Model fitting

HyperSpy can perform curve fitting of one-dimensional signals (spectra) and two-dimensional signals (images) in n-dimensional data sets. Models are defined by adding individual functions (components in HyperSpy’s terminology) to a BaseModel instance. Those individual components are then summed to create the final model function that can be fitted to the data, by adjusting the free parameters of the individual components.

Models can be created and fit to experimental data in both one and two dimensions i.e. spectra and images respectively. Most of the syntax is identical in either case. A one-dimensional model is created when a model is created for a Signal1D whereas a two- dimensional model is created for a Signal2D.

Note

Plotting and analytical gradient-based fitting methods are not yet implemented for the Model2D class.

• Creating a model
• Model components
  • Pre-defined model components
  • Define components from a mathematical expression
  • Define new components from a Python function
  • Define components from a fixed-pattern
• Adding components to the model
• Indexing the model
• Getting and setting parameter values and attributes
  • Getting parameter values
  • Setting Parameters free / not free
  • Setting twin parameters
  • Batch setting of parameter attributes
• Fitting the model to the data
  • Optimization algorithms
  • Loss functions
  • Using gradient information
  • Bounded optimization
  • Linear least squares
  • Optimization results
  • Goodness of fit
  • Visualizing the model
  • Setting the initial parameters
  • Exclude data from the fitting process
  • Fitting multidimensional datasets
  • Visualising the result of the fit
• Storing models
  • Saving and loading the result of the fit
  • Exporting the result of the fit
• Fitting big data
• Smart Adaptive Multi-dimensional Fitting (SAMFire)
  • The idea
  • Strategies
  • Local strategy family
  • Global strategy family
  • Seed points
  • Usage