An inhibitory weight initialization improves the speed and quality of recurrent neural networks learning.
Jean-Philippe DrayeDavor PavisicGuy CheronGaetan LibertPublished in: Neurocomputing (1997)
Keyphrases
- recurrent neural networks
- cascade correlation
- learning process
- feedforward neural networks
- neural network
- feed forward
- supervised learning
- high quality
- active learning
- real time
- reinforcement learning
- recurrent networks
- learning systems
- learning tasks
- learning algorithm
- incremental learning
- biologically plausible
- long short term memory
- online learning
- long term
- learning environment