Based Method for Training Recurrent Networks

Abstract

Trained recurrent networks are powerful tools for modeling dynamic neural computations. We present a target-based method for modifying the full connectivity matrix of a recurrent network to train it to perform tasks involving temporally complex input/output transformations. The method introduces a second network during training to provide suitable "target" dynamics useful for performing the task. Because it exploits the full recurrent connectivity, the method produces networks that perform tasks with fewer neurons and greater noise robustness than traditional least-squares (FORCE) approaches. In addition, we show how introducing additional input signals into the target-generating network, which act as task hints, greatly extends the range of tasks that can be learned and provides control over the complexity and nature of the dynamics of the trained, task-performing network.

arXiv comment20 pages, 8 figures

References in corpus (3)

• Complexity without chaos: Plasticity within random recurrent networks generates robust timing and motor control
• Recurrent Network Models Of Sequence Generation And Memory
• Stimulus-Dependent Suppression of Chaos in Recurrent Neural Networks

Cited by in corpus (27)

• Physical reservoir computing -- An introductory perspective
• A Survey on Reservoir Computing and its Interdisciplinary Applications Beyond Traditional Machine Learning
• Dynamical learning of dynamics
• Designing spontaneous behavioral switching via chaotic itinerancy
• How single neuron properties shape chaotic dynamics and signal transmission in random neural networks
• Training dynamically balanced excitatory-inhibitory networks
• Interpreting recurrent neural networks behaviour via excitable network attractors
• A Theoretical Framework for Target Propagation
• Beyond spiking networks: the computational advantages of dendritic amplification and input segregation
• Simulations Approaching Data: Cortical Slow Waves in Inferred Models of the Whole Hemisphere of Mouse
• Input correlations impede suppression of chaos and learning in balanced rate networks
• Target spiking patterns enable efficient and biologically plausible learning for complex temporal tasks
• Interpreting Deep Learning: The Machine Learning Rorschach Test?
• Composite FORCE learning of chaotic echo state networks for time-series prediction
• Emergent Computations in Trained Artificial Neural Networks and Real Brains
• Error-based or target-based? A unifying framework for learning in recurrent spiking networks
• Different eigenvalue distributions encode the same temporal tasks in recurrent neural networks
• R-FORCE: Robust Learning for Random Recurrent Neural Networks
• Dynamics and computation in mixed networks containing neurons that accelerate towards spiking
• Path classification by stochastic linear recurrent neural networks
• Recurrent Neural Network Learning of Performance and Intrinsic Population Dynamics from Sparse Neural Data
• A model of reward-modulated motor learning with parallelcortical and basal ganglia pathways
• Neuronal architecture extracts statistical temporal patterns
• Transfer between long-term and short-term memory using Conceptors
• Spiking Linear Dynamical Systems on Neuromorphic Hardware for Low-Power Brain-Machine Interfaces
• Unfolding recurrence by Green's functions for optimized reservoir computing
• Learning recurrent dynamics in spiking networks