| id |
ndltd-NEU--neu-893
|
| record_format |
oai_dc
|
| spelling |
ndltd-NEU--neu-8932021-05-26T05:10:54ZNovel function approximation techniques for large-scale reinforcement learningFunction approximation can be used to improve the performance of reinforcement learners. Traditional techniques, including Tile Coding and Kanerva Coding, can give poor performance when applied to large-scale problems. In our preliminary work, we show that this poor performance is caused by prototype collisions and uneven prototype visit frequency distributions. We describe our adaptive Kanerva-based function approximation algorithm, based on dynamic prototype allocation and adaptation. We show that probabilistic prototype deletion with prototype splitting can make the distribution of visit frequencies more uniform, and that dynamic prototype allocation and adaptation can reduce prototoype collsisions. This approach can significantly improve the performance of a reinforcement learner.http://hdl.handle.net/2047/d20000932
|
| collection |
NDLTD
|
| sources |
NDLTD
|
| description |
Function approximation can be used to improve the performance of reinforcement learners. Traditional techniques, including Tile Coding and Kanerva Coding, can give poor performance when applied to large-scale problems. In our preliminary work, we show that this poor performance is caused by prototype collisions and uneven prototype visit frequency distributions. We describe our adaptive Kanerva-based function approximation algorithm, based on dynamic prototype allocation and
adaptation. We show that probabilistic prototype deletion with prototype splitting can make the distribution of visit frequencies more uniform, and that dynamic prototype allocation and adaptation can reduce prototoype collsisions. This approach can significantly improve the performance of a reinforcement learner.
|
| title |
Novel function approximation techniques for large-scale reinforcement learning
|
| spellingShingle |
Novel function approximation techniques for large-scale reinforcement learning
|
| title_short |
Novel function approximation techniques for large-scale reinforcement learning
|
| title_full |
Novel function approximation techniques for large-scale reinforcement learning
|
| title_fullStr |
Novel function approximation techniques for large-scale reinforcement learning
|
| title_full_unstemmed |
Novel function approximation techniques for large-scale reinforcement learning
|
| title_sort |
novel function approximation techniques for large-scale reinforcement learning
|
| publishDate |
|
| url |
http://hdl.handle.net/2047/d20000932
|
| _version_ |
1719406459793440768
|
