Loading...
Please wait, while we are loading the content...
Opleiding Informatica Psychology-Inspired Memory Sampling in a Deep Q Network
| Content Provider | Semantic Scholar |
|---|---|
| Author | Meyer, Samuel J. |
| Copyright Year | 2019 |
| Abstract | Deep Q-Learning uses a method called experience replay which re-uses past experiences of an agent in learning to complete some task. This increases the learning speed and information efficiency of the agent. Several psychology-inspired methods for memory sampling from replay memory were implemented. Performance of these was compared to two established methods for an exploration task in a partially observable environment. The psychology-inspired methods performed approximately as well as the standard method of random sampling, but worse than prioritized memories sampling. Due to being computationally much more expensive this makes them not recommendable. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://theses.liacs.nl/pdf/2018-2019-MeyerS.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |