StackedML
Practice
Labs
Questions
Models
Pricing
Sign in
Questions
/
Deep Learning
/
Foundations
/
Activation functions
← Previous
Next →
765.
Tanh vs Sigmoid for Hidden Layers
medium
The tanh activation function is often preferred over sigmoid for hidden layers. Why?
A
It is more computationally efficient than sigmoid since it avoids the exponential in the denominator
B
It produces sparser activations than sigmoid since negative inputs are mapped to exactly zero
C
It is zero-centered, producing outputs in (-1, 1) that keep gradients better balanced during backpropagation
D
It has a larger gradient than sigmoid at all input values, accelerating convergence for the same learning rate
Sign in to verify your answer
← Back to Questions