Phil 11.2.17

ASRC MKT 7:00 – 4:30

  • Add a switch to the GPM that makes the adversarial herders point in opposite directions, based on this: Russia organized 2 sides of a Texas protest and encouraged ‘both sides to battle in the streets’
  • It’s in and running. Here’s a screenshot: 2017-11-02 There are some interesting things to note. First, the vector is derived from the average heading of the largest group (green in this case). This explains why the green agents are more tightly clustered than the red ones. In the green case, the alignment is intrinsic. In the red case, it’s extrinsic. What this says to me is that although adversarial herding works well when amplifying the heading already present, it is not as effective when enforcing a heading that does not already predominant. That being said, when we have groups existing in opposition to each other, that is a tragically easy thing to enhance.
  • Hierarchical Representations for Efficient Architecture Search
    • We explore efficient neural architecture search methods and present a simple yet powerful evolutionary algorithm that can discover new architectures achieving state of the art results. Our approach combines a novel hierarchical genetic representation scheme that imitates the modularized design pattern commonly adopted by human experts, and an expressive search space that supports complex topologies. Our algorithm efficiently discovers architectures that outperform a large number of manually designed models for image classification, obtaining top-1 error of 3.6% on CIFAR-10 and 20.3% when transferred to ImageNet, which is competitive with the best existing neural architecture search approaches and represents the new state of the art for evolutionary strategies on this task. We also present results using random search, achieving 0.3% less top-1 accuracy on CIFAR-10 and 0.1% less on ImageNet whilst reducing the architecture search time from 36 hours down to 1 hour.
  • Continuing with the schema. Here’s where we are today: polarizationgameone

2 thoughts on “Phil 11.2.17

  1. Cindy Flatley

    Notes on our discussion today:
    In response to my suggestion of “The Buffalo Jump Effect” title for the new paper on adversarial herding (Phil suggested “pishkun” and the generalized term is “game driving systems” but those terms might be confusingly broad in connotation), we talked a bit about game driving systems research.
    – An overview of game driving systems- http://www.bioone.org/doi/abs/10.1657/1523-0430(2005)037%5B0425:TGDAAC%5D2.0.CO;2
    Tundra Game Drives: an Arctic-Alpine Comparison
    Stone fences and blinds built by prehistoric hunters to gather and ambush elk and bighorn sheep above timberline in the Colorado Front Range are similar in concept and function to structures built by the Copper Inuit and their predecessors for hunting caribou near Bathurst Inlet, in the Central Canadian Arctic. Four principal differences exist: (1) Circular blinds and continuous rock walls are more numerous in the Front Range than in the Arctic, where arcuate breastworks and lines of widely spaced cairns predominate. Differences in prey-species behavior are the most probable explanation. (2) Stone house foundations, meat-drying facilities, meat caches, kayak-storage racks, and fox and wolf traps occur near drive sites along caribou migration routes in the Bathurst Inlet region. The structures imply long-term habitation made possible by a plentiful meat supply. Comparable structures are absent above timberline in the Front Range because people retreated to warmer environments in winter, and because steep terrain and deep snow discouraged return visits to high-altitude caches. (3) The technique was adopted much earlier in Colorado than in the Central Canadian Arctic. The oldest Front Range drive systems were constructed while the Laurentide Ice Sheet still covered the Bathurst Inlet landscape. (4) Pedestrian game-drive hunting was abandoned in the Front Range soon after arrival of the horse (ca. a.d. 1700), but remained an integral part of Copper Inuit subsistence until the mid twentieth century. The rich ethnographic and oral history record of communal hunting in the Arctic is invaluable for interpreting the Colorado structures.
    – This article has maps and diagrams that might be helpful for simulation of game drive systems- https://s3.amazonaws.com/academia.edu.documents/38968758/FINAL_HOLT_UL_V2.pdf?AWSAccessKeyId=AKIAIWOWYYGZ2Y53UL3A&Expires=1509650511&Signature=1DIguG0L0qyhgvit2yYBt2%2BM2xE%3D&response-content-disposition=inline%3B%20filename%3DKuzitrin_Lake_and_Twin_Calderas_an_Examp.pdf
    Kuzitrin Lake and Twin Calderas: an Example of Optimal Land Use in the Late Holocene in Seward Peninsula, Alaska
    – It looks like the primary expert in this field was JB Benedict. Unfortunately, he died in 2011. However, it looks like they set up a memorial foundation dedicated to mountain archaeology in his name and Colorado State University might have some experts to talk to- http://central.colostate.edu/news/the-james-b-benedict-fund/
    Contributions of people who worked with JB Benedict can be found here:
    http://www.academia.edu/9231747/Footprints_in_the_Snow_Papers_in_Honor_of_James_B._Benedict_2012_edited_by_Jason_LaBelle_Steve_Cassells_and_Mike_Metcalf_
    – It would be interesting to produce an agent-based simulation of a buffalo jump, but I’m guessing Phil wouldn’t want to spend his time on it when there’s so much other fun stuff to explore in belief space.
    – It occurs to me that the visual effects field for movies might already have created agent-based simulations for herd behavior, maybe even buffalo jumps. I’m not sure what the best way would be to investigate that. I’ve read that London has a relatively stable visual effects industry based on tax incentives, whereas in the US the visual effects industry moves often because of new tax incentives being offered in different states. If Phil is interested, I can research who we could talk to about it.

    Phil suggested that we add rooms/subreddit type categories of discussions. I think that’s a good idea, as it could allow us to alert users to discussions that might interest specific users. – My initial thoughts of types of topic divisions:
    * Ethical Dilemmas
    * Conspiracy Theories
    * Pop Culture (Fan Theories)
    * Current Events/Politics/International Relations (maybe include forecasting, like from the Good Judgment Project/Superforecasting?)
    * Family/Interpersonal Issues – Parenting, Elder/Disability Care, Pets, Relationships, etc.
    * Science/Technology
    * Economics/Finance/Business
    * Societal Issues (Addiction, Race, Gender, etc.)
    * Health/Fitness/Beauty
    * Religion/Spirituality
    – Reddit might be a good place to look for discussion categories, so here’s a list of Reddit categories. The nested-bullet list in the middle of the page is particularly interesting:
    https://www.reddit.com/r/ListOfSubreddits/wiki/listofsubreddits#wiki_general_content
    – I think, at least initially, the Open Discussion Lobby and the Completed Discussions pages should be laid out in groups by category. I’d think this would need to grow significantly for people to go to specific pages. But it could be good to offer a click-through option.
    – Perhaps we should ask scenario submitters to mark their scenarios with category tags, though we might get different tags than we would initially assign. For instance, I might assign “conspiracy theory” to discussion of whether the moon landing was faked, but someone who tagged it might tag it “government propaganda.” It’s a place to start, though, and might produce some interesting data.

    The other day I mentioned that a friend of a friend went through a series of polarizing scenario discussions when trying out to be a cast member for The Bachelor. I’ve been looking at who we might talk to about getting access to this kind of data (surely an NDA agreement would need to be involved, but it still might be worth pursuing if there is a lot of data available). I came across two leads that might be interesting.
    – Kelsey Porter is a casting recruiter who worked on 25 several reality shows, so she should at minimum understand the field and be able to direct us in the remotely right direction. Also, she said, “If anybody’s interested in getting into [casting], they can e-mail me or contact me. I’m always happy to help anybody who’s interested in the casting world,” in this article – http://getinmedia.com/articles/film-tv-careers/kelsey-porter-casting-reality-tv. Also from the article: “Kelsey Porter can be reached at KelseyAlayneCasting@gmail.com or on Twitter @KiwiPorter.” If you’re okay with me doing so, I’ll send her an email to see what she says. If that doesn’t work, we could reach out to the author of the article, Christina Couch – http://getinmedia.com/users/christina-couch.
    – Sarah Gertrude Shapiro is a former producer of The Bachelor, who might be able to help from a non-attached-to-the-industry perspective. She now has a TV show satirizing reality television called UnReal – https://twitter.com/GertShap?ref_src=twsrc%5Egoogle%7Ctwcamp%5Eserp%7Ctwgr%5Eauthor
    – Also, the psychologist in charge of psychological testing for The Bachelor was listed as Dr. Catherine Selden. I found a listing for a Beverly-Hills-based psychologist with that name, though I can’t be 100% certain that it’s the same one: http://www.wellness.com/dir/3874981/psychologist/ca/beverly-hills/selden-catherine-dr#referrer

    Reply
  2. Cindy Flatley

    Also, when it comes to oppositional adversarial herding (making stampedes run in opposite directions), you’ve said it was easier to get the first group to clump than it was to get the opposite one. I wonder if that is part of why, in the article I sent you today (Vox, America is facing an epistemic crisis, https://www.vox.com/policy-and-politics/2017/11/2/16588964/america-epistemic-crisis), according to the Harvard study of online media (https://cyber.harvard.edu/publications/2017/08/mediacloud), “conservative media is more partisan and more insular than the left.” If the Russian bot army is herding them as the anchor and then herding the left away in opposition, it might be less coalesced based on the forces your model demonstrates.

    Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.