Sparse Reconstruction with Compressed Sensing
Author: Ngoc Ly (SysEn 5800 Fall 2021)
Introduction
Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle x \in \mathbb{R}^N} often not really sparse but aproximently sparse
Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \Phi \in \mathbb{R}^{M \times N}} for iFailed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle M \ll N} s a Random Gaussian or Bernoulli matrix
Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle y \in \mathbb{R}^M} are the observed y samples
Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle e \in \mathbb{R}^M} noise vector Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \| e \|_2 \leq \eta} k e k 2 ≤ η
s.t.
How can we reconstruct x from Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle y = \Phi x + e}
The goal is to reconstruct x 0 ∈ R N Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle x \in \mathbb{R}^N} given Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle y} and Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \Phi}
Sensing matrix Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \Phi} must satisfy RIP i.e. Random Gaussian or Bernoulli ma-
traxies satsifies (cite)
let Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle [ N ] = \{ 1, \dots , N \} } be an index set
Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle [N]} enumerates the colums of Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \Phi} and Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle x}
Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \Phi} is an underdetermined systems with infinite solutions. Why Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle l_2} norm does
not work
What is compression is sunomunus with to the sparcity.
The problem formulation is to recover sparse data x 0 ∈ R N Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \mathbf{x} \in \mathbb{R}^N }
The support of Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \mathbf{x}} is Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle supp(\mathbf{x}) = \{i \in [N] : \mathbf{x}_i \neq 0 \}}
we say Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \mathbf{x}} is Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle k} sparse when Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle |supp(x)| \leq k}
We are interested in the smallest Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle supp(x)} , i.e. Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle min(supp(x))}
Restricted isometry property (RIP) Introduced by Candes, Tao
Random Gausian and Bernoulli satsifies RIP
Let Φ ∈ R M × N Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \Phi \in \mathbb{R}^{M \times N}} satsify RIP, Let Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle [N]} be an index set
Before we get into RIP lets talk about RIC
Resiricted Isometry Constant (RIC) is the smallest δ | s s.t. s ⊆ [ N ] Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \delta_{|s} \ s.t. \ s \subseteq [N]} that satsi-
fies the RIP condition
For Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle s} is a restrciton on Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \mathbf{x}} denoted by Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle x_{|s}} Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle x \in \mathbb{R}^N} to Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle s} k-sparse Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \mathbf{x}} s.t. RIP is
satisified the s = | Γ | i.e. s ⊆ [ N ] and Φ | s ⊆ Φ where the columns of Φ | s is
indexed by i ∈ S
�
x i , i f i ∈ S
( x | S ) i =
0 otherwise
RIP defined as
( 1 − δ s )k x k 22 ≤ k Φx k 22 ≤ ( 1 + δ s )k x k 22
3 Lemmas Page 267 Blumensath Davies IHT for CS
Lemma 1(Blumensath, Davis 2009 Iterative hard thresholding for compressed
sensing), For all index sets Γ and all Φ for which RIP holds with s = | Γ | that is
s = supp ( x )
1k Φ Γ T k 2 ≤
q
1 + δ | Γ | k y k 2
( 1 − δ | Γ | )k x Γ k 22 ≤ k Φ Γ T Φ Γ x Γ k 22 ≤ ( 1 + δ | Γ | )k x Γ k 22
and
k( I − Φ Γ T Φ Γ )k 2 ≤ δ | Γ | k x Γ k 2
SupposeΓ ∩ Λ = ∅
k Φ Γ T Φ Λ ) x Λ k 2 ≤ δ s k x Λ k 2
Lemma 2 (Needell Tropp, Prop 3.5 in CoSaMP: Iterative signal recovery
from incomplete and inaccurate √ samples)
If Φ satisfies RIP k Φx s k 2 ≤ 1 + δ s k x s k 2 , ∀ x s : k x s k 0 ≤ s, Then ∀ x
k Φx k 2 ≤
p
1 + δ s k x k 2 +
p
1 + δ s
k x k 1
sqrts
Lemma 3 (Needell Tropp, Prop 3.5 in CoSaMP: Iterative signal recovery
from incomplete and inaccurate samples)
Let x s be the best s-term approximation to x. Let x r = x − x s Let
y = Φx + e = Φx s + Φx r + e = Φx s + ẽ
If Φ statisfies RIP for sparcity s, then the norm of error ẽ is bounded by
k ẽ k 2 ≤
p
1 + δ s k x − x s k 2 +
p
1 + δ s
k x − x s k 1
√
+ k e k 2
s
∀ x
Theory
Gel’fend n-width
Errors E ( S, Φ, D )
Definition Mutual Coherence
LetA ∈ R M × N , themutualcoherenceµ A isde f inedby :
µ A =
|h a i , a j i|
k a i kk a j k
i 6 = j
We want a small µ A because it will be close ot the normal matrix, which
will satisfie RIP
Numerical Example
Basis Persuit
Iterative Hard Thresholding IHT
Solver
2