Natural gradient algorithms for training deep neural networks
<p>In this thesis, we reduce the computation cost of a particular class of learning algorithms for Deep Neural Networks (DNNs; human-brain inspired artificial intelligence models). Natural Gradient Descent (NGD) algorithms have favorable properties as compared to Stochastic Gradient methods. K...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
2023
|
| Subjects: |