Harvard OpenDP Hackathon
classification
Just a sec!
Please log in to access your credentials.
import antigranular as ag
session = ag.login(
"********", "************",
competition = "Harvard OpenDP Hackathon")About the competition
Caution: Excessive epsilon spending may significantly impact your score. To familiarize yourself with how the system works, practice spending epsilon efficiently in our sandbox competition.
Tip: Export important variables and save to a file for easy retrieval in case the kernel stops.
About the dataset
Submission
1)Using session object to submit and print the result
Just a sec!
Please log in to access your credentials.
result = session.submit_predictions(df_pred)
print(result)NOTE: session here is the session object assigned against ag.login
2)Using the ag_utils submit_predictions function
Just a sec!
Please log in to access your credentials.
submit_predictions(pd.DataFrame(y_pred))NOTE: here y_pred is the predicted (Private) DataFrame object
About the prize
We have €5500 total in prizes:
Category 1: Very accurate prizes - €3300 up for grabs
- 3 prizes will be awarded for the top submissions.
- You can earn €500 if you are in the top 3 with a 70% score.
- An additional €20 will be grated for every accuracy point above 70% (so you will get €1100 if you score 100% accuracy).
- A €10 deduction is applicable for each point below 70% (so you will get no money if your score is below 20% accuracy even if you are in the top 3).
Category 2: Best notebooks - €2000 up for grabs
- 20 prizes of €100 each will be awarded for the best notebook submissions.
- Good coding practices and detailed comments to explain the code are required (you have to have to make the notebook public to claim the prize).
- You can be eligible for this prize even if your accuracy score is not high.
Category 3: Special Open DP prize - €200 up for grabs
- 1 prize for the best use of OpenDP and AG in one solution.
All prizes are contingent on the submission of Python code and / or Notebooks.
Competition rules
Getting started
Sign Up and Log In!
First things first, create an account on the platform. You can register via email or just use your Google or GitHub credentials.
Explore the competitions
Dive into the available competitions and choose the one you're interested in. Seems you have already picked this one!
Get your Jupyter Notebook ready
You can use any available option, including Google Colab or your local system.
Install the Antigranular package
Type in a "pip install antigranula" command to connect to Antigranular.
Secure enclave access
Connect to the secure enclave by copying the code block at the top right corner of the competition page which includes your custom credentials. Paste them into your Jupyter Notebook to connect to the secure enclave. Check the sample notebooks if in doubt.
You're in!
Upon successful login, you'll see the session ID and the %%ag cell magic registered to your system.
Ready to make a submission?
Submit a prediction by simply typing “session.submit_predictions (your_prediction_dataframe)” to find out how you rank on the leaderboard.
Spend wisely
Antigranular is not just about accuracy but also about using the least amount of privacy budget. Navigate the trade-off like a boss to come out on top. Head to the Antigranular Docs for details of the scoring system and epsilon best practice.
Got questions?
Join our Discord channel, where our team is ready to give you guidance and troubleshoot problems.