Lernkarten
realizable
Trade off Accuracy vs. Generalization
Tradeoff between complex hypothesis that fit well and a simpler hypothesis that generalize well
Supervised learning (x,y)? y=f(x)? and h?
(x,y) input-output pair
y = f(x) true function to be approximated
h hypothesis
inducitve learning
learning a general function (generalization)
deductive learning
from known general function to a new rule
reinforcment learning
agent decides at the end by its own if id did right (exampel with taxi driver - tip or no tip)
or how to act or behave when given occasional reward
No free lunch theorem
No universally best Model. A set of assumptions work well for a Problem A but necesseraly for Problem B. Diffrent Models lead to diffrent Algo.
Types of Machine Learning
Supervised: input-output value pair
Unsupervised: classification
Reinforcment:
