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> - 64 - Universal Induction & Intelligence<br />

Universal Inductive Inference: Summary<br />

Universal Solomon<strong>of</strong>f prediction solves/avoids/meliorates many problems<br />

<strong>of</strong> (Bayesian) <strong>in</strong>duction. We discussed:<br />

+ general total bounds for generic class, prior, and loss,<br />

+ i.i.d./universal-specific <strong>in</strong>stantaneous and future bounds,<br />

+ the D n bound for cont<strong>in</strong>uous classes,<br />

+ <strong>in</strong>difference/symmetry pr<strong>in</strong>ciples,<br />

+ the problem <strong>of</strong> zero p(oste)rior & confirm. <strong>of</strong> universal hypotheses,<br />

+ reparametrization and regroup<strong>in</strong>g <strong>in</strong>variance,<br />

+ the problem <strong>of</strong> old evidence and updat<strong>in</strong>g,<br />

+ that M works even <strong>in</strong> non-computable environments,<br />

+ how to <strong>in</strong>corporate prior knowledge,<br />

− the prediction <strong>of</strong> short sequences,<br />

− the constant fudges <strong>in</strong> all results and the U-dependence,<br />

− M’s <strong>in</strong>computability and crude practical approximations.

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

Saved successfully!

Ooh no, something went wrong!