Karteikarten
12 Learning
Karten in diesem Stapel (9)
- 1What is the definition of learning?Improving performance on future tasks after making observations about the world
- 2Which types of learning are there?-Unsupervised learning: in input
-Reinforcement learning: from series of rewards or punishments
-Supervised learning (!): observing example input-output pairs & function that maps from input to output - 3What is crucial for supervised learning?Hypothesis selection: maximum probability that a hypothesis belongs to a data set - check whether hypothesis generalizes wll & predicts other values outside the training set
- 4Which strategy does a decision-tree-learning algorithm adopt?Greedy divide-and-conquer:
-Test most important attribute first (splits examples into subsets that are ideally "all positive" or "all negative")
-Test divides problem into two smaller subproblems that can be resolved recursively - 5What can you do when remaining examples are all positive or all negative?Return yes or no and done
- 6What can you do when some examples are positive and some are negative?Choose best attribute to split them
- 7What can you do when there are no examples left?Return default value from plurality classification of all examples that were used in constructing the node´s parent (most common output value)
- 8What can you do when there are no attributes left, but positive and negative examples?There is a conflict because of noise in the data and domain is nondeterministic -> return plurality classification
- 9How can the importance of an attribute be quantified?Entropy: measure of the uncertainty of a random variable - acquisition of information corresponds to a reduction in entropy
Informationen
- Beschreibung
- Artificial Intelligence Karteikarten am 12 Learning, erstellt von Timo Bierwirth am 25/02/2018.
- Erstellt von
- Timo Bierwirth
- Aufrufe
- 1
- Bewertung