Machine learning

HyperSpy provides easy access to several “machine learning” algorithms that can be useful when analysing multi-dimensional data. In particular, decomposition algorithms, such as principal component analysis (PCA), or blind source separation (BSS) algorithms, such as independent component analysis (ICA), are available through the methods described in this section.

Hint

HyperSpy will decompose a dataset, \(X\), into two new datasets: one with the dimension of the signal space known as factors (\(A\)), and the other with the dimension of the navigation space known as loadings (\(B\)), such that \(X = A B^T\).

For some of the algorithms listed below, the decomposition results in an approximation of the dataset, i.e. \(X \approx A B^T\).

• Decomposition
  • Available algorithms
  • Singular value decomposition (SVD)
  • Principal component analysis (PCA)
  • Poissonian noise
  • Maximum likelihood principal component analysis (MLPCA)
  • Robust principal component analysis (RPCA)
  • Non-negative matrix factorization (NMF)
  • Robust non-negative matrix factorization (RNMF)
  • Custom decomposition algorithms
• Blind Source Separation
  • Available algorithms
  • Orthomax
  • Independent component analysis (ICA)
  • Custom BSS algorithms
• Cluster analysis
  • Introduction
  • Nomenclature
  • Clustering functions HyperSpy
  • Pre-processing
  • Examples
• Visualizing results
  • Scree plots
  • Decomposition plots
  • Blind source separation plots
  • Clustering plots
• Export results
  • Obtain the results as BaseSignal instances
  • Save and load results