Optimization Landscapes — Roadmap Step & Resources
Convexity, local minima and saddle points in the loss surfaces neural networks navigate
- Level: beginner
- Estimated time: 1 days
- Roadmap: Deep Learning & Neural Networks
Before this step
Interview questions
- Why are saddle points a bigger practical problem than local minima in high-dimensional deep learning?
Study resources
- Deep Learning Book: Optimization (Article) — The free official Deep Learning book's chapter on optimization for training deep networks.
- Visualizing the Loss Landscape of Neural Nets Explained (Video) — StatQuest's gradient descent explanation, foundational to understanding loss landscapes.
Part of
- Mathematics for Deep Learning — section
- Deep Learning & Neural Networks roadmap — the full learning path