StackedML
Practice
Labs
Questions
Models
Pricing
Sign in
Questions
/
Optimization
/
Gradient Methods
/
Convex vs non-convex
← Previous
Next →
540.
Non-Convexity in Deep Learning
easy
Why are most deep learning loss functions non-convex?
A
They are compositions of nonlinear activation functions and weight matrices, which destroys global convexity
B
They optimize over discrete parameter spaces where convexity is undefined for non-continuous functions
C
They contain regularization terms that break the quadratic structure required for convexity
D
They use stochastic gradient estimates that introduce noise preventing the loss from being convex in expectation
Sign in to verify your answer
← Back to Questions