Wraps lmfit, which is a wrapper for scipy.optimize.
- Models that exist:
- Superconducting transition (resistance vs. temperature)
- Complex IQ transmission resonance (S21 vs. frequency)
- NS and NSN proximity effect models for calulating Tc vs film thickness
Each model must contain at least one method that implements a model and returns a residual. Ideally, each model will also implement a parameter guessing method that parses the data, makes some reasonable guesses and returns an lmfit.Paramters object. Signatures of functions must be as follows:
model_function(params, xvals, yvals=None, **kwargs)
#code goes here
return residual
params_function(*args, **kwargs)
params = lmfit.Parameters()
#code goes here to modify params
return params
It is highly encouraged to have your model function also have an option to return a model calculated at the xvals if no data is passed, and it is good practice to weight the residual with some uncertainty so that chi-squared values make sense.
It is also highly encouraged to include docstrings in both functions to describe the options.
Running an actual fit is as simple as calling the do_lmfit function from fit_tools.py and passing it a model function, a parameters object, the data, and any extra keywords that the model supports.