Why do neural networks need activation functions?medium
Answer
Activations add non-linearity so networks can model complex patterns.
Explanation
Without non-linear activations, stacked layers collapse into a linear transformation and lose expressive power.
Follow-upWhy is ReLU popular?