I don’t understand the economics of Brex. I used to work for a large, analytics-savvy credit card company, and people would freak out if the percentage of customers going bad (not paying their debt) exceeded something like 3-5%. Given that VC-funded startups fail at a rate of 11/12, isn’t Brex effectively throwing money down a massive black hole, assuming they’re extending credit? And this is at the top of credit cycle, probably within a few years of a recession, where a higher than normal percentage of good debt will go bad too.

After in-depth data exploration and feature engineering I often spent hours or days experimenting with various sklearn models and performing hyper parameter grid search. Now I use an open source package, TPOT (https://epistasislab.github.io/tpot/), that fully automates not only grid search but also selection of the learning algorithm (SVM, Random Forest, K-NN, etc). If your data set is small enough for sklearn, then you should not be writing sklearn code manually.

For those that are interested, TPOT uses evolutionary algorithms to perform its search.

Further down in the study it says that removing the gene likely occurred because it made humans more resistant to malaria, and somehow it also enhanced the ability to run long distances.