Reinforcement learning environments with musculoskeletal models
Completed
Submissions
Participants
Views
Posted by Marcel_Salathé about 1 year ago | Quote
Huge thanks to you @spMohanty - without your efforts above and beyond, this challenge would not have been possible!!
BIG thanks also to challenge organizer Lukasz Kidzinski (Stanford) and crowdAI developer Sean Carroll, who made all of this happen!
Posted by Lukasz about 1 year ago | Quote
Great thanks to all the competitors, organizers from the crowdAI side (Marcel, @spMohanty, and Sean) our RL advisor Sergey Levine and the entire OpenSim team (Stanford’s NMBL lab of Scott Delp). OpenSim team seemed silent in the challenge but they have been developing the musculoskeletal simulation software for over 10 years now. The challenge is just a cherry on top of all other great solutions enabled by this software.
The challenge is over, but now an even bigger challenge starts – turning the results into meaningful biomedical tools
Posted by Henryk about 1 year ago | Quote
Congratulations to the winners on the behalf of the deepsense.ai team (Piotr Miłoś, Błażej Osiński and me)!!! I would like to express thanks to all who contributed to the organization of this very interesting competition, in particular to spMohanty who tirelessly addressed all requests of participants!!! See you at the NIPS workshop!
Posted by QinYongliang about 1 year ago | Quote
This announcement is so official, that I think I should first appreciate the hard work the developers/organizers have gone through, then name a few institutions that supported this with abundant resources, and finally express gratitude to the honorable invitation.
Yea but let’s not forget those who spent their time to make the competition meaningful - without all those hundreds of participants’ efforts, we won’t be able to confirm/conclude that the winning entry is indeed the state-of-the-art. It’s those who didn’t receive the prize made the prize heavier.
I’d like to mention a few of them in acknowledgment, the Chinese speaking participants in particular:
- USTC-ICML (USTC
- hzwer (PKU
- JGeek (THU
- J1 (UIUC
- SDU_VSIS (SDU
- Yujin
- VsonicV
Greatest respect for them and all other participants.
Thank you everyone for taking part in the NIPS 2017: Learning to Run Challenge. The level of engagement by all of you in this challenge, more than justified all the effort that went into setting this up. Congratulations to every single one of you for having made that Skeleton run !!
We finally have a winner !!
Congratulations to @NNAISENSE for winning the NIPS 2017: Learning to Run Challenge with a top score of
45.97in Round 2.The runners up are @PKU-hzwer and @reason8-ai with top scores of
41.66and38.42respectively. Many congratualtions to you guys too.The seeds used for grading the submissions in Round-2 were :
[7898, 98754, 2420, 22142, 75023, 8536, 28194, 42895, 89646, 85729]and here are the score breakdowns for the top-3 participants in round 2 across all the 10 simulations :@NNAISENSE :
[45.750899629645666, 46.53111152686293, 45.571554566763176, 45.71011646542387, 45.56303399366884, 46.46241976938794, 46.11001265670976, 46.22358191551397, 45.886406626376896, 45.846110215678486]@PKU-hzwer:
[41.80766896924946, 41.43754450225851, 41.63060117493113, 38.93953241563359, 42.07344184297542, 40.82957871280918, 43.251539553334915, 41.67377496365344, 42.42926225572355, 42.547970378117036]@reason8-ai:
[39.21013402035226, 37.8836218623784, 38.02101691718449, 38.069635054231355, 38.40310998692697, 38.57608903483628, 39.91650968241817, 37.53747605857543, 38.30373582175789, 38.30493160781083]@NNAISENSE is the proud owner of a NVIDIA DGX-Station now. The other sub prizes include, a travel grant plus an invitation to present their work at the NIPS-2017 Challenge Track workshop in Long Beach, California; and a travel grant plus an invitation to present their work at Applied Machine Learning Days, EPFL, Switzerland.
We would also like to have a special mention for @qin-yongliang (aka @ctmarko), for going above and beyond in helping other participants getting started in the challenge, from making introductory blog posts about the challenge, to releasing his own code for other participants to use. To honor his efforts, and celebrate this amazing spirit of open dissemination of ideas, we happily extend a travel grant and an invitation for @qin-yongliang to Applied Machine Learning Days at EPFL, Switzerland to present his work and his experience in this challenge.
We will be getting in touch individually with the participants regarding the logistics of the prizes.
Apart from that, we will be getting in touch with some of the top participants in both Round-1 and Round-2 to contribute to the book chapter for the NIPS 2017 Challenge Track. If you are interested to contribute to the same, please do get in touch with us.
Oh..and if any of you write blog posts around your solutions for this challenge, do drop us a line, we would be happy to both promote them, and refer them in the resources section of this challenge.
And finally, a challenge like this would not be possible without the efforts of a serious group of really committed people !! While many many people contributed in different ways to make this challenge happen; this challenge simply would not have been possible without Lukasz Kidzinski, Sean Carroll and Marcel Salathé. Thank you so much guys !!
Cheers,
Mohanty (on behalf of the CrowdAI team and the Learning2Run Challenge organizing team)