 #jsDisabledContent { display:none; } My Account | Register | Help Flag as Inappropriate This article will be permanently flagged as inappropriate and made unaccessible to everyone. Are you certain this article is inappropriate?          Excessive Violence          Sexual Content          Political / Social Email this Article Email Address:

# Non-negative least squares

Article Id: WHEBN0041740999
Reproduction Date:

 Title: Non-negative least squares Author: World Heritage Encyclopedia Language: English Subject: Collection: Publisher: World Heritage Encyclopedia Publication Date:

### Non-negative least squares

In mathematical optimization, the problem of non-negative least squares (NNLS) is a constrained version of the least squares problem where the coefficients are not allowed to become negative. That is, given a matrix A and a (column) vector of response variables y, the goal is to find

\operatorname*{arg\,min}_\mathbf{x} \|\mathbf{Ax} - \mathbf{y}\|_2 subject to x ≥ 0.

Here, x ≥ 0 means that each component of the vector x should be non-negative and ‖·‖₂ denotes the Euclidean norm.

Non-negative least squares problems turn up as subproblems in matrix decomposition, e.g. in algorithms for PARAFAC and non-negative matrix/tensor factorization. The latter can be considered a generalization of NNLS.

Another generalization of NNLS is bounded-variable least squares (BLVS), with simultaneous upper and lower bounds αx ≤ β.:291

## Contents

The NNLS problem is equivalent to a quadratic programming problem

\operatorname*{arg\,min}_\mathbf{x \ge 0} \frac{1}{2} \mathbf{x}^\mathsf{T} \mathbf{Q}\mathbf{x} + \mathbf{c}^\mathsf{T} \mathbf{x},

where Q = AA and c = Ay. This problem is convex as Q is positive semidefinite and the non-negativity constraints form a convex feasible set.

## Algorithms

The first widely used algorithm for solving this problem is an active set method published by Lawson and Hanson in their 1974 book Solving Least Squares Problems.:291 In pseudocode, this algorithm looks as follows:

• Inputs:
• a real-valued matrix A of dimension m × n
• a real-valued vector y of dimension m
• a real value t, the tolerance for the stopping criterion
• Initialize:
• Set P = ∅
• Set R = {1, ..., n}
• Set x to an all-zero vector of dimension n
• Set w = Aᵀ(yAx)
• Main loop: while R ≠ ∅ and max(w) > t,
• Let j be the index of max(w) in w
• Remove j from R
• Let AP be A restricted to the variables included in P
• Let SP = ((AP)ᵀ AP)-1 (AP)ᵀy
• While min(SP ≤ 0):
• Let α = min(xi/xi - si) for i in P where si ≤ 0
• Set x to x + α(s - x)
• Move to R all indices j in P such that xj = 0
• Set sP = ((AP)ᵀ AP)-1 (AP)ᵀy
• Set sR to zero
• Set x to s
• Set w to Aᵀ(yAx)

This algorithm takes a finite number of steps to reach a solution and smoothly improves its candidate solution as it goes (so it can find good approximate solutions when cut off at a reasonable number of iterations), but is very slow in practice, owing largely to the computation of the pseudoinverse ((Aᴾ)ᵀ Aᴾ)⁻¹. Variants of this algorithm are available in Matlab as the routine lsqnonneg and in SciPy as optimize.nnls.

Many improved algorithms have been suggested since 1974. Fast NNLS (FNNLS) is an optimized version of the Lawson—Hanson algorithm. Other algorithms include variants of Landweber's gradient descent method and coordinate-wise optimization based on the quadratic programming problem above.