FitBasis - GTSAM Docs

Overview¶

FitBasis performs least-squares regression to fit a basis function to sample data. It builds a factor graph from samples and solves for basis parameters that best explain the data under a Gaussian noise model.

Authors: Frank Dellaert, et al. (see THANKS for the full author list)

See LICENSE for the license information

# Install GTSAM from pip if running in Google Colab
try:
    import google.colab
    %pip install --quiet gtsam-develop
except ImportError:
    pass  # Not in Colab

Key Functionality / API¶

FitBasis(sequence, model, N) constructs and solves the least-squares problem.
parameters() returns the fitted parameter vector.
NonlinearGraph(...) and LinearGraph(...) expose intermediate graphs.

C++ Usage Example¶

std::map<double, double> samples = {{0.0, 1.0}, {0.5, 0.2}, {1.0, -0.1}};
auto model = gtsam::noiseModel::Isotropic::Sigma(1, 0.1);
size_t N = 5;
gtsam::FitBasis<gtsam::Chebyshev2> fit(samples, model, N);
gtsam::Vector params = fit.parameters();

Python Example¶

import numpy as np
import gtsam

np.set_printoptions(precision=3, suppress=True)

sequence = {0.0: 1.0, 0.5: 0.2, 1.0: -0.1}
model = gtsam.noiseModel.Isotropic.Sigma(1, 0.1)
fit = gtsam.FitBasisFourierBasis(sequence, model, 3)
params = fit.parameters()
print("FitBasisFourierBasis parameters:", params)

FitBasisFourierBasis parameters: [ 2.242 -1.242 -1.986]

Source¶

FitBasis.h