Loss functions

Loss functions guide your optimizer to take a step in the correct direction (direction of the global minima). For different tasks, different loss functions are used.

A complete list of all the available loss functions is available on the fastai docs → here.

The default loss function is CrossEntropy, depending on your problem you would need to change the loss function.

Here is a small list of tasks which require different loss functions,

1. Classification
2. Object Detection
3. Segmentation
4. Problems involving class-imbalance
5. Custom loss function (in case of hackathons, etc.)