Golden Rules: Evaluation

<aside> 💡

In the end , what matters is your chosen metrics, not the loss. The loss is just a function we’ve give the computer to help us to optimize.o

</aside>

Machines learn by means of a loss function. It’s a method of evaluating how well specific algorithm models the given data. If predictions deviates too much from actual results, loss function would cough up a very large number.

Thinks like Cross Entropy , binary-cross-entropy and SoftMax & ArgMax , Mean Absolute Error (MAE), Root mean Squared Error (RMSE),… are all loss functions. See overview here.

A great explanation, given by fastai, can be found here. It covers the math and explains how and why to apply it.

SoftMax & ArgMax

Cross Entropy