[FRIAM] All Your Them Are One Model To Us Learn

[Marcus Daniels]

Roger writes:

“So this doesn't mean much to simulation modelers for the moment.”

Consider predicting a frame of 24 bit color values instead of a binary scalar (Labradoodle vs. Chicken).  With such an encoding, one could possibly `find’ the equations of motion in the learned neural net by watching a Labradoodle run.   In that sense I think it is of interest to simulation modelers.   Agent rules for free.

Marcus

From: Friam [mailto:friam-bounces at redfish.com] On Behalf Of Roger Critchlow
Sent: Tuesday, June 20, 2017 11:29 AM
To: The Friday Morning Applied Complexity Coffee Group <friam at redfish.com>
Subject: Re: [FRIAM] All Your Them Are One Model To Us Learn

Different kind of model, Merle, for a deep neural net the model is the architecture of untrained layers and the connections between them.  Then you train the model on a discrimination task, like distinguishing labradoodles from fried chicken.  So this doesn't mean much to simulation modelers for the moment.

-- rec --

[Inline image 1]

On Tue, Jun 20, 2017 at 11:56 AM, Merle Lefkoff <merlelefkoff at gmail.com<mailto:merlelefkoff at gmail.com>> wrote:
Wow, Roger.  I don't get all this, but tell me---does this change everything for modelers?

On Tue, Jun 20, 2017 at 9:39 AM, Roger Critchlow <rec at elf.org<mailto:rec at elf.org>> wrote:
Apropos my babbling about all patterns being patterns and all the mechanisms that recognize patterns being an incomprehensible jumble of mechanisms, Google shares One Model to Learn Them All, https://arxiv.org/abs/1706.05137, in which it turns out that throwing all the architectural elements from all kinds of deep learning into a single model ends up working pretty well.  Adding the recursive elements used to parse the linear sequences of elements in languages never hurts and mostly improves the performance of image classifiers and object recognizers.  Go figure.

Found from https://venturebeat.com/2017/06/19/google-advances-ai-with-one-model-to-learn-them-all/amp/

-- rec --


============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove



--
Merle Lefkoff, Ph.D.
President, Center for Emergent Diplomacy
emergentdiplomacy.org<http://emergentdiplomacy.org>
Santa Fe, New Mexico, USA
Visiting Professor in Integrative Peacebuilding
Saint Paul University
Ottawa, Ontario, Canada

merlelefkoff at gmail.com<mailto:merlelefoff at gmail.com>
mobile:  (303) 859-5609<tel:(303)%20859-5609>
skype:  merle.lelfkoff2

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/ by Dr. Strangelove

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20170620/774ad991/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.png
Type: image/png
Size: 743951 bytes
Desc: image002.png
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20170620/774ad991/attachment-0001.png>