# Convex Iteration

(Difference between revisions)
 Revision as of 13:19, 24 November 2011 (edit)← Previous diff Current revision (21:18, 3 October 2016) (edit) (undo) (→constraining cardinality of signed variable) (5 intermediate revisions not shown.) Line 6: Line 6: ==constraining cardinality of signed variable== ==constraining cardinality of signed variable== - Excert from Chapter 4 [http://meboo.convexoptimization.com/Meboo.html Semidefinite Programming]: + Excerpt from Chapter 4 [http://meboo.convexoptimization.com/Meboo.html Semidefinite Programming]:
Consider a feasibility problem equivalent to the classical problem from linear algebra
Consider a feasibility problem equivalent to the classical problem from linear algebra $A_{}x_{\!}=_{\!}b$ ,  $A_{}x_{\!}=_{\!}b$ ,  - but with an upper bound $k$ on cardinality $\|x\|_0$ : + but with an upper bound $k$ on cardinality $||x||_0$ : -  for vector $b\!\in\!\mathcal{R}(A)$ +  for vector $b\!\in\mathcal{R}(A)$ - $\begin{array}{rl}\text{find}&_{}x\in_{}\mathbb{R}^n\\ + [itex]\begin{array}{rl}{\text find}&_{}x\in_{}\mathbb{R}^n\\ \mbox{subject to}&A_{}x=b\\ \mbox{subject to}&A_{}x=b\\ - &\|x\|_0\leq_{}k + &||x||_0\leq_{}k - \end{array}~~~~~~~~~~(1)$ + \end{array}[/itex]           (1) - where $\|x\|_{0\!}\leq_{_{}\!}k$ means vector $x$ has at most $k$ nonzero entries; + where $||x||_0\leq k$ means vector $x$ has at most $k$ nonzero entries; such a vector is presumed existent in the feasible set. such a vector is presumed existent in the feasible set.
Convex iteration works with a nonnegative variable; absolute value $|x|$ is therefore needed.
Convex iteration works with a nonnegative variable; absolute value $|x|$ is therefore needed. Line 24: Line 24: for $\varepsilon$ a relatively small positive constant, for $\varepsilon$ a relatively small positive constant, - $\begin{array}{cl}\text{minimize}_{x_{}\in_{_{}}\mathbb{R}^{^n},~t_{}\in_{_{}}\mathbb{R}^{^n}}&\langle t_{}\,,\,y+\varepsilon^{}\mathbf{1}\rangle\\ + [itex]\begin{array}{cl}\text{minimize}_{x_{}\in_{_{}}\mathbb{R}^n,~t_{}\in_{_{}}\mathbb{R}^n}&\langle t_{}\,,\,y+\varepsilon^{}\mathbf{1}\rangle\\ \mbox{subject to}&A_{}x=b\\ \mbox{subject to}&A_{}x=b\\ &-t\preceq x\preceq_{_{}}t &-t\preceq x\preceq_{_{}}t - \end{array}~~~~~~~~~~~~~~(2)$ + \end{array}[/itex]              (2) is iterated with is iterated with - $\begin{array}{cl}\text{minimize}_{y_{}\in_{_{}}\mathbb{R}^{^n}}&\langle t^*,\,y+\varepsilon^{}\mathbf{1}\rangle\\ + [itex]\begin{array}{cl}\text{minimize}_{y_{}\in_{_{}}\mathbb{R}^n}&\langle t^*,\,y+\varepsilon^{}\mathbf{1}\rangle\\ \mbox{subject to}&0\preceq y\preceq\mathbf{1}\\ \mbox{subject to}&0\preceq y\preceq\mathbf{1}\\ &y^T\mathbf{1}=n-_{}k &y^T\mathbf{1}=n-_{}k - \end{array}~~~~~~~~~~~~~~~~~~~(3)$ + \end{array}[/itex]                   (3) to find ''direction vector'' $y$. to find ''direction vector'' $y$. Line 45: Line 45: $\begin{array}{ccc} [itex]\begin{array}{ccc} - \begin{array}{cl}\text{minimize}_{x_{}\in_{_{}}\mathbb{R}^{^n},~t_{}\in_{_{}}\mathbb{R}^{^n}}&\langle t\,,_{}\mathbf{1}\rangle\\ + \begin{array}{cl}{\text minimize}_{x\in\mathbb{R}^n,~t\in\mathbb{R}^n}&\langle t,\mathbf{1}\rangle\\ - \mbox{subject to}&A_{}x=b\\ + \mbox{subject to}&Ax=b\\ - &-t\preceq x\preceq_{_{}}t + &-t\preceq x\preceq t \end{array} \end{array} - &\equiv&~ + &\equiv& - \begin{array}{cl}\text{minimize}_{x_{}\in_{_{}}\mathbb{R}^{^n}}&\|x\|_1\\ + \begin{array}{cl}{\text minimize}_{x\in\mathbb{R}^n}&||x||_1\\ - \mbox{subject to}&A_{}x=b + \mbox{subject to}&Ax=b \end{array} \end{array} - \end{array}~~~~~~~~~~(4)$ + \end{array}[/itex]          (4) Subsequent iterations of problem (2) engaging cardinality term $\langle t_{}\,,\,y\rangle$ Subsequent iterations of problem (2) engaging cardinality term $\langle t_{}\,,\,y\rangle$ - can be interpreted as corrections to this 1-norm problem leading to a 0-norm solution. + can be interpreted as corrections to this 1-norm problem leading to a $0$-norm solution.

Iteration (2) (3) always converges to a locally optimal solution by virtue of Iteration (2) (3) always converges to a locally optimal solution by virtue of Line 67: Line 67: ===equivalent=== ===equivalent=== Several simple equivalents to linear programs (2) (3) are easily devised, Several simple equivalents to linear programs (2) (3) are easily devised, - but their geometrical interpretation is not as apparent: ''e.g.'', equivalent in the limit $\,\varepsilon\!\rightarrow\!0^+\,$ + but their geometrical interpretation is not as apparent: ''e.g.'', equivalent in the limit $\,\varepsilon\!\rightarrow0^+\,$ - $\begin{array}{cl}\text{minimize}_{x_{}\in_{_{}}\mathbb{R}^{^n},~t_{}\in_{_{}}\mathbb{R}^{^n}}&\langle t\,,\,y\rangle\\ + [itex]\begin{array}{cl}{\text minimize}_{x_{}\in_{_{}}\mathbb{R}^n,~t_{}\in_{_{}}\mathbb{R}^n}&\langle t\,,\,y\rangle\\ \text{subject to}&_{}A_{}x=b\\ \text{subject to}&_{}A_{}x=b\\ &-t\preceq x\preceq_{_{}}t &-t\preceq x\preceq_{_{}}t \end{array}$ \end{array}[/itex] - [itex]\begin{array}{cl}\text{minimize}_{y_{}\in_{_{}}\mathbb{R}^{^n}}&\langle |x^*|\,,\,y\rangle\\ + [itex]\begin{array}{cl}{\text minimize}_{y_{}\in_{_{}}\mathbb{R}^n}&\langle |x^*|\,,\,y\rangle\\ \text{subject to}&0\preceq y\preceq\mathbf{1}\\ \text{subject to}&0\preceq y\preceq\mathbf{1}\\ &y^{\rm T}\mathbf{1}=n-_{}k &y^{\rm T}\mathbf{1}=n-_{}k

## Current revision

Convex iteration is method for constraining rank or cardinality in an otherwise convex optimization problem.
A rank or cardinality constraint is replaced by a linear regularization term added to the objective.
Then two convex problems are iterated until convergence where, ideally, solution to the original problem is found.

## constraining cardinality of signed variable

Excerpt from Chapter 4 Semidefinite Programming:
Consider a feasibility problem equivalent to the classical problem from linear algebra $LaTeX: A_{}x_{\!}=_{\!}b$ ,  but with an upper bound $LaTeX: k$ on cardinality $LaTeX: ||x||_0$ :  for vector $LaTeX: b\!\in\mathcal{R}(A)$

$LaTeX: \begin{array}{rl}{\text find}&_{}x\in_{}\mathbb{R}^n\\ \mbox{subject to}&A_{}x=b\\ &||x||_0\leq_{}k \end{array}$           (1)

where $LaTeX: ||x||_0\leq k$ means vector $LaTeX: x$ has at most $LaTeX: k$ nonzero entries; such a vector is presumed existent in the feasible set.
Convex iteration works with a nonnegative variable; absolute value $LaTeX: |x|$ is therefore needed.
We propose that nonconvex problem (1) can be equivalently written as a convex iteration: for $LaTeX: \varepsilon$ a relatively small positive constant,

$LaTeX: \begin{array}{cl}\text{minimize}_{x_{}\in_{_{}}\mathbb{R}^n,~t_{}\in_{_{}}\mathbb{R}^n}&\langle t_{}\,,\,y+\varepsilon^{}\mathbf{1}\rangle\\ \mbox{subject to}&A_{}x=b\\ &-t\preceq x\preceq_{_{}}t \end{array}$              (2)

is iterated with

$LaTeX: \begin{array}{cl}\text{minimize}_{y_{}\in_{_{}}\mathbb{R}^n}&\langle t^*,\,y+\varepsilon^{}\mathbf{1}\rangle\\ \mbox{subject to}&0\preceq y\preceq\mathbf{1}\\ &y^T\mathbf{1}=n-_{}k \end{array}$                   (3)

to find direction vector $LaTeX: y$. The cardinality constraint has been moved to the objective as a linear regularization.
The term $LaTeX: \langle t\,,_{_{}}\varepsilon^{}\mathbf{1}\rangle$ in (2) is necessary to determine absolute value $LaTeX: |x|_{\!}=_{}t^{*_{}}$ because vector $LaTeX: y$ can take zero values in its entries.

By initializing direction vector $LaTeX: y$ to $LaTeX: (1\!-_{}\!\varepsilon)\mathbf{1}_{}$, the first iteration of problem (2) is a 1-norm problem; i.e.,

$LaTeX: \begin{array}{ccc} \begin{array}{cl}{\text minimize}_{x\in\mathbb{R}^n,~t\in\mathbb{R}^n}&\langle t,\mathbf{1}\rangle\\ \mbox{subject to}&Ax=b\\ &-t\preceq x\preceq t \end{array} &\equiv& \begin{array}{cl}{\text minimize}_{x\in\mathbb{R}^n}&||x||_1\\ \mbox{subject to}&Ax=b \end{array} \end{array}$          (4)

Subsequent iterations of problem (2) engaging cardinality term $LaTeX: \langle t_{}\,,\,y\rangle$ can be interpreted as corrections to this 1-norm problem leading to a $LaTeX: 0$-norm solution.
Iteration (2) (3) always converges to a locally optimal solution by virtue of a monotonically nonincreasing objective sequence.
There can be no proof of global optimality, defined by an optimal objective equal to 0.
Local solutions are therefore detected by nonzero optimal objective.
Heuristics for reinitializing direction vector $LaTeX: y$ can lead to a globally optimal solution.

### equivalent

Several simple equivalents to linear programs (2) (3) are easily devised, but their geometrical interpretation is not as apparent: e.g., equivalent in the limit $LaTeX: \,\varepsilon\!\rightarrow0^+\,$

$LaTeX: \begin{array}{cl}{\text minimize}_{x_{}\in_{_{}}\mathbb{R}^n,~t_{}\in_{_{}}\mathbb{R}^n}&\langle t\,,\,y\rangle\\ \text{subject to}&_{}A_{}x=b\\ &-t\preceq x\preceq_{_{}}t \end{array}$

$LaTeX: \begin{array}{cl}{\text minimize}_{y_{}\in_{_{}}\mathbb{R}^n}&\langle |x^*|\,,\,y\rangle\\ \text{subject to}&0\preceq y\preceq\mathbf{1}\\ &y^{\rm T}\mathbf{1}=n-_{}k \end{array}$

For a coded numerical example, see Candes.m