I am new to machine learning and just came across this website with few tournament trails. It seems an interesting way to protect user’s algorithms and host’s data while having cooperation in-between. However, as the features of the subjects are anonymous, things seems getting bored because I have no idea what I am working on the features. Does anyone have the same feeling?
Nope, if we knew what the features were, we would bring in our own “biases” from our own experiences (be that financial or scientific). Without knowing what the features are, pure machine learning techniques are the only things that matter.
It also helps Numerai in that anyone can participate. If the features were known, a lot of non-financial people (like me) might not even be bothered to take part, because everyone trained in finance might have an edge over me. With unknown features, everyone starts on the same level.
Yep, what themicon said. If this were not obfuscated data, I probably wouldn’t bother. I’ve been considering working on financial problems for a long time, but just never bother getting into it because there’s too much background and data and too little time to work on things in my spare time. But then I found Numerai and it’s pretty ideal. I get to work on my machine learning skills in a difficult dataset with lots of noise, and I don’t have to worry about domain knowledge at all because it’s just pure data to me. Plus I get to earn some money. Win win.