API at a glance
Optimization
Unconstrained
|
BFGS solver. |
|
Gradient Descent solver. |
|
LBFGS solver. |
|
scipy.optimize.minimize wrapper |
|
Nonlinear conjugate gradient solver. |
Constrained
|
L-BFGS-B solver. |
|
Mirror descent solver. |
|
Projected gradient solver. |
|
scipy.optimize.minimize wrapper. |
Quadratic programming
|
Coordinate descent solver for box-constrained QPs. |
|
Operator Splitting Solver for Quadratic Programs. |
|
Wraps CVXPY's quadratic solver with implicit diff support. |
|
Quadratic programming with equality constraints only. |
|
OSQP solver for general quadratic programming. |
Non-smooth
|
Proximal gradient solver. |
|
Block coordinate solver. |
Stochastic
|
SGD with Armijo line search. |
|
Optax solver. |
|
SGD with Polyak step size. |
Loss functions
|
Binary logistic loss. |
|
Binary sparsemax loss. |
|
Binary hinge loss. |
|
Binary perceptron loss. |
Sparse plus function. |
|
Sparse sigmoid function. |
|
|
Huber loss. |
|
Multiclass logistic loss. |
|
Multiclass sparsemax loss. |
|
Multiclass hinge loss. |
Binary perceptron loss. |
Linear system solving
|
Solves |
|
Solves |
|
Solves |
|
Solves the normal equation |
|
Solves |
|
Solves |
|
Iterativement refinement algorithm. |
Nonlinear least squares
|
Gauss-Newton nonlinear least-squares solver. |
|
Levenberg-Marquardt nonlinear least-squares solver. |
Root finding
|
One-dimensional root finding using bisection. |
|
Limited-memory Broyden solver. |
|
scipy.optimize.root wrapper. |
Fixed point resolution
|
Fixed point iteration method. |
|
Anderson acceleration. |
|
Wrapper for accelerating JAXopt solvers. |
Implicit differentiation
|
Decorator for adding implicit differentiation to a root solver. |
Decorator for adding implicit differentiation to a fixed point solver. |
|
|
Jacobian-vector product of a root. |
|
Vector-Jacobian product of a root. |
Line search
|
Backtracking line search. |
|
Hager-Zhang line search. |
Perturbed optimizers
Transforms a function into a differentiable version with perturbations. |
|
Turns an argmax in a differentiable version of the max with perturbations. |
|
Transforms a function into a differentiable version with perturbations. |
|
Gumbel distribution. |
|
Normal distribution. |
Isotonic regression
|
Solves an isotonic regression problem using PAV. |
Tree utilities
|
Tree addition. |
|
Tree subtraction. |
|
Tree multiplication. |
|
Tree division. |
|
Compute scalar * tree_x. |
|
Compute tree_x + scalar * tree_y. |
|
Compute the inner product <tree_x, tree_y>. |
|
Compute sum(tree_x). |
|
Compute the l2 norm ||tree_x||. |
|
Creates an all-zero tree with the same structure as tree_x. |