Open in TheBrain

Structural information theory SIT Leeuwenberg van der Helm

<p>Jun 29, 2011 09:08 PM &#160;SIT grew out of Gestalt .... it is an extension of Gestalt Psychology.

&#160;

SIT initiated in 1960s by Emanuel Leeuwenberg

&#160;

developed further by Peter van der Helm &#160;(PhD supervisor was Leeuwenberg.

&#160;

Most prominently, he advocated<b> the simplicity principle</b>

<b>which states that, from among all possible interpretations of a pattern, the visual system</b>

<b>selects the interpretation with the simplest description</b>.

&#160;

<b>SIT has been argued to be the best defined and most successful extension of </b><a title="Gestalt psychology" href="/wiki/Gestalt_psychology"><b>Gestalt ideas</b></a><b>.</b><sup id="cite_ref-3" class="reference"><a href="#cite_note-3"><b>[</b>4]</a></sup>&#160;

&#160;

It is the only Gestalt approach providing a <a title="Calculus" href="/wiki/Calculus">formal calculus</a> that generates plausible <a class="mw-redirect" title="Percept" href="/wiki/Percept">perceptual interpretations</a>.

&#160;

SIT : &#160;the perceptually preferred interp. is the one with the SIMPLEST CODE.

&#160;

simplest code = minimum info load, ie &#160;a code that enables reconstruction using a

minimum # of descriptive parameters &#160;(note relation to AIT below &#160;(Marcus Hutter et al))

&#160;

Such a code is obtained by capturing a <b>maximum amount of visual regularity and</b>

<b>yields a hierarchical organization of the stimulus in terms of wholes and parts.</b>

&#160;

The assumption that the <b>visual system prefers simplest interpretations is called the simplicity principle</b>.<sup id="cite_ref-4" class="reference"><a href="#cite_note-4">[5]</a></sup>&#160;

&#160;

Historically, the <b>simplicity principle is an information-theoretical descendant of the Gestalt law of Prägnanz</b>,<sup id="cite_ref-5" class="reference"><a href="#cite_note-5">[6]</a></sup>&#160;

&#160;

which was based on the <b>natural tendency of physical systems to settle into stable minimum-energy states</b>.

&#160;

&#160;Furthermore, just as the later-proposed<b>&#160;</b>

<a title="Minimum description length" href="/wiki/Minimum_description_length"><b>minimum description length principle</b></a><b> in </b><a title="Algorithmic information theory" href="/wiki/Algorithmic_information_theory"><b>algorithmic information theory</b></a><b> (AIT),</b>

<b> it can be seen as a formalization of </b><a class="mw-redirect" title="Occam's Razor" href="/wiki/Occam%27s_Razor"><b>Occam's Razor</b></a><b> in which the best hypothesis for a given set of data is the one that leads to the largest compression of the data</b>.

&#160;

Structural vs Algorithmic Information Theory...

&#160;

SIT encodes for a restricted set of perceptually relevant kinds of regularities

in SIT, the relevant outcome of an encoding is a hierarchical organization.

&#160;

&#160;

&#160;

&#160;

&#160;

&#160;</p>

Attachments