Inductive Learning

The inductive learning problem $\mathcal{P}:=\langle\mathcal{H}, f, \approx\rangle$ is made up of a

• Hypothesis Space $\mathcal{H}$
• the to be found hypothesis $h\in \mathcal{H}$
• and a consistent training set $f$ of examples ($x,y$ pairs)

We want to find the hypothesis $h$ such that

which means we want to find a function $h$ that is on the domain of $f$ roughly equal to $f$.

For inductive learning we assume

• no prior knowledge
• deterministic Fully Observable Environment
• examples are given
• the Agent wants to learn $f$

Consistency and Simplicity

Ockhams Razor