Sleep prevents catastrophic forgetting in spiking neural networks by forming joint synaptic weight representations

doi:10.1371/journal.pcbi.1010628

Original source · Paper JSON

NeuroAI · 2020-01-01

Cites (2) · Cited by (0)

Complete cleaned Markdown · Source Markdown including bibliography

Read in parts

1. Part 1
2. Part 2
3. Part 3
4. Part 4
5. Part 5

Contents

• Sleep prevents catastrophic forgetting in spiking neural networks by forming joint synaptic weight representations
• Abstract
• Introduction
• Results
• Complementary complex foraging tasks can be robustly learned
• Catastrophic forgetting occurs following sequential but not interleaved training
• Periods of sleep allow for sequential training without catastrophic forgetting
• Receptive fields of decision-making neurons after sleep represent multiple tasks
• Periods of sleep allow reintegration of new task without interference through renormalization of task-relevant synapses
• Periods of sleep push the network towards the intersection of the solution manifolds representing Task 1 and Task 2 specific weight configurations
• Discussion
• Methods
• Environment
• Network structure
• Policy
• Neuron models
• Synaptic plasticity
• Simulated Sleep
• Support Vector Machine Training
• 2-D Synaptic Weight distributions (Figure 6)
• Distance from Solution Manifolds (Figure 7)
• Author Contributions
• Conflicts of Interest
• Acknowledgements