Flashcards
Artificial Intelligence
Cards in this deck (7)
- 1What is supervised learning?Supervised learning is a learning that the computer is given sample inputs and desired output. Computers mainly learn the model.
- 2What is unsupervised learning?Unsupervised learning is a learnin that the computer does not receive any desired output unlike supervised learning. The computer is expected to discover and present the intereseting structure in the data. The compute mainly learns the model.
- 3What is reinforcement learning?The reinforcement learning is a learning that the computer received the algorithmw with rewards and punishments. The computer learns controlling. RL uses training information that evaluates the actions rather than instructs by giving correct actions. The evaluative feedback is how good an action is.
- 4What is N-armed Bandit Problem?The problem that one tries to maximize the final reward from n different options. The greedy action is to choose the most rewarding one based on previous knowledge, exploits it,but exploring unknown actions would be helpful to find a better rewad in the long run.
- 5What is ε(Epsilon)-Greedy Action Selection rule?It is near-greedy rule that behave greedily most of the time, but every once in a while, with a small probabilility ε, they explore unknown actions.
ε Random
1-ε Greedy - 6What is one drawback of Epsolon-greedy selection?The drawback is to explore equally all actions with the same probability. With softmap, the non-greedy actions will be selected with probability proportional to their value estimate. However, if two actions were with the same estimated value, it makes sense to explore the action that is more uncertain.
- 7—
Information
- Description
- Flashcards on Artificial Intelligence, created by Masaki Nakamura on 15/01/2018.
- Created by
- Masaki Nakamura · Jan 2018
- Updated
- Jan 15, 2018
- Views
- 0
- Rating