A convenience library to fit truncated series (like polynomials) to data. The resulting calibration can be stored as a dictionary, wherein string expressions encode functions.
This library uses eval, which is not safe.
Generate noisy data:
from pylab import *
x = linspace( 0, 1, 50 )
y = 5*x + 0.5*sin( 2*pi*x ) + 0.5*rand( len(x) )
plot( x, y, '.' )Create a fit series with a first-order polynomial, and a sine function:
from LeastSquareCalib import CalibSeries
calib = CalibSeries().from_dict( [
dict( expression = 'x**i', coeffs = [0]*2 ),
dict( expression = 'np.sin( x*2*np.pi*( i + 1) )', coeffs = [0]*1 )
], safe = True )The initial values of the coefficients don't matter. Fit to data:
calib.fit_to_data( x, y )
print( calib.to_dict() )>>> [{'expression': 'x**i', 'coeffs': [0.28142438610775405, 5.011667743660484]}, {'expression': 'np.sin( x*2*np.pi*( i + 1) )', 'coeffs': [0.4983928797062334]}]
To see the result:
plot( x, calib.evaluate(x) )