Optimization

Optimization Functions

BOBE.optim.optimize_optax(fun, fun_args=(), fun_kwargs={}, num_params=1, bounds=None, x0=None, optimizer_options={'early_stop_patience': 25, 'lr': 0.001, 'name': 'adam'}, maxiter=200, n_restarts=1, verbose=False)

SEQUENTIAL OPTIMIZER: Minimize a function using JAX + optax, iterating through restarts with a Python for-loop.

Requires optax to be installed. Install with: pip install ‘BOBE[nn]’ or pip install optax

Return type:

Tuple[Array, float]

BOBE.optim.optimize_optax_vmap(fun, fun_args=(), fun_kwargs={}, num_params=1, bounds=None, x0=None, optimizer_options={'early_stop_patience': 25, 'lr': 0.001, 'name': 'adam'}, maxiter=200, n_restarts=1, verbose=False)

VECTORIZED OPTIMIZER: Minimize a function using JAX + optax, vectorizing over restarts with jax.vmap for parallel execution. Only to be used with EI.

Requires optax to be installed. Install with: pip install ‘BOBE[nn]’ or pip install optax

Return type:

Tuple[Array, float]

BOBE.optim.optimize_scipy(fun, fun_args=(), fun_kwargs={}, num_params=1, bounds=None, x0=None, optimizer_options={'ftol': 1e-06, 'gtol': 1e-06, 'method': 'L-BFGS-B'}, maxiter=200, n_restarts=4, verbose=False)

Standalone method to minimize a function using scipy.optimize.minimize.

Parameters:

• fun (Callable) – The objective function to minimize

• fun_args (Optional[Tuple]) – Additional arguments to pass to the function

• fun_kwargs (Optional[dict]) – Additional keyword arguments to pass to the function

• num_params (int) – Number of parameters.

• bounds (Union[List, Tuple, Array, None]) – Parameter bounds in shape (2, num_params).

• x0 (Optional[Array]) – Initial guess for the parameters, rescaled to unit space.

• optimizer_kwargs (Optional[dict]) – Additional keyword arguments to pass to the optimizer.

• maxiter (int) – Maximum number of iterations.

• n_restarts (int) – Number of restarts for the optimization.

• verbose (bool) – Whether to print progress messages.

Returns:

The best parameters found and the corresponding function value.

Return type:

Tuple[Array, float]