Function dual_obj_grad

ot/smooth.py:260–301

r""" Compute objective value and gradients of dual objective. Parameters ---------- alpha: array, shape = len(a) beta: array, shape = len(b) Current iterate of dual potentials. a: array, shape = len(a) b: array, shape = len(b) Input histograms (should be

(alpha, beta, a, b, C, regul)

def dual_obj_grad(alpha, beta, a, b, C, regul):
    r"""
    Compute objective value and gradients of dual objective.

    Parameters
    ----------
    alpha: array, shape = len(a)
    beta: array, shape = len(b)
        Current iterate of dual potentials.
    a: array, shape = len(a)
    b: array, shape = len(b)
        Input histograms (should be non-negative and sum to 1).
    C: array, shape = (len(a), len(b))
        Ground cost matrix.
    regul: Regularization object
        Should implement a `delta_Omega(X)` method.

    Returns
    -------
    obj: float
        Objective value (higher is better).
    grad_alpha: array, shape = len(a)
        Gradient w.r.t. `alpha`.
    grad_beta: array, shape = len(b)
        Gradient w.r.t. `beta`.
    """
    obj = np.dot(alpha, a) + np.dot(beta, b)
    grad_alpha = a.copy()
    grad_beta = b.copy()

    # X[:, j] = alpha + beta[j] - C[:, j]
    X = alpha[:, np.newaxis] + beta - C

    # val.shape = len(b)
    # G.shape = len(a) x len(b)
    val, G = regul.delta_Omega(X)

    obj -= np.sum(val)
    grad_alpha -= G.sum(axis=1)
    grad_beta -= G.sum(axis=0)

    return obj, grad_alpha, grad_beta


def solve_dual(