Equadratures v10

equadratures v10 will be released early next year! Some of the planned features are listed below.

  • Bayesian polynomial models

    • Polynomial priors and posteriors for experimental vs. computational misfits

    • Polynomial priors for multi-fidelity / multi-level models

    • Polynomial coregional models for correlating similar polynomials

    • Bayesian polynomial quadrature strategies

  • Logistic regression in equadratures

  • Network polynomials

    • A network lasso-type framework with simultaneous clustering and regression

    • Sensitivity analysis for each cluster

    • Subspace-based analysis

    • Both L_1 and L_2 norm based cluster distance metrics

  • A minimum-code eqgo class focused on end-to-end results

  • Orthogonal polynomial expansions’ root finding for speeding up optimisations

  • Adaptive quadrature rules for polynomial least squares

  • Addition / subtraction / multiplication of polynomials

  • Basic spline fitting functionality

  • Enhanced syntax for streamlining the code

    • Moving away from
      param = Parameter(distribution='Normal', shape_parameter_A=2.5, shape_parameter_B=4.5, order=2)
      param = Normal(mean=2.5, variance=4.5, order=2)
      whilst maintaining backward compatibility.

    • Having correlations defined as part of a parameter list class (that extends the default list class).