学习 spams --- 帮助文档摘要(1)

来源:互联网 发布:苏州危机网络公关公司 编辑:程序博客网 时间:2024/05/16 11:47

spams(sparse modeling software):开源优化工具箱,稀疏估计,机器学习,信号处理,稀疏正则化。
(1)字典学习和矩阵分解工具箱
(1.1)字典学习,稀疏编码
(1.2)稀疏主成分分析
(1.3)非负矩阵分解
(1.4)非负稀疏编码
(1.5)结构化稀疏字典学习
(1.6)原型分析

(2) 稀疏分解工具箱
(2.1)正交匹配追踪(OMP)(前向选择)[35,27]
(2.2)the LARS/homotopy 算法(求解Lasso和Elastic-Net问题的变体)
(2.3)加权LARS
(2.4)OMP and LARS when data comes with a binary mask
(2.5)a coordinate-descent algorithm for L1-decomposition problems[12,10,36]
(2.6)a greedy solver for simultaneous signal approximation as defined in [34,33]
(2.7)a solver for simultaneous signal approximation with L1/L2-regularization based on block-coordinate descent
(2.8)a homotopy method for the Fused-Lasso Signal Approximation as defined in [10] with the homotopy method presented in the appendix of [21]
(2.9)a tool for projecting efficiently onto a few convex sets inducing sparsity such as the L1-ball using the method of [3,18,8];
and Elastic-Net or Fused Lasso constraint sets as proposed in the appendix of [21]
(2.10)an active-set algorithm for simplex decomposition problems[37]

(3)逼近工具箱:逼近方法(ISTA,FISTA[1]),基于duality gaps控制优化质量。
regularizations:
(3.1)Tikhonov regularization(平方L2-范数)
(3.2)L1,L2,L∞范数
(3.3)Elastic-Net
(3.4)Fused Lasso
(3.5)L2范数的树结构的和
(3.6)L∞范数的树结构的和
(3.7)L∞范数的general sum
(3.8)mixed L1/L2-norms on matrices
(3.9)mixed L1/L∞-norms on matrices
(3.10)mixed L1/L2-norms on matrices plus L1
(3.11)mixed L1/L∞-norms on matrices plus L1
(3.12)group-lasso with L2 or L∞-norms
(33.13)group-lasso + L1
(3.14)multi-task tree-structured sum of L∞-norms
(3.15)trace norm
(3.16)L0 pseudo-norm(ISTA)
(3.17)tree-structured L0(ISTA))
(3.18)rank regularization for matrices(ISTA)
(3.19)the path-coding penalties of [24]
losses:
(3.1)square loss
(3.2)square loss with missing observations
(3.3)logistic loss, weightd logistic loss
(3.4)multi-class logistic
others:
(3.1)non-negativity constraints, intercepts and sparse matrices
(3.2)stochastic and incremental proximal gradient solvers

(4)几个线性代数的工具:共轭梯度算法,稀疏矩阵、图

0 0
原创粉丝点击