Resources
- Tensorflow, harder to use when it is introduced, getting better
- PyTorch, very pythonic
- sk-learn, not a fan, but worth to mention
- xgboost, gbm
Topics
- data preparation, transformation, imputation, normalization
- model comparison, evaluation
- model interpretability
- productionization and reproducibility
- AutoML is a thing, worth look into
Recommendations
Write your own perceptron, loss function, optimizer, etc. for an plain vanilla network to get started.
And then get into imagine recognition, speech recognition, time series, RL with existing libraries.
And also be ready to read research papers, heck lotsa papers. :=)
Do an kaggle competition!