Result

class pylit.core.data_classes.Result(eps, residuals, mu, sigma, coefficients, S, exp_S, std_S, moments_S, forward_S, eps_S, max_eps_S)

Bases: object

Represents the output of the optimizer.

Parameters:

• eps (ndarray) – The values of the objective functions evaluated at the solutions coefficients.

• residuals (ndarray) – np.ndarray The residual norms from (1) evaluated at the solutions coefficients.

• mu (ndarray) – Discrete support points of the models in the frequency domain \(\omega\).

• sigma (ndarray) – Kernel widths of the models.

• coefficients (ndarray) – The final iterates of the optimization, representing the solutions.

• S (ndarray) –

  Evaluated model in the frequency domain at \(\omega\):

  \[S_i = \sum_j \text{coefficients}_j K_j(\omega_i).\]

  The scaling and detailed balance corrections are not included here, as they are applied automatically by linear_scaling_decorator() and detailed_balance_decorator().

• eps_S (ndarray)

• exp_S (ndarray) –

  np.ndarray Expected value (first moment) of S, computed as

  \[\rho_i = \frac{\max(S_i, 0)}{\sum_j \max(S_j, 0)}, \quad \langle S \rangle = \sum_i \omega_i \rho_i\]

• std_S (ndarray) –

  Standard deviation of S, computed as

  \[\sigma_S = \sqrt{\sum_i \rho_i (\omega_i - \langle S \rangle)^2}\]

• moments_S (ndarray) –

  np.ndarray Higher-order moments of S for indices \(\alpha = -1, 0, 1, \dots, 10\), computed as

  \[\mu_\alpha = \sum_i \omega_i^\alpha S_i\]

• forward_S (ndarray) –

  Forward-transformed S onto the original \(\tau\) grid, computed via a kernel or Laplace transform:

  \[\text{forward_S}_i = \sum_j \text{coefficients}_j \mathcal{L}(K)(\tau_i)\]

  The scaling and detailed balance corrections are not included here, as they are applied automatically by linear_scaling_decorator() and detailed_balance_decorator().

• eps_S – Pointwise reconstruction error between the forward-transformed model and the observed data.

• max_eps_S (ndarray) – Maximum absolute reconstruction error in eps_S.