LTPDA Toolbox™ contents Method: ao/svd_fit   Method: ao/tfe

Method ao/tdfit

  TDFIT fit a set of smodels to a set of input and output signals..
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
  DESCRIPTION: TDFIT fit a set of smodels to a set of input and output signals. 
  
 
  CALL:        b = tdfit(outputs, pl)
 
  INPUTS:      outputs  - the AOs representing the outputs of a system.
               pl       - parameter list (see below)
 
  OUTPUTs:     b  - a pest object containing the best-fit parameters,
                    goodness-of-fit reduced chi-squared, fit degree-of-freedom
                    covariance matrix and uncertainties. Additional
                    quantities, like the Information Matrix, are contained 
                    within the procinfo. The best-fit model can be evaluated
                    from pest\eval.
 
  Parameters Description
 
  EXAMPLES:
 
  % 1) Sine-wave stimulus of a simple system
  
    % Sine-wave data
    data = ao(plist('tsfcn', 'sin(2*pi*3*t) + 0.01*randn(size(t))', 'fs', 100, 'nsecs', 10));
    data.setName;
  
    % System filter
    pzm = pzmodel(1, 1, 10);
    f = miir(pzm);
  
    % Make output signal
    dataf = filter(data, f);
    dataf.setName;
  
    % fit model to output
    mdl = smodel('(a.*(b + 2*pi*i*f)) ./ (b*(a + 2*pi*i*f))');
    mdl.setParams({'a', 'b'}, {2*pi 10*2*pi});
    mdl.setXvar('f');
    params = tdfit(dataf, plist('inputs', data, 'models', mdl, 'P0', [1 1]));
  
    % Evaluate fit
    mdl.setValues(params);
    BestModel = fftfilt(data, mdl);
    BestModel.setName;
    iplot(data, dataf, BestModel, plist('linestyles', {'-', '-', '--'}))
  
    % recovered parameters (in Hz)
    params.y/2/pi
  
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Method Details
Access public
Defining Class ao
Sealed 0
Static 0

Parameter Description

Sets for this method …
Default

Default
no description
Key Default Value Options Description
tdfit
INPUTS [] none An array of input AOs, one per each experiment.
MODELS [] none An array of transfer function SMODELs.
PADRATIO 1 none PadRatio is defined as the ratio between the number of zero-pad points and the data length.
Define how much to zero-pad data after the signal.
Being tdfit a fft-based algorithm, no zero-padding might bias the estimation, therefore it is strongly suggested to do that.
WHFLTS [] none An array of FILTERBANKs containing the whitening filters per each output AO.
PNAMES {} [0x0] none A cell-array of parameter names to fit.
P0 [] none An array of starting guesses for the parameters.
LB [] none Lower bounds for the parameters.
This improves convergency. Mandatory for Monte Carlo.
UB [] none Upper bounds for the parameters.
This improves the convergency. Mandatory for Monte Carlo.
ALGORITHM 'fminsearch'
  • 'fminsearch'
  • 'fminunc'
  • 'fmincon'
  • 'patternsearch'
  • 'ga'
  • 'simulannealbnd'
 A string defining the fitting algorithm.
fminunc, fmincon require 'Optimization Toolbox' to be installed.
patternsearch, ga, simulannealbnd require 'Genetic Algorithm and Direct Search' to be installed.
OPTSET '' none An optimisation structure to pass to the fitting algorithm.
See fminsearch, fminunc, fmincon, optimset, for details.
See patternsearch, psoptimset, for details.
See ga, gaoptimset, for details.
See simulannealbnd, saoptimset, for details.
SYMDIFF 0
  • 0
  • 1
 Use symbolic derivatives or not. Only for gradient-based algorithm or for LinUnc option.
DIFFORDER 1
  • 1
  • 2
 Symbolic derivative order. Only for SymDiff option.
FITUNC 1
  • 1
  • 0
 Fit parameter uncertainties or not.
UNCMTD 'hessian'
  • 'hessian'
  • 'jacobian'
 Choose the uncertainties estimation method.
For multi-channel fitting hessian is mandatory.
LINUNC 1
  • 1
  • 0
 Force linear symbolic uncertainties.
GRADSEARCH 0
  • 0
  • 1
 Do a preliminary gradient-based search using the BFGS Quasi-Newton method.
MONTECARLO 0
  • 0
  • 1
 Do a Monte Carlo search in the parameter space.
Useful when dealing with high multiplicity of local minima. May be computer-expensive.
Note that, if used, P0 will be ignored. It also requires to define LB and UB.
NPOINTS 100000 none Set the number of points in the parameter space to be extracted.
NOPTIMS 10 none Set the number of optimizations to be performed after the Monte Carlo.
SINGLEINPUTSINGLEOUTPUT 0
  • 0
  • 1
 Specify whether the model should be considered as Single-Input/Single-Output model. This is for performance.

Example
plist('INPUTS', [[]], 'MODELS', [[]], 'PADRATIO', [1], 'WHFLTS', [[]], 'PNAMES', cell(0,0), 'P0', [[]], 'LB', [[]], 'UB', [[]], 'ALGORITHM', 'fminsearch', 'OPTSET', '', 'SYMDIFF', [false], 'DIFFORDER', [1], 'FITUNC', [true], 'UNCMTD', 'hessian', 'LINUNC', [true], 'GRADSEARCH', [false], 'MONTECARLO', [false], 'NPOINTS', [100000], 'NOPTIMS', [10], 'SINGLEINPUTSINGLEOUTPUT', [false])

back to top back to top

Some information of the method ao/tdfit are listed below:
Class name ao
Method name tdfit
Category Signal Processing
Package name ltpda
VCS Version 967b0eec0dece803a81af8ef54ad2f8c784b20b2
Min input args 1
Max input args -1
Min output args 1
Max output args -1
Can be used as modifier 0
Supported numeric types {'double'}



Method: ao/svd_fit  Method: ao/svd_fit   Method: ao/tfe Method: ao/tfe

©LTP Team