Course contentsShow
← Mathematics
Lesson 2113 of 3,12144. Linear Algebra — Matrices and SystemsFree lesson

Partial Pivoting Strategy

Choosing the largest available pivot to minimize numerical errors and improve computational stability.

Partial Pivoting Strategy

What you'll learn: How to choose the best pivot element during Gaussian elimination to keep numerical errors under control and improve computational accuracy.


Why Pivoting Matters

When performing Gaussian elimination, you use pivot positions to eliminate entries below them. But not all pivots are created equal! If your pivot is a very small number (close to zero), dividing by it creates huge multipliers that amplify rounding errors—especially on computers where numbers have limited precision.

Think of it like building a tower: if your foundation block is tiny and wobbly, every level above it magnifies that instability.

The Partial Pivoting Strategy

Core idea: Before eliminating entries in a column, scan downward from your current pivot position. Swap rows to bring the largest absolute value entry to the pivot position.

Example

Suppose you're working on column 2 and your current pivot row is row 2:

[  1   2   3 ]
[  0   0.01 5 ]  ← current pivot (tiny!)
[  0   4    7 ]

The element 0.01 is a terrible pivot. But 4 is much larger! Partial pivoting swaps rows 2 and 3:

[  1   2   3 ]
[  0   4   7 ]  ← now the pivot (stable!)
[  0   0.01 5 ]

Now elimination using 4 produces small, controlled multipliers (like 0.01/4 = 0.0025) instead of giant ones.

When to Use It

Apply partial pivoting at every step of forward elimination, right before you eliminate entries below the current pivot. It's a simple row swap that drastically improves numerical stability—critical for computer algorithms and large systems.


Key Takeaway: Partial pivoting selects the largest available pivot in the current column to minimize rounding errors and prevent numerical instability during Gaussian elimination.