L1 norm
The L1 norm of a vector is the sum of the absolute values of its coordinates.
On this page 7 sections
01The L1 norm adds the size of each coordinate#
The L1 norm adds the size of each coordinate. For , the horizontal contribution is and the vertical contribution is .
For , take each absolute value before adding. The negative coordinate contributes .
Add the coordinate magnitudes.
The subscript names this rule. Every entry contributes its absolute value, including entries after the first two.
For a vector with n coordinates.
Compute the L1 norm of .
Show answer and explanation
The contributions are .
02Do not add signed coordinates first#
Do not add signed coordinates first. For , the signed sum is , but the L1 norm is . Only the all-zero vector has L1 norm zero.
A zero entry adds nothing.
Three controls change by −2, 5 and −1. What is the total unsigned adjustment?
Show answer and explanation
Add the magnitudes of the three changes: .
03Check your understanding#
For , repair the L1 rule .
Show answer and explanation
Add .
Take each absolute value first, then add every contribution.
- Compute the L1 norm of a vector
Sources & further reading
- [1]Goodfellow, Bengio and Courville, Deep Learning (2016), 2.5 Norms ↗Deep Learning textbook · Book