Skip to main content

compression:apply svd

Version: 0.0.0

Description

Computes the coefficients (=U*Sigma) and VT components from SVD.

Inputs

InputNameExpected type(s)Description
Pin 0
Required
field_contaner_to_compressfields_containerFields container with data to be compressed
Pin 1scalar_intint32Number of vectors (r) to keep for the future reconstraction of the matrix A, ex.A[m,n] = coef[m,r] * VT[r,n], where coef = U * Sigma
Pin 2scalar_doubledoubleThreshold (precision) as a double (Default : 1e-7). If both pin1 and pin2 are provided, choose the min r-vectors
Pin 3booleanboolApply SVD on the initial data (Default : false), otherwise use reduced data (square matrix with the smallest dimensions).

Outputs

OutputNameExpected type(s)Description
Pin 0us_svdfields_containerThe output entity is a fields container (time dependant); it contains the product of two matrices, U and S, where A=U.S.Vt
Pin 1vt_svdfields_containerThe output entity is a field container (space dependant), containing the Vt, where A=U.S.Vt
Pin 2sigmafield, fields_containerThe output entity is a field (or a field container if input fc contains several labels, where field contains results per label), containing singular (S) values of the input data, where A=U.S.Vt

Configurations

NameExpected type(s)Default valueDescription
mutexboolfalseIf this option is set to true, the shared memory is prevented from being simultaneously accessed by multiple threads.
num_threadsint320
run_in_parallelbooltrueLoops are allowed to run in parallel if the value of this config is set to true.

Scripting

Category: compression

Plugin: N/A

Scripting name: apply_svd

Full name: compression.apply_svd

Internal name: svd_operator

License: any_dpf_supported_increments

Changelog

  • Version 0.0.0: Initial release.