Kalmangrad is a python package that calculates smooth derivatives of non-uniformly sampled time series data using Bayesian filtering techniques to mitigate noise. Traditional numerical differentiation methods amplify noise, leading to inaccurate results. This package offers the ability to compute derivatives of any order, providing smoother and more accurate estimates in the presence of noise and non-uniform sampling. Users can easily integrate this package into existing projects due to its simple API and minimal dependencies. A unique feature is the automatic time step adjustment for non-uniformly sampled data. The package also includes functions for state transition, observation extraction, and Jacobian computation.
https://github.com/hugohadfield/kalmangrad