lyelibi

The issue is with the framing, the justification for the added overhead in training and productionalizing deep learning models is to take advantage of scaling laws: If you have more compute, and more data you should get better performance especially over xgboost architectures. So if we are to take what you said seriously it's even less reasons to invest in tabular transformers.

I have never seen tabular transformer models beat xgboost/catboost in industrial context where datasets is gigantic. They most produce these results on relatively small datasets, clearly not in the tens of millions of rows.

It's not just curve fitting because basis functions have characteristics which make them desirable for the kind of decomposition one is trying to find. We typically assume in factor analysis that factors are gaussian random variables without clear and repeating patterns. Fourrier transforms force us to think in similar terms but accounting for specific dynamics factor (I. E. Basis functions) should capture.

Also how do we construct those orthogonal basis functions for any downstream task is an interesting research question!

seems broken, at least the ML role

  Location: Houston, TX
  Remote: Yes
  Willing to relocate: No but can travel.
  Technologies: Python
  Résumé/CV: linkedin.com/in/lionelyelibi
  Email: [email protected]

I am a research scientist currently working in financial services. My background is in physics, data science and machine learning. I have broad interests so hit me to talk about your project ideas or your firm's mission. Over the last +8yrs I have worked as IC or lead in various academic and industrial research projects. I'm looking for a real challenge and a group of like-minded folks to build with.

Hey, I'd like to know more on what you guys are looking to do. Do I DM on twitter?