Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2022
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2207.11439 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909340445704192 |
|---|---|
| author | Martín-Sánchez, Guillermo Bohté, Sander Otte, Sebastian |
| author_facet | Martín-Sánchez, Guillermo Bohté, Sander Otte, Sebastian |
| contents | Backpropagation through time (BPTT) is the de facto standard for training recurrent neural networks (RNNs), but it is non-causal and non-local. Real-time recurrent learning is a causal alternative, but it is highly inefficient. Recently, e-prop was proposed as a causal, local, and efficient practical alternative to these algorithms, providing an approximation of the exact gradient by radically pruning the recurrent dependencies carried over time. Here, we derive RTRL from BPTT using a detailed notation bringing intuition and clarification to how they are connected. Furthermore, we frame e-prop within in the picture, formalising what it approximates. Finally, we derive a family of algorithms of which e-prop is a special case. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2207_11439 |
| institution | arXiv |
| publishDate | 2022 |
| record_format | arxiv |
| spellingShingle | A Taxonomy of Recurrent Learning Rules Martín-Sánchez, Guillermo Bohté, Sander Otte, Sebastian Machine Learning Backpropagation through time (BPTT) is the de facto standard for training recurrent neural networks (RNNs), but it is non-causal and non-local. Real-time recurrent learning is a causal alternative, but it is highly inefficient. Recently, e-prop was proposed as a causal, local, and efficient practical alternative to these algorithms, providing an approximation of the exact gradient by radically pruning the recurrent dependencies carried over time. Here, we derive RTRL from BPTT using a detailed notation bringing intuition and clarification to how they are connected. Furthermore, we frame e-prop within in the picture, formalising what it approximates. Finally, we derive a family of algorithms of which e-prop is a special case. |
| title | A Taxonomy of Recurrent Learning Rules |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2207.11439 |