Logarithmic Regret for Learning Linear Quadratic Regulators Efficiently
We consider the problem of learning in Linear Quadratic Control systems whose transition parameters are initially unknown. Recent results in this setting have demonstrated efficient learning algorithms with regret growing with the square root of the number of decision steps. We present new efficient...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Published: |
2020
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2002.08095 http://arxiv.org/abs/2002.08095 |
| Tags: |
No Tags, Be the first to tag this record!
|