menu
Kyla Jeffrey HeroX team
Sept. 4, 2018
10:30 a.m. PDT

Announcing the GF Challenge Final Workshop

Dear Solvers,

As noted in the rules update last week, all Solvers are invited to attend the GF Challenge Final Workshop taking place in the Northern Virginia area from October 23-24; this workshop runs in conjunction with the HFC Principal Investigators’ meeting.

The top 5 Solvers on the leaderboard on September 23rd will be invited to present their solution at this Final Workshop. These Solvers will receive a $2,500 Workshop Presenter Prize, as outlined in the updated Rules Document. Please keep an eye out for an email announcing the winners of this Workshop Presenter Award.

While we only have a few presenter slots, we invite all GF Solvers to join us at this special event in celebration of your participation in the GF Challenge. It’s a great opportunity to network with HFC performers, IARPA officials, and fellow Solvers. More information on the event, schedule, and invitations to come.

Cheers!
GF Challenge Team
24 Replies

Lars Ericson
Sept. 15, 2018
5:24 a.m. PDT
@Ian Pitchford
I'm in the Below 1000 bin along with most everybody else. Luckily I was there much longer than the Consensus due to an early frequency-based estimate that I didn't update (because my data uploader for FluNet broke and I didn't notice). They've updated the language from "each Monday thereafter" to "fish or cut bait": https://www.herox.com/IARPAGFChallenge/forum/thread/3444
Attachments

Ian Pitchford
Sept. 15, 2018
5:39 a.m. PDT
The consensus is 43% on Carbon.
Attachments

Lars Ericson
Sept. 15, 2018
5:49 a.m. PDT
That's what they're showing you. Their real consensus is the bottom line in the picture above which peaks towards the end. The reasoning for my model is in the picture below of my homebrew system. The top of the page shows my model, the current consensus, how they differ, and what I think my NBP should be if the model is right (calculated branchwise, which is not totally correct, but at least indicative). The first picture shows the history of Australia Flu. On the right hand of the picture, tilted sideways, is a picture of the frequency distribution (PDF) for flu, with colored markers to break it up into the different branch levels. The bottom picture is a closeup of the flu PDF, with the current level (the last time I measured it) marked. My model is just saying that, going by the PDF, there is a 71% chance that the level will be below 1000, based on prior history for that time of year (what "seasonal" means in the model description). So it is actually completely ignoring what the current level is (note to self: fix that, especially towards the end of a question).
Attachments

Ian Pitchford
Sept. 15, 2018
6:02 a.m. PDT
The bottom line is “between 2500 and 3200 inclusive”.
Attachments