- Adult is a 48 000 examples binary classification dataset
- Contains 14 features, 7 of which categoricals
- Very imbalanced dataset (75% - 25%)
Although Random Forests have been known to be very efficient on this dataset
I tried to find ways to reach the 90% accuracy bar, which seemed difficult
Very easy to adjust the code and do any architecture you want
Contains routine to load data from csv and re organize them in order to be used with pytorch's dataloader
Next interesting step would be to normalize non-categorical variables