2020: Kaggle RL; 12 live contests; $1.5m+ in prizes
ML Contests: February 2020 Update
Hello!
Some news since the last update: expect to see more reinforcement learning (RL) competitions in 2020, including some from Kaggle. As opposed to the traditional supervised learning contests that Kaggle popularised, where you're submitting static answers and ranked on prediction accuracy, this type of competition requires you to submit an agent which can take actions in an environment. Kaggle only recently added support for these, and right now has one trial competition up: Connect X. Time to brush up on RL! I highly recommend Lilian Weng's overview.
A few interesting contests which went live in the last few weeks:
Bengali.AI Handwritten Grapheme Classification: despite Bengali being the 5th most spoken language in the world, OCR for Bengali script is currently not very good. In this challenge the goal is to improve recognition on three components of Bengali characters - and with 4 weeks left, there's $10k in prize money to be won.
Segmenting Buildings for Disaster Resilience: build models to identify building footprints from aerial imagery across diverse African cities. $15k in prizes, 4 weeks left.
Can ML models match simulations for predicting how different shapes bend under pressure? Asahi Kasei Corporation wants to find out. 2 weeks left, $21k in prizes.
Two contests for the ICLR Computer Vision for Agriculture Workshop: spotting crop diseases, and classifying fields by crop type. Just over a month to go, with $5k in prizes for each - and a trip to ICLR, one of the top ML conferences!
I've made a few tweaks to the website which should make it more usable on mobile - please hit reply and let me know if you have any feedback.
Happy ML hacking ;)
PS: I've started playing around with Paperspace for cloud GPUs. Use my code ZKGRZB5 to sign up and get $10 free credit!
Create your profile
Only paid subscribers can comment on this post
Check your email
For your security, we need to re-authenticate you.
Click the link we sent to , or click here to sign in.