# 4- 一般模式（General Patterns）

terms, separable objective and constraints 和 smooth objective terms.

## 4-1 近似算子（Proximity Operator）

x-update 为

[41] P. L. Combettes and J. C. Pesquet, “Proximal Splitting Methods in Signal Processing,” arXiv:0912.3522, 2009.

### 4-2-2 利用稀疏（Exploiting Sparsity）

F=P+ρATA$F = P + ρA^{T}A$，当 F 是稀疏时，
- if P and A are diagonal n × n matrices, then both the factor and solve costs are
O(n).
- If P and A are banded, then so is F.
- If F is banded with bandwidth k, the factorization cost is O(nk2)$O(nk^{2})$ and the back-solve cost is O(nk). In this case, the x-update can be carried out at a cost O(nk2)$O(nk^{2})$, plus the cost of forming F.

## 4-4 分解（Decomposition）

### 4-4-2 组件可分离（Component Separability）

x- 最小化可以通过 n 标量最小化 执行。

### 4-4-3 软阈值（Soft Thresholding）

In the language of §4.1, soft thresholding is the proximity operator of the L1 norm.

## 凸优化交替方向乘子法

• jzwong
• 2016-05-24 15:51
• 2033