47 ms·
> but you don't have to know all state of the arts techniques to be dangerous Often times you can be dangerous with just scikit learn, xgboost and Keras if you
by visarga 10y ago
> but you don't have to know all state of the arts techniques to be dangerous
Often times you can be dangerous with just scikit learn, xgboost and Keras if you can do data prepping, pipelines and stacking/ensembles. Besides wonderful frameworks such as TF, Theano and Torch there are also "zoos" of ready made models, word vectors and such that can be used out of the box.
- disgruntledphd2 10y agoDo read introduction to statistical learning though, as that book helps to give (some) understanding of what's actually happening in the models, which is important when (not if) things go horribly wrong.