28.01.2015 Views

Slides in PDF - of Marcus Hutter

Slides in PDF - of Marcus Hutter

Slides in PDF - of Marcus Hutter

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Marcus</strong> <strong>Hutter</strong> - 80 - Universal Induction & Intelligence<br />

Particularly Interest<strong>in</strong>g Environments<br />

• Sequence Prediction, e.g. weather<br />

√<br />

or stock-market prediction.<br />

Strong result: Vµ ∗ − Vµ pξ = O( ), m =horizon.<br />

K(µ)<br />

m<br />

• Strategic Games: Learn to play well (m<strong>in</strong>imax) strategic zero-sum<br />

games (like chess) or even exploit limited capabilities <strong>of</strong> opponent.<br />

• Optimization: F<strong>in</strong>d (approximate) m<strong>in</strong>imum <strong>of</strong> function with as few<br />

function calls as possible. Difficult exploration versus exploitation<br />

problem.<br />

• Supervised learn<strong>in</strong>g: Learn functions by present<strong>in</strong>g (z, f(z)) pairs<br />

and ask for function values <strong>of</strong> z ′ by present<strong>in</strong>g (z ′ , ) pairs.<br />

Supervised learn<strong>in</strong>g is much faster than re<strong>in</strong>forcement learn<strong>in</strong>g.<br />

AIξ quickly learns to predict, play games, optimize, and learn supervised.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!