TFiltersPy

Bayesian state estimation for Python – sklearn-compatible, scalable, modular.

TFiltersPy provides 5 production-ready Bayesian filters with a unified fit / predict / score API. Built on NumPy/SciPy with optional Dask parallelism for large-scale problems.

Part of the Ubunye AI Ecosystems.

Filters at a Glance

Filter	Best For	Linearity	Jacobians?	Scales To
KalmanFilter	GPS tracking, signal denoising	Linear	N/A	10K+ steps
ExtendedKalmanFilter	Radar, navigation	Nonlinear	Required	10K+ steps
UnscentedKalmanFilter	Highly nonlinear systems	Nonlinear	Not needed	10K+ steps
EnsembleKalmanFilter	Weather, ocean models	Nonlinear	Not needed	High-dim states
ParticleFilter	Robot localization, multimodal	Any	Not needed	Non-Gaussian

Quick Example

import numpy as np
from tfilterspy import KalmanFilter

F = np.eye(2)
H = np.eye(2)
Q = np.eye(2) * 0.1
R = np.eye(2) * 0.5

kf = KalmanFilter(F, H, Q, R, x0=np.zeros(2), P0=np.eye(2))
kf.fit(measurements)        # (n_steps, 2)
filtered = kf.predict()     # (n_steps, 2)
smoothed, _ = kf.smooth()   # RTS smoother

Key Features

5 Filters: KF, EKF, UKF, EnKF, Particle Filter – covering linear to fully nonlinear, Gaussian to arbitrary distributions
sklearn API: fit() / predict() / score() / get_params() / set_params()
RTS Smoothing: Backward pass for KF and EKF to refine estimates
Online Filtering: filter_step(z) for real-time streaming data
Forecasting: forecast(n_steps) to predict future states
Dask Parallelism: Ensemble and particle propagation on distributed clusters
Memory Efficient: store_covariances=False saves ~80% memory on long series
Numerically Stable: Joseph form covariance update, log-weight arithmetic

Contents: