• Quantitative Modelling, Data Science

Mentors :

  • Shubh Kumar, Sankalp Parashar & Ashwin Ramachandran

Mentees :

  • 2-10

Numer.AI is a large and Decentralised Data Science Contest supported by many big names in the field, including Co-founders of Rennaisance Technologies. Proclaimed as the last hedge fund on Earth. Numer.AI is about building generalised Data Science Strategies for the actual market, which the people at the firm use to actually trade in the markets, and pay you back in their own Cryptocurrency.

This is not a machine learning project, in the sense that we won't actually be using any heavy training requiring models. Instead we'll try to use the advanced versions of basic techniques in Bayesian Analysis for our purposes.
The Goal of this SoC Project would be to design a model which we could collectively submit, and henceforth participate in the contest.
We would also be ourselves designing models to participate in the contest and the mentees' models would be complementing ours using the ensemble learning strategies we learn about.
Prerequisites: Python, A strong hold on Probability, Having done any course equivalent to CS 215 (Data Analysis and Interpretation) is a plus, so is any prior experience in Data Science or Ensemble Learning.

Tentative Timeline :

Week Work
Week 1 Go through material on Data Science and light weight Gradient-based Optimization and getting acquainted with Bayesian Learning
Week 2 Solve some basic challenges in Data Science and Bayesian Learning
Week 3 Explore various papers which seem fit for your purposes in Numer.AI
Week 4 Start Implementing versions of the paper
Week 5 Continue implementation, while at the same time exploring Methods for ensembling the models
Week 6 Final Debugging and submission