Deep Learning Quick Reference

Drawbacks

As before, a simpler model might do the job as well or better than a deep learning model. All else being equal, you should probably favor the simpler model. However, the cons of deep neural network complexity will often diminish as the number of classes increases. In order to accommodate many classes, many other models have to become significantly more complex in their implementation and some may even require you optimizing the multiclass strategy used for the model as a hyperparameter.