Deep Q Learning
Before diving into the details of the deep Q learning process, let's first discuss the disadvantages of the traditional tabular Q learning process, and then we will look at how combining deep learning with Q learning can help us to resolve these disadvantages of tabular methods.
The following describes several disadvantages of the tabular Q learning approach:
- Performance issues: When the state spaces are very large, the tabular iterative lookup operations will be much slower and more costly.
- Storage issues: Along with the performance issues, storage will also be costly when it comes to storing the tabular data for large combinations of state and action spaces.
- The tabular method will work well only when an agent comes across seen discrete states that are present in the Q table. For the unseen states that are not present in the Q table, the agent's performance may be the optimal performance.
- For continuous state spaces for the...
