Gradient-based learning drives robust representations in recurrent neural networks by balancing compression and expansion.
Matthew FarrellStefano RecanatesiTimothy MooreGuillaume LajoieEric Shea-BrownPublished in: Nat. Mach. Intell. (2022)
Keyphrases
- recurrent neural networks
- learning algorithm
- learning process
- recurrent networks
- neural network
- cascade correlation
- learning systems
- reservoir computing
- echo state networks
- feedforward neural networks
- online learning
- supervised learning
- prior knowledge
- learning tasks
- feed forward
- long term
- support vector
- machine learning