Covariance Matrix To Feature Vector

Summary

• Plugin name : Covariance Matrix To Feature Vector
• Version : 0.1
• Author : Thibaut Monseigne
• Company : Inria
• Short description : Transforms a covariance matrix into a feature std::vector
• Documentation template generation date : Apr 20 2022

Description

Transforms a covariance matrix (size : NxN) into a feature std::vector (size : N(N+1)/2). The Setting Tangent Space define if the transformation into std::vector is done in the tanget space or if it is a squeeze of the upper triangular matrix

This box transforms the matrix into a vector for use in a classic classifier. Featurization for more details.

Inputs

1. Input Covariance Matrix

The covariance matrix on which the Feature Vector needs to be calculated.

• Type identifier : Streamed matrix (0x544a003e, 0x6dcba5f6)

Outputs

1. Output Feature Vector

Feature Vector generated.

• Type identifier : Feature vector (0x17341935, 0x152ff448)

Settings

1. Tangent Space

Method of calculating the Feature Vector :

True : The matrix is transposed into the tangent space.

False : The upper triangular matrix is used.

• Type identifier : Boolean (0x2cdb2f0b, 0x12f231ea)
• Default value : [ true ]

2. Filename to Reference Matrix (CSV, empty for Identity)

Link to the Reference Matrix CSV. A square matrix of size NxN with N the number of Features. The reference matrix is the same size as the input covariance matrices. The reference matrix is useful for calculating the feature vector on the tangent space.\nRemarks : If no reference an identity matrix is used.

• Type identifier : Filename (0x330306dd, 0x74a95f98)
• Default value : [ ${Player_ScenarioDirectory}/Mean.csv ]

3. Log Level

Log Level (None to see nothing)

• Type identifier : Log level (0xa88b3667, 0x0871638c)
• Default value : [ Information ]

Examples

Miscellaneous