Python for Finance Cookbook

There's more...

Some noteworthy features of fastai for tabular datasets include:

• Using callbacks while training neural networks. Callbacks are used for inserting some custom code/logic into the training loop at different times (begin_epoch, after_step, begin_fit, and so on). They are stored in the callbacks module and can be added as a list via the callback_fns argument of tabular_learner. Some of the interesting callbacks include ShowGraph (this plots the losses while training the model), OverSamplingCallback (used for oversampling data in case of class imbalance), and SaveModelCallback (used for resuming the training following an interruption).
• Using custom indices for training and validation sets. This feature comes in handy when we are ...