Best of $100k Causal Discovery Challenge
The last weeks of November were exciting for the causal discovery community.
After 12 weeks of active competition, @ADIAlab's and @crunchdao 's $100k causal discovery challenge...
1/🧵
#CausalSky
The last weeks of November were exciting for the causal discovery community.
After 12 weeks of active competition, @ADIAlab's and @crunchdao 's $100k causal discovery challenge...
1/🧵
#CausalSky
Comments
We announce events like this + share curated papers and other resources.
9/9 ✋🏼
I think there are some things to think about here -- both potential benefits and drawbacks -- I have some ideas for a blog post discussing them.
Would be great to have a discussion on this topic.
8/
Congrats to all winners!
It's inspiring to see all creative ideas developed during the competition.
We're looking fwd to learning more about them soon
7/
Her ML model for predicting graph structure achieved 70.37% acc on the test set.
Hoang's Deakin colleagues Bao Duong, Quang-Duy Tran, & Thin Nguyen secured the 7th, 9th, and 10th spots, respectively.
6/
His solution achieved 70.50% acc, roughly 5% less than the ShanghaiTech & Xiamen University team, whose solution secured 3rd place
6/
His solution was followed by a hybrid approach developed by Andre Franca & Alexis Gassman
They combined a modified PC algo with a VAE-based model, and a graph search algo, all integrated using an ensemble layer, achieving 74.47%
5/
🥈 $20k: Former R&D VP at causaLens, Andre Franca (Brazil) and Alexis Gassman (France)
🥇 $40k: Hicham Hallak (France)
The winning solution, a CNN-based algorithm developed by an applied data scientist, Hicham Hallak, achieved 76.70% acc.
4/
The top three solutions were:
3/
This globally accessible competition attracted researchers and practitioners from around the world.
2/
And yes, no prior expert knowledge was taken into account.