I have a very simple question, but I didn't figure out how to solve this.I have the function definition below:
function model = oasis(data, class_labels, parms)
% model = oasis(data, class_labels, parms)
%
% Code version 1.3 May 2011 Fixed random seed setting
% Code version 1.2 May 2011 added call to oasis_m.m
% Code version 1.1 May 2011 handle gaps in class_labels
%
% Input:
% -- data - Nxd sparse matrix (each instance being a ROW)
% -- class_labels - label of each data point (Nx1 integer vector)
% -- parms (do sym, do_psd, aggress etc.)
%
% Output:
% -- model.W - dxd matrix
% -- model.loss_steps - a binary vector: was there an update at
% each iterations
% -- modeo.parms, the actual parameters used in the run (inc. defaults)
%
% Parameters:
% -- aggress: The cutoff point on the size of the correction
% (default 0.1)
% -- rseed: The random seed for data point selection
% (default 1)
% -- do_sym: Whether to symmetrize the matrix every k steps
% (default 0)
% -- do_psd: Whether to PSD the matrix every k steps, including
% symmetrizing them (defalut 0)
% -- do_save: Whether to save the intermediate matrices. Note that
% saving is before symmetrizing and/or PSD in case they exist
% (default 0)
% -- save_path: In case do_save==1 a filename is needed, the
% format is save_path/part_k.mat
% -- num_steps - Number of total steps the algorithm will
% run (default 1M steps)
% -- save_every: Number of steps between each save point
% (default num_steps/10)
% -- sym_every: An integer multiple of "save_every",
% indicates the frequency of symmetrizing in case do_sym=1. The
% end step will also be symmetrized. (default 1)
% -- psd_every: An integer multiple of "save_every",
% indicates the frequency of projecting on PSD cone in case
% do_psd=1. The end step will also be PSD. (default 1)
% -- use_matlab: Use oasis_m.m instead of oasis_c.c
% This is provided in the case of compilation problems.
%
I want to use this function, but I don't figure how to set the parameters, or use the default values. What is the variable parms in this case, it is an object that keep all the other variables? Can I make something python like syntax where we put the name of the parameter plus value? For example:
model = oasis(data_example, labels_example, agress = 0.2)
Additionally, If I have understood correctly, I get two Objects in the Output, which is model and modeo, so I need to make this call to receive all contents this function returns?
[model,modeo] = oasis(data_example, labels_example, ?(parms)?)
From your function definition it seems like params is simply a placeholder for the parameters. Typically the parameters themselves are passed as pairs of inputs in the form:
model = oasis(data, class_labels, 'do_sym',do_symValue, 'do_psd', do_psdValue,...)
where do_symValue and do_psdValue are the values you want to pass as the respective parameters.
As for the functions return value, it returns a single struct with members W, loss_steps, and parms. I believe that what you thought as a second output (modelo) is simply a typo in the text - at least based on the function's definition.
From the documentation above, I don't know which one is the right, but there are two common ways for optional parameters in matlab.
parameter value pairs:
model = oasis(data, class_labels, 'do_sym',1,'do_psd',0)
structs:
params.do_sym=1
params.do_psd=0
model = oasis(data, class_labels, params)
Probably one of these two possibilities is right.
Related
I am using the input parser functionality a lot but now I am facing the following issue:
I am coding an object that computes the rate-of-change of time series, i.e. one input argument is a matrix of doubles (timeSeries). A second input parameter is the time period used to compute the rate-of-change (lag). lag can be a vector wherby different rate-of-changes will be computed. A third input vector is called weightVector. weightVector will be used to compute an average of all rate-of-changes and applies the appropriate weights to the respective rate-of-change results.
Now, I always like to define some default values when using input parser. I would like to define weightVector to be an equal weighted vector. However, the length of the default weightVector is dependent on the length of lag. For example, if lag = [1,2,3], then weightVectorshould equal [1/3, 1/3, 1/3]. How am I supposed to code a situation like this? My current code for the constructor looks like this:
function obj = roc(timeSeries, varargin)
%% Input parser
% 1. Create input parser instance
p = inputParser;
% 2. Default values for input arguments
default_lag = 1;
default_weightVector = 1/length(lag); % This line is causing
% problems as LAG isn't
% defined, yet.
% 3. Validation of input arguments
valid_lag = {'vector', 'nonempty', 'integer', 'positive'};
check_lag = #(x) validateattributes(x, {'numeric'}, valid_lag);
valid_weightVector = {'vector', 'nonempty'};
check_weightVector = #(x) validateattributes(x, {'numeric'}, ...
valid_weightVector);
% 4. Add input arguments to input scheme
p.addRequired('timeSeries');
p.addParamater('lag', default_lag, check_lag);
p.addParameter('weightVector', default_weightVector, check_weightVector);
% 5. Parse input arguments
parse(p, timeSeries, varargin{:});
% 6. Assign results to variables
lag = p.Results.lag;
weightVector = p.Results.weightVector;
%% Main code
end % Constructor
function obj = roc(timeSeries, varargin)
%% Input parser
% 1. Create input parser instance
p = inputParser;
% 2. Default values for input arguments
default_lag = 1;
default_weightVector = 1;
% 3. Validation of input arguments
valid_lag = {'vector', 'nonempty', 'integer', 'positive'};
check_lag = #(x) validateattributes(x, {'numeric'}, valid_lag);
% 4. Add input arguments to input scheme
p.addRequired('timeSeries');
p.addParameter('lag', default_lag, check_lag);
p.addParameter('weightVector', default_weightVector);
% 5. Parse input arguments
parse(p, timeSeries, varargin{:});
% 6. Assign results to variables
lag = p.Results.lag;
if check_weightVector(p.Results.weightVector, lag) == true
weightVector = p.Results.weightVector;
end
function vout = check_weightVector(weightVector, lag) % validation function
if length(lag) ~= length(weightVector)
error('lengthWeightVector:WrongNumberOfElements', 'The number of elements in "weightVector" must correspond to the number of elements in "lag"');
elseif sum(weightVector) ~= 1
error('sumWeightVector:SumNotEqualToOne', 'The sum of elements in "weightVector" must equal 1');
end
vout = true;
end
%% Main code
end
I am working on a simple System Object in Matlab/Simulink.
It looks like this :
classdef realtime_header_detectorSO < matlab.System & matlab.system.mixin.Propagates
% correlateHeader
%
% This template includes the minimum set of functions required
% to define a System object with discrete state.
properties
Header
%nrOfBitsInPreviousStep=0;
s=100;
d=zeros(1,s);
end
properties (DiscreteState)
end
properties (Access = private)
max_nr_of_packets=20;
max_packet_length_in_bytes=300;
current_packet=1;
% Pre-computed constants.
% h = commsrc.pn('GenPoly', [9 5 0],'NumBitsOut', 8,'InitialStates',ones(1,9));
% data=logical(zeros(max_nr_of_packets,max_packet_length_in_bytes*8));
end
methods (Access = protected)
function setupImpl(obj,u)
% Implement tasks that need to be performed only once,
% such as pre-computed constants.
end
function [maxCorr]= stepImpl(obj,u)
eml.extrinsic('num2str');
coder.extrinsic('sprintf');
% Implement algorithm. Calculate y as a function of
% input u and discrete states.
%y = size(u,1);
symbols=sign(u);
c=abs(conv(flipud(obj.Header),[symbols; symbols]));
maxCorr=max(c);
% maxCorr
if(maxCorr==36)
idx36=find(c(1:size(symbols,1))==36);
disp('header(s) detected at the following location(s) in bytes:');
disp(sprintf('%15.4f \n',idx36/8));
nrOfSymbols=size(symbols,1);
disp(['out of nr. of total symbols: ' num2str(nrOfSymbols)]);
disp('------');
% maxCorr
end
% y=obj.pBufferIdx;
end
function resetImpl(obj)
% Initialize discrete-state properties.
end
function varargout = isOutputFixedSizeImpl(~)
varargout = {true};
end
function varargout = getOutputSizeImpl(obj)
varargout = {[1 1]};
end
end
end
However when I compile/run it I get the following error:
The System object name 'realtime_header_detectorSO' specified in MATLAB System block 'freqScanningRT/Sync and
Find Header/detect header' is invalid.
Caused by:
Undefined function or variable 's'.
However (!) the following code compiles and runs just fine :
classdef realtime_header_detectorSO < matlab.System & matlab.system.mixin.Propagates
% correlateHeader
%
% This template includes the minimum set of functions required
% to define a System object with discrete state.
properties
Header
%nrOfBitsInPreviousStep=0;
s=100;
% d=zeros(1,s);
end
properties (DiscreteState)
end
properties (Access = private)
max_nr_of_packets=20;
max_packet_length_in_bytes=300;
current_packet=1;
% Pre-computed constants.
% h = commsrc.pn('GenPoly', [9 5 0],'NumBitsOut', 8,'InitialStates',ones(1,9));
% data=logical(zeros(max_nr_of_packets,max_packet_length_in_bytes*8));
end
methods (Access = protected)
function setupImpl(obj,u)
% Implement tasks that need to be performed only once,
% such as pre-computed constants.
end
function [maxCorr]= stepImpl(obj,u)
eml.extrinsic('num2str');
coder.extrinsic('sprintf');
% Implement algorithm. Calculate y as a function of
% input u and discrete states.
%y = size(u,1);
disp(obj.s);
symbols=sign(u);
c=abs(conv(flipud(obj.Header),[symbols; symbols]));
maxCorr=max(c);
% maxCorr
if(maxCorr==36)
idx36=find(c(1:size(symbols,1))==36);
disp('header(s) detected at the following location(s) in bytes:');
disp(sprintf('%15.4f \n',idx36/8));
nrOfSymbols=size(symbols,1);
disp(['out of nr. of total symbols: ' num2str(nrOfSymbols)]);
disp('------');
% maxCorr
end
% y=obj.pBufferIdx;
end
function resetImpl(obj)
% Initialize discrete-state properties.
end
function varargout = isOutputFixedSizeImpl(~)
varargout = {true};
end
function varargout = getOutputSizeImpl(obj)
varargout = {[1 1]};
end
end
end
So I can access s in the stepImpl(obj,u) as obj.s but I cannot access s inside the properties block, where it is defined !
Now this is confusing.
Is there way to access s inside the properties block ?
The problem is that I have to use the properties block because if I try this :
function setupImpl(obj,u)
% Implement tasks that need to be performed only once,
% such as pre-computed constants.
d=zeros(1,obj.s);
end
then I get :
Error due to multiple causes.
Caused by:
Problem creating simulation target MEX-file for model 'freqScanningRT'.
Simulink detected an error
'Computed maximum size is not bounded.
Static memory allocation requires all sizes to be bounded.
The computed size is [1 x :?].'.
The error occurred for MATLAB System block 'freqScanningRT/Sync and Find Header/detect header'. See line
34, column 15 in file
'path/realtime_header_detectorSO.m'.
The error was detected during code generation phase.
Start code generation report.
To prevent this error, use one of the following:
* Modify the System object to avoid code that does not support code generation.
* Change 'Simulate using' parameter to 'Interpreted Execution'.
Any idea how to refer to variables in the properties blocks ?
There must be a way to do this.
You should be able to use
properties
s = 100;
d = zeros(1,100);
end
right? If you already have the 100 as a default for s, you should also be able to provide this as part of the default for d.
I'm guessing that you're trying to avoid doing that because you feel uncomfortable repeating the "100". But I'd also guess that really, "100" is some sort of magic number that your system depends on; so really, you should try to pull it out as a constant in any case, whether it's repeated or not.
So in that case, you might improve things with
properties (Constant)
MYCONSTANT = 100;
end
properties
% Reference Constant class properties with the class name
s = realtime_header_detectorSO.MYCONSTANT;
d = zeros(1, realtime_header_detectorSO.MYCONSTANT);
end
You're not going to be able to do what you're originally trying to do - it's not possible to reference the name of one property within the property definition block when defining another property (even though you can perfectly well reference it within a method). I guess I understand why you find that confusing - but to clear up your confusion, note that you have no guarantee over the order in which MATLAB instantiates default values for properties. If it tried to create d before it had created s, it would obviously fail.
You can avoid this problem all together by initializing the properties in the constructor.
I have Matlab script that I am trying to compile as an executable that will run on a headless server as part of a large batch process. The script calls functions and classes written by several non-programmers (scientists) over the span of more than a decade, and I am having a difficult time getting it to compile.
The script will run in a Matlab instance, but MCC gives me an error that I do not quite understand:
Compiler version: 5.1 (R2014a)
Dependency analysis by REQUIREMENTS.
Error using matlab.depfun.internal.MatlabSymbol/proxy (line 405)
Internal1 Error: Could not determine class of method
"/home/me/MyCode/#tsd/Data.m". Number of classes checked:
17.
#tsd/tsd.m looks like this:
function tsa = tsd(t, Data)
% tsd Creates tsd object (tsd = class of timestamped arrays of data)
%
% tsa = tsd(t, Data)
%
% INPUTS:
% t - list of timestamps - must be sequential, but don't have to be continuous
% Data - data, possibly an array. If data is n-dimensional, then time should be the FIRST axis.
% OUTPUTS:
% tsa - a tsd object
%
% Completely compatible with ctsd.
%
% Methods
% tsd/Range - Timestamps used
% tsd/Data - Returns the data component
% tsd/DT - Returns the DT value (mean diff(timestamps))
% tsd/StartTime - First timestamp
% tsd/EndTime - Last timestamp
% tsd/Restrict - Keep data within a certain range
% tsd/CheckTS - Makes sure that a set of tsd & ctsd objects have identical start and end times
% tsd/cat - Concatenate ctsd and tsd objects
% tsd/Mask - Make all non-mask values NaN
%
% ADR 1998, version L4.0, last modified '98 by ADR
% RELEASED as part of MClust 2.0
% See standard disclaimer in ../Contents.m
if nargin == 0
tsa.t = NaN;
tsa.data = NaN;
tsa = class(tsa, 'tsd');
return
end
if nargin < 2
error('tsd constructor must be called with T, Data');
end
tsa.t = t;
tsa.data = Data;
tsa = class(tsa, 'tsd');
and the Data.m file:
function v = Data(tsa, ix)
% tsd/Data Retrieves data from tsd
%
% d = Data(tsa)
% d = Data(tsa, ix)
%
% INPUTS:
% tsa - tsd object
% ix - alignment list (timestamps)
% OUTPUTS:
% v - the tsa.Data
%
% if called with alignment list, returns those tsa.Data(ix)
% if called without, returns complete tsa.Data
%
% ADR 1998, version L4.1, last modified '98 by ADR
% RELEASED as part of MClust 2.0
% See standard disclaimer in ../Contents.m
switch nargin
case 2
f = findAlignment(tsa, ix);
v = SelectAlongFirstDimension(tsa.data,f);
case 1
v = tsa.data;
otherwise
error('Unknown number of input arguments');
end
So the script calling the tsd function runs just find in the Matlab session, but the compiler throws the error described above. This is my first time working with Matlab and I am totally stumped. There is another class with a method named "Data", but that shouldn't cause this problem, could it?
I was having the exact same problem and was able to come up with a workaround.
I was facing this problem when I was compiling the MATLAB library using the command
mcc -W cpplib:LibName -T link:lib main.m -a './' -d 'dll_destintaion_dir'
LibName: DLL name
main.m: MATLAB entry point
dll_destintaion_dir: destination where DLL created will be stored
The error was gone when I compiled using this command instead
mcc -W cpplib:LibName -T link:lib main.m -a 'full_path_of_project_folder/*'
-d 'dll_destintaion_dir'
Note: This method does not add folders recursively to the archive. So subfolders have to be added explicitly using another -a flag in the command shown above.
I'm new to Matlab.
I'm trying to apply PCA function(URL listed below)into my palm print recognition program to generate the eigenpalms. My palm print grey scale images dimension are 450*400.
Before using it, I was trying to study these codes and add some codes to save the eigenvector as .mat file. Some of the %comments added by me for my self understanding.
After a few days of studying, I still unable to get the answers.
I decided to ask for helps.I have a few questions to ask regarding this PCA.m.
PCA.m
What is the input of the "options" should be? of "PCA(data,details,options)"
(is it an integer for reduced dimension? I was trying to figure out where is the "options" value passing, but still unable to get the ans. The msgbox of "h & h2", is to check the codes run until where. I was trying to use integer of 10, but the PCA.m processed dimension are 400*400.)
The "eigvector" that I save as ".mat" file is ready to perform Euclidean distance classifier with other eigenvector? (I'm thinking that eigvector is equal to eigenpalm, like in face recognition, the eigen faces. I was trying to convert the eigenvector matrix back to image, but the image after PCA process is in Black and many dots on it)
mySVD.m
In this function, there are two values that can be changed, which are MAX_MATRIX_SIZE set by 1600 and EIGVECTOR_RATIO set by 0.1%. May I know these values will affect the results? ( I was trying to play around with the values, but I cant see the different. My palm print image dimension is set by 450*400, so the Max_matrix_size should set at 180,000?)
** I hope you guys able to understand what I'm asking, please help, Thanks guys (=
Original Version : http://www.cad.zju.edu.cn/home/dengcai/Data/code/PCA.m
mySVD: http://www.cad.zju.edu.cn/home/dengcai/Data/code/mySVD.m
% Edited Version by me
function [eigvector, eigvalue] = PCA(data,details,options)
%PCA Principal Component Analysis
%
% Usage:
% [eigvector, eigvalue] = PCA(data, options)
% [eigvector, eigvalue] = PCA(data)
%
% Input:
% data - Data matrix. Each row vector of fea is a data point.
% fea = finite element analysis ?????
% options.ReducedDim - The dimensionality of the reduced subspace. If 0,
% all the dimensions will be kept.
% Default is 0.
%
% Output:
% eigvector - Each column is an embedding function, for a new
% data point (row vector) x, y = x*eigvector
% will be the embedding result of x.
% eigvalue - The sorted eigvalue of PCA eigen-problem.
%
% Examples:
% fea = rand(7,10);
% options=[]; %store an empty matrix in options
% options.ReducedDim=4;
% [eigvector,eigvalue] = PCA(fea,4);
% Y = fea*eigvector;
%
% version 3.0 --Dec/2011
% version 2.2 --Feb/2009
% version 2.1 --June/2007
% version 2.0 --May/2007
% version 1.1 --Feb/2006
% version 1.0 --April/2004
%
% Written by Deng Cai (dengcai AT gmail.com)
%
if (~exist('options','var'))
%A = exist('name','kind')
% var = Checks only for variables.
%http://www.mathworks.com/help/matlab/matlab_prog/symbol-reference.html#bsv2dx9-1
%The tilde "~" character is used in comparing arrays for unequal values,
%finding the logical NOT of an array,
%and as a placeholder for an input or output argument you want to omit from a function call.
options = [];
end
h2 = msgbox('not yet');
ReducedDim = 0;
if isfield(options,'ReducedDim')
%tf = isfield(S, 'fieldname')
h2 = msgbox('checked');
ReducedDim = options.ReducedDim;
end
[nSmp,nFea] = size(data);
if (ReducedDim > nFea) || (ReducedDim <=0)
ReducedDim = nFea;
end
if issparse(data)
data = full(data);
end
sampleMean = mean(data,1);
data = (data - repmat(sampleMean,nSmp,1));
[eigvector, eigvalue] = mySVD(data',ReducedDim);
eigvalue = full(diag(eigvalue)).^2;
if isfield(options,'PCARatio')
sumEig = sum(eigvalue);
sumEig = sumEig*options.PCARatio;
sumNow = 0;
for idx = 1:length(eigvalue)
sumNow = sumNow + eigvalue(idx);
if sumNow >= sumEig
break;
end
end
eigvector = eigvector(:,1:idx);
end
%dt get from C# program, user ID and name
evFolder = 'ev\';
userIDName = details; %get ID and Name
userIDNameWE = strcat(userIDName,'\');%get ID and Name with extension
filePath = fullfile('C:\Users\***\Desktop\Data Collection\');
userIDNameFolder = strcat(filePath,userIDNameWE); %ID and Name folder
userIDNameEVFolder = strcat(userIDNameFolder,evFolder);%EV folder in ID and Name Folder
userIDNameEVFile = strcat(userIDNameEVFolder,userIDName); % EV file with ID and Name
if ~exist(userIDNameEVFolder, 'dir')
mkdir(userIDNameEVFolder);
end
newFile = strcat(userIDNameEVFile,'_1');
searchMat = strcat(newFile,'.mat');
if exist(searchMat, 'file')
filePattern = strcat(userIDNameEVFile,'_');
D = dir([userIDNameEVFolder, '*.mat']);
Num = length(D(not([D.isdir])))
Num=Num+1;
fileName = [filePattern,num2str(Num)];
save(fileName,'eigvector');
else
newFile = strcat(userIDNameEVFile,'_1');
save(newFile,'eigvector');
end
You pass options in a structure, for instance:
options.ReducedDim = 2;
or
options.PCARatio =0.4;
The option ReducedDim selects the number of dimensions you want to use to represent the final projection of the original matrix. For instance if you pick option.ReducedDim = 2 you use only the two eigenvectors with largest eigenvalues (the two principal components) to represent your data (in effect the PCA will return the two eigenvectors with largest eigenvalues).
PCARatio instead allows you to pick the number of dimensions as the first eigenvectors with largest eigenvalues that account for fraction PCARatio of the total sum of eigenvalues.
In mySVD.m, I would not increase the default values unless you expect more than 1600 eigenvectors to be necessary to describe your dataset. I think you can safely leave the default values.
Since it is not possible to have a script and a function definition in the same file , I thought to echo the function which I want to attach in the script such that I get a script with function code and then some usage with this function .
For example -
func1.m
function [result] = func1(x)
result=sqrt(x) ;
end
script1.m
echo(func1.m) ;
display(func1(9))
Desire output for script1.m
function [result] = func1(x)
result=sqrt(x) ;
end
display(func1(9))
3
Have you any idea for that ?
Since a convoluted solution was already proposed, why not stating the obvious?
Matlab has a built-in command that does exactly what you want. It is called type:
>> type('mean')
will give you this:
function y = mean(x,dim)
%MEAN Average or mean value.
% For vectors, MEAN(X) is the mean value of the elements in X. For
% matrices, MEAN(X) is a row vector containing the mean value of
% each column. For N-D arrays, MEAN(X) is the mean value of the
% elements along the first non-singleton dimension of X.
%
% MEAN(X,DIM) takes the mean along the dimension DIM of X.
%
% Example: If X = [0 1 2
% 3 4 5]
%
% then mean(X,1) is [1.5 2.5 3.5] and mean(X,2) is [1
% 4]
%
% Class support for input X:
% float: double, single
%
% See also MEDIAN, STD, MIN, MAX, VAR, COV, MODE.
% Copyright 1984-2005 The MathWorks, Inc.
% $Revision: 5.17.4.3 $ $Date: 2005/05/31 16:30:46 $
if nargin==1,
% Determine which dimension SUM will use
dim = min(find(size(x)~=1));
if isempty(dim), dim = 1; end
y = sum(x)/size(x,dim);
else
y = sum(x,dim)/size(x,dim);
end
You could use this:
function echo(mfile)
filename=which(mfile);
if isempty(filename)
fprintf('Invalid input - check you are inputting a string.');
return;
end
fid=fopen(filename,'r');
if (fid<0)
fprintf('Couldn''t open file.');
end
file=fread(fid,Inf);
fclose(fid);
fprintf('%s',file);
end
This will open a file, read it, and print it. Note that you need to provide the input as a string, i.e. with single quotes around it, and need to have '.m' at the end:
echo('fread.m')
Not
echo(fread.m) % This won't work
echo('fread') % This won't work
Just for completeness, there's also dbtype which prepends line numbers.