optimal_features_SFS: Function performing optimal feature selection based on...

In jbferet/prospect: PROSPECT leaf radiative transfer model and inversion routines

Function performing optimal feature selection based on sequential forward feature slection

Description

Function performing optimal feature selection based on sequential forward feature slection

Usage

optimal_features_SFS(
  Refl = NULL,
  Tran = NULL,
  lambda,
  BiochTruth,
  Target,
  Parms2Estimate = "ALL",
  InitValues = data.frame(CHL = 40, CAR = 10, ANT = 0.1, BROWN = 0.01, EWT = 0.01, LMA =
    0.01, PROT = 0.001, CBC = 0.009, N = 1.5, alpha = 40),
  xlub = data.frame(CHL = c(1e-04, 150), CAR = c(1e-04, 25), ANT = c(0, 50), BROWN = c(0,
    4), EWT = c(1e-08, 0.1), LMA = c(1e-06, 0.06), PROT = c(1e-07, 0.006), CBC = c(1e-06,
    0.054), N = c(0.5, 4), alpha = c(10, 90)),
  SpecPROSPECT,
  spectral_domain,
  spectral_width,
  number_features,
  PROSPECT_version = "D",
  MeritFunction = "Merit_PROSPECT_RMSE",
  Est_alpha = FALSE,
  verbose = FALSE,
  nbCPU = 1,
  Continue = FALSE,
  AlreadyDone = NULL
)

Arguments

Refl	numeric. matrix of reflectances (n spectral bands x p samples)
Tran	numeric. matrix of transmittances (n spectral bands x p samples)
lambda	numeric. spectral bands corresponding to reflectance and transmittance measurements
BiochTruth	numeric. value of biophysical/biochemical parameter to estimate for each sample
Target	character. name of the parameter. Should be picked between "CHL", "CAR", "ANT", "BROWN", "EWT", "PROT", "CBC", "N"
Parms2Estimate	character vector. Parameters to estimate (can be 'ALL')
InitValues	data.frame. Default values of PROSPECT parameters. During optimization, they are used either as initialization values for parameters to estimate, or as fix values for other parameters. Parameters not compatible with PROSPECT_version are not taken into account.
xlub	data.frame. Boundaries of the parameters to estimate. The data.frame must have columns corresponding to Parms2Estimate first line being the lower boundaries and second line the upper boundaries.
SpecPROSPECT	list. Optical constants for a spectral domain broader than spectral_domain
spectral_domain	vector. defines minimum and maximum wavelengths of the spectral domain (in nm). Assumes 1nm spectral sampling
spectral_width	vector. width of individual spectral features (in nm)
number_features	vector. number of features to be identified
PROSPECT_version	character. Version of prospect model used for the inversion: 'D', 'PRO'
MeritFunction	character. name of the function to be used as merit function with given criterion to minimize (default = RMSE)
Est_alpha	boolean. should alpha be estimated or not?
verbose	boolean. set to TRUE to display sample number to be inverted
nbCPU	numeric. defines number of CPU for multithread processing
Continue	boolean. set to TRUE if the function has already been run for a lower value of number_features
AlreadyDone	list. contains output of function optimal_features_SFS previouly run with lower value of number_features

Value

list containing :

• SpectralFeatures = the Spectral Features identified as most relevant for estimation of Target (vector)

• SpectralFeatures_List = the Spectral Features identified as most relevant for estimation of Target (list)

• EstimatedParm = the estimated Target parameter for each additional spectral feature

• RMSE = the minimum RMSE corresponding to the estimation of Target for each asditional spectral feature

jbferet/prospect documentation built on Feb. 10, 2025, 9:35 a.m.