单目标跟踪mexLasso解决办法64位win7资源-CSDN文库

共2449个文件

png：716个

html：538个

m：385个

5星 · 超过95%的资源需积分: 50 169 浏览量 2014-05-13 11:01:47 上传评论 4 收藏 18.49MB RAR 举报

资源推荐

资源详情

资源评论

收起资源包目录

单目标跟踪mexLasso解决办法 64位win7 （2449个子文件）

sift.1 5KB

mser.1 4KB

vlfeat.7 3KB

aib 10KB

aib 9KB

aib 8KB

center.asv 4KB

initial_tracking.asv 3KB

drawbox.asv 2KB

drawtrackresult.asv 1KB

main.bib 36KB

vlfeat.bib 4KB

doc_spams.blg 1KB

covdet.c 74KB

sift.c 71KB

kmeans.c 52KB

hog.c 37KB

generic.c 36KB

imopv.c 34KB

mser.c 29KB

sift.c 26KB

dsift.c 24KB

vl_covdet.c 24KB

vl_svmpegasos.c 23KB

kdtree.c 23KB

mathop.c 23KB

aib.c 21KB

scalespace.c 21KB

vl_pegasos.c 20KB

vl_alldist2.c 17KB

homkermap.c 17KB

mser.c 17KB

pegasos.c 16KB

vl_sift.c 14KB

host.c 14KB

quickshift.c 14KB

slic.c 14KB

pgm.c 14KB

stringop.c 13KB

vl_hog.c 10KB

vl_imsmooth.c 10KB

vl_mser.c 9KB

vl_kmeans.c 9KB

getopt_long.c 9KB

rodrigues.c 9KB

lbp.c 9KB

vl_dsift.c 8KB

vl_ubcmatch.c 8KB

mathop_sse2.c 8KB

vl_maketrainingset.c 8KB

imopv_sse2.c 8KB

hikmeans.c 7KB

vl_aib.c 7KB

vl_siftdescriptor.c 7KB

vl_localmax.c 7KB

random.c 6KB

ikmeans.c 6KB

vl_ihashsum.c 6KB

vl_inthist.c 6KB

svmdataset.c 6KB

vl_hikmeanspush.c 6KB

vl_imwbackwardmx.c 6KB

vl_hikmeans.c 6KB

vl_alldist.c 6KB

array.c 6KB

vl_erfill.c 6KB

vl_homkermap.c 5KB

vl_kdtreequery.c 5KB

vl_twister.c 5KB

vl_kdtreebuild.c 5KB

vl_ihashfind.c 4KB

vl_ikmeans.c 4KB

vl_slic.c 4KB

vl_aibhist.c 4KB

test_heap-def.c 4KB

vl_imdisttf.c 4KB

vl_binsum.c 4KB

vl_cummax.c 4KB

test_stringop.c 4KB

vl_quickshift.c 4KB

vl_ikmeanspush.c 4KB

vl_sampleinthist.c 3KB

vl_imintegral.c 3KB

test_imopv.c 3KB

vl_lbp.c 2KB

test_getopt_long.c 2KB

vl_irodr.c 2KB

vl_rodr.c 2KB

test_threads.c 2KB

test_svd2.c 2KB

vl_tpsumx.c 2KB

vl_binsearch.c 2KB

vl_version.c 2KB

test_mathop.c 2KB

test_vec_comp.c 2KB

aib.c 1KB

test_mathop_abs.c 1KB

vl_simdctrl.c 985B

vl_getpid.c 852B

共 2449 条

SPAMS: a SPArse Modeling Software, v2.2

Julien Mairal

julien.mairal@m4x.org

March 23, 2012

1 Introduction

SPAMS (SPArse Modeling Software) is an open-source optimization toolbox under licence

GPLv3. It implements algorithms for solving various machine learning and signal processing

problems involving sparse regularizations.

The library is coded in C++, is compatible with Linux, Mac, and Windows 32bits and

64bits Operating Systems. It is interfaced with Matlab, but can be called from any C++

application. A R and Python interface has been developed by Jean-Paul Chieze.

It requires an implementation of BLAS and LAPACK for performing eﬃcient linear

algebra operations such as the one provided by matlab/R, atlas, netlib, or the one provided

by Intel (Math Kernel Library). It also exploits multi-core CPUs when this feature is

supported by the compiler, through OpenMP.

The current licence is GPLv3 available at http://www.gnu.org/licenses/gpl.html,

which limits its usage. For other usages (such as the use in proprietary softwares), please

contact the author.

Version 2.2 of SPAMS is divided into three main “toolboxes” and has a few additional

miscellaneous functions:

• The Dictionary learning and matrix factorization toolbox contains the online

learning technique of [18, 19] and its variants for solving various matrix factorization

problems:

– Dictionary Learning for sparse coding.

– Sparse principal component analysis.

– Non-negative matrix factorization.

– Non-negative sparse coding.

• The Sparse decomposition toolbox contains eﬃcient implementations of

– Orthogonal Matching Pursuit, (or Forward Selection) [28, 21].

– The LARS/homotopy algorithm [8] (variants for solving Lasso and Elastic-Net

problems).

– A weighted version of LARS.

– OMP and LARS when data come with a binary mask.

– A coordinate-descent algorithm for `

-decomposition problems [10, 9, 29].

– A greedy solver for simultaneous signal approximation as deﬁned in [27, 26]

(SOMP).

– A solver for simulatneous signal approximation with `

-regularization based

on block-coordinate descent.

– A homotopy method for the Fused-Lasso Signal Approximation as deﬁned in [9]

with the homotopy method presented in the appendix of [19].

– A tool for projecting eﬃciently onto a few convex sets inducing sparsity such

as the `

-ball using the method of [3, 16, 7], and Elastic-Net or Fused Lasso

constraint sets as proposed in the appendix of [19].

• The Proximal toolbox: An implementation of proximal methods (ISTA and FISTA [1])

for solving a large class of sparse approximation problems with diﬀerent combinations

of loss and regularizations. One of the main features of this toolbox is to provide a

robust stopping criterion based on duality gaps to control the quality of the opti-

mization, whenever possible. It also handles sparse feature matrices for large-scale

problems. The following regularizations are implemented:

– Tikhonov regularization (squared `

-norm).

– `

-norm, `

, `

∞

-norms.

– Elastic-Net [31].

– Fused Lasso [25].

– Tree-structured sum of `

-norms (see [13, 14]).

– Tree-structured sum of `

∞

-norms (see [13, 14]).

– General sum of `

∞

-norms (see [20]).

– Mixed `

-norms on matrices [30, 23].

– Mixed `

∞

-norms on matrices [30, 23].

– Mixed `

-norms on matrices plus `

[24].

– Mixed `

∞

-norms on matrices plus `

– group-lasso with `

or `

∞

-norms.

– group-lasso+`

– multi-task tree-structured sum of `

∞

-norms (see [20]).

– trace norm.

– `

pseudo-norm (only with ISTA)

– Tree-structured `

(only with ISTA)

– rank regularization for matrices (only with ISTA)

All of these regularization functions can be used with the following losses

– square loss.

– square loss with missing observations.

– logistic loss, weighted logistic loss.

– multi-class logistic

This toolbox can also enforce positivity constraints and handles sparse matrices.

• A few tools for performing linear algebra operations such as a conjugate gradient

algorithm, and manipulating sparse matrices.

The toolbox was written by Julien Mairal when he was at INRIA, with the collabora-

tion of Francis Bach (INRIA), Jean Ponce (Ecole Normale Supérieure), Guillermo Sapiro

(University of Minnesota) and Rodolphe Jenatton (INRIA).

A R and Python interface has been written by Jean-Paul Chieze (INRIA), and a few

contributors have helped us making compilation scripts for various platforms.

2 Installation

The toolbox used to come with pre-compiled binaries for various plateforms. Now the

sources are available and you will have to compile it yourself. Pre-compiled binaries will

still be available for Linux 64 bits plateforms.

The user has the choice of the BLAS library, but the Intel MKL is recommended for

the best performance. Note that the builtin blas library of Matlab is a version of the Intel

MKL (not the most recent one though).

The folder doc contains the documentation in pdf and html. build/ contains the binary

ﬁles, including the help for each command. test_release contains various matlab scripts

which call the diﬀerent functions of this toolbox.

The software package comes also with a script bash that has to be used to launch

matlab for Linux and/or Mac OS versions. The install procedure is described in a ﬁle

called HOW_TO_INSTALL.

3 Dictionary Learning and Matrix Factorization Toolbox

This is the section for dictionary learning and matrix factorization, corresponding to [18, 19].

3.1 Function mexTrainDL

This is the main function of the toolbox, implementing the learning algorithms of [19].

Given a training set x

, . . . ,. It aims at solving

min

D∈C

lim

n→+∞

i=1

min



− Dα

+ ψ(α

)



. (1)

ψ is a sparsity-inducing regularizer and C is a constraint set for the dictionary. As shown

in [19] and in the help ﬁle below, various combinations can be used for ψ and C for solving

diﬀerent matrix factorization problems. What is more, positivity constraints can be added

to α as well. The function admits several modes for choosing the optimization parame-

ters, using the parameter-free strategy proposed in [18], or using the parameters t

and

ρ presented in [19]. Note that for problems of a reasonable size, and when ψ is

the `

-norm, the function mexTrainDL_Memory can be faster but uses more

memory.

% Usage: [D [model]]=mexTrainDL(X,param[,model]);

% model is optional

% Name: mexTrainDL

% Description: mexTrainDL is an efficient implementation of the

% dictionary learning technique presented in

% "Online Learning for Matrix Factorization and Sparse Coding"

% by Julien Mairal, Francis Bach, Jean Ponce and Guillermo Sapiro

% arXiv:0908.0050

% "Online Dictionary Learning for Sparse Coding"

% by Julien Mairal, Francis Bach, Jean Ponce and Guillermo Sapiro

% ICML 2009.

% Note that if you use param.mode=1 or 2, if the training set has a

% reasonable size and you have enough memory on your computer, you

% should use mexTrainDL_Memory instead.

% It addresses the dictionary learning problems

% 1) if param.mode=0

% min_{D in C} (1/n) sum_{i=1}^n (1/2)||x_i-Dalpha_i||_2^2 s.t. ...

% ||alpha_i||_1 <= lambda

% 2) if param.mode=1

% min_{D in C} (1/n) sum_{i=1}^n ||alpha_i||_1 s.t. ...

% ||x_i-Dalpha_i||_2^2 <= lambda

% 3) if param.mode=2

% min_{D in C} (1/n) sum_{i=1}^n (1/2)||x_i-Dalpha_i||_2^2 + ...

% lambda||alpha_i||_1 + lambda_2||alpha_i||_2^2

% 4) if param.mode=3, the sparse coding is done with OMP

% min_{D in C} (1/n) sum_{i=1}^n (1/2)||x_i-Dalpha_i||_2^2 s.t. ...

% ||alpha_i||_0 <= lambda

% 5) if param.mode=4, the sparse coding is done with OMP

% min_{D in C} (1/n) sum_{i=1}^n ||alpha_i||_0 s.t. ...

% ||x_i-Dalpha_i||_2^2 <= lambda

%% C is a convex set verifying

% 1) if param.modeD=0

% C={ D in Real^{m x p} s.t. forall j, ||d_j||_2^2 <= 1 }

% 2) if param.modeD=1

% C={ D in Real^{m x p} s.t. forall j, ||d_j||_2^2 + ...

% gamma1||d_j||_1 <= 1 }

% 3) if param.modeD=2

% C={ D in Real^{m x p} s.t. forall j, ||d_j||_2^2 + ...

% gamma1||d_j||_1 + gamma2 FL(d_j) <= 1 }

% 4) if param.modeD=3

% C={ D in Real^{m x p} s.t. forall j, (1-gamma1)||d_j||_2^2 + ...

% gamma1||d_j||_1 <= 1 }

% Potentially, n can be very large with this algorithm.

% Inputs: X: double m x n matrix (input signals)

% m is the signal size

% n is the number of signals to decompose

% param: struct

% param.D: (optional) double m x p matrix (dictionary)

% p is the number of elements in the dictionary

% When D is not provided, the dictionary is initialized

% with random elements from the training set.

% param.lambda (parameter)

% param.lambda2 (optional, by default 0)

% param.iter (number of iterations). If a negative number is

% provided it will perform the computation during the

% corresponding number of seconds. For instance param.iter=-5

% learns the dictionary during 5 seconds.

% param.mode (optional, see above, by default 2)

% param.posAlpha (optional, adds positivity constraints on the

% coefficients, false by default, not compatible with

% param.mode =3,4)

% param.modeD (optional, see above, by default 0)

% param.posD (optional, adds positivity constraints on the

% dictionary, false by default, not compatible with

% param.modeD=2)

% param.gamma1 (optional parameter for param.modeD >= 1)

% param.gamma2 (optional parameter for param.modeD = 2)

% param.batchsize (optional, size of the minibatch, by default

% 512)

% param.modeParam (optimization mode).

% 1) if param.modeParam=0, the optimization uses the

% parameter free strategy of the ICML paper

% 2) if param.modeParam=1, the optimization uses the

% parameters rho as in arXiv:0908.0050

% 3) if param.modeParam=2, the optimization uses exponential

% decay weights with updates of the form

% A_{t} <- rho A_{t-1} + alpha_t alpha_t^T

% param.rho (optional) tuning parameter (see paper arXiv:0908.0050)

% param.clean (optional, true by default. prunes

% automatically the dictionary from unused elements).

% param.numThreads (optional, number of threads for exploiting

% multi-core / multi-cpus. By default, it takes the value -1,

% which automatically selects all the available CPUs/cores).

% Output:

% param.D: double m x p matrix (dictionary)

% Note: this function admits a few experimental usages, which have not

% been extensively tested:

% - single precision setting

% Author: Julien Mairal, 2009

3.2 Function mexTrainDL_Memory

Memory-consuming version of mexTrainDL. This function is well adapted to small/medium-

size problems: It requires storing all the coeﬃcients α and is therefore impractical for very

large datasets. However, in many situations, one can aﬀord this memory cost and it is

better to use this method, which is faster than mexTrainDL. Note that unlike mexTrainDL

this function does not allow warm-restart. _Memory.m

3.3 Function nmf

This function is an example on how to use the function mexTrainDL for the problem of non-

negative matrix factorization formulated in [15]. Note that mexTrainDL can be replaced

by mexTrainDL_Memory in this function for small or medium datasets.

% Usage: [U [,V]]=nmf(X,param);

% Name: nmf

% Description: mexTrainDL is an efficient implementation of the

% non-negative matrix factorization technique presented in

评论收藏

内容反馈

星空利明

2016-05-22

这个程序怎么用啊，我运行不起来

囧神柴囧
上传者
2016-07-22

我忘了，呵呵。我传的资源应该没问题的，亲测可用才传的。
zhoumengjiao12255710

2015-12-01

请问，您在找到源码后，如何编译成mexLasso.mexw64的呢？用来编译mexLasso.mexw64的源码具体用到了源码中的那些文件？最近我也在做这方面工作，麻烦您多多指教
redstar649

2015-11-19

感谢分享，很有帮助！
qss2012

2015-10-07

不错，可以下载，好用
daimao1412hao

2015-06-24

能用，请问目标跟踪代码对应的是哪篇论文？