Machine Learning using landmarks

How to use the MACHINE LEARNING section of XYOM with landmarks data ?

  • Do not enter the RAW COORDINATES of LANDMARKS as inputs
  • Unknown and reference matrices must have the same landmarks (which also means the same number of columns) in exactly the same order
  • Please consider performing the following steps:
    • 1. Concatenate raw landmarks of unknown and reference individuals           <– MISCELLANEOUS (Working with data files, Concatenate, By Rows)
    • 2. Perform a GPA (Generalised Procrustes Analysis) on the single file containing the concatenated data, and save the resulting PCs (principal components) of orthogonal projections (ORP) <– CHARACTERIZATION
    • 3. Extract and save the PCs corresponding to the unknown specimens           <– MISCELLANEOUS (Working with data files, Extraction)
    • 4.Extract and save the PCs corresponding to the reference specimens           <– MISCELLANEOUS (Working with data files, Extraction)
    • 5. These 2 PC files are the correct input files for the Multilayer Perceptron. Enter the subdivision of the reference file, activate the SUBMIT button and see the report
    • 6. Do not keep these weights as valid weights for new identification analyses