cotengra.experimental.hyper_scipy¶

Hyper optimization using scipy gradient-free optimizers.

Supported methods: differential_evolution, dual_annealing, direct, shgo. Since these optimizers use a callback-style objective, a background thread is used to invert the control flow into an ask/tell interface. Multiple workers are maintained per method to support parallel pre-dispatch (ask-ask-…-tell-tell-…).

Attributes¶

_OPTIMIZER_NAMES

Exceptions¶

_StopOptimization

Raised inside the objective to abort the scipy optimizer.

Classes¶

`ScipyAskTell`	Ask/tell wrapper around a scipy global optimizer.
`HyperScipySampler`	Per-method optimizer that wraps a pool of `ScipyAskTell`
`ScipyOptLib`	Hyper-optimization using scipy gradient-free optimizers with

Module Contents¶

cotengra.experimental.hyper_scipy._OPTIMIZER_NAMES¶

exception cotengra.experimental.hyper_scipy._StopOptimization¶

Bases: Exception

Raised inside the objective to abort the scipy optimizer.

class cotengra.experimental.hyper_scipy.ScipyAskTell(method, bounds, **kwargs)¶

Ask/tell wrapper around a scipy global optimizer.

The optimizer runs in a background thread. Each time it needs an objective evaluation it posts the candidate vector to _ask_q and blocks on _tell_q. The caller drives progress by alternating ask() / tell() calls from the main thread.

Parameters:

method (str) – One of the supported scipy optimizer names.
bounds (list[tuple[float, float]]) – Bounds for every raw dimension.
kwargs – Forwarded to the underlying scipy optimizer.

method¶

bounds¶

kwargs¶

_ask_q¶

_tell_q¶

_thread = None¶

_stop¶

done = False¶

_get_optimizer_fn()¶

_objective(x)¶

_run()¶

start()¶: Launch the optimizer in a background thread.

ask()¶

Block until the optimizer requests an evaluation.

Returns:: x – The candidate vector, or None if the optimizer finished.
Return type:: ndarray or None

tell(score)¶: Provide the objective value back to the optimizer.

stop()¶: Signal the background thread to stop and wait for it.

class cotengra.experimental.hyper_scipy.HyperScipySampler(space, method='differential_evolution', n_workers=1, exponential_param_power=None, **kwargs)¶

Per-method optimizer that wraps a pool of ScipyAskTell workers with the Param-based space mapping.

Each ScipyAskTell worker is strictly serial (ask-tell-ask-tell), so to support parallel pre-dispatch (multiple ask calls before any tell), we maintain a pool of workers. Each ask grabs a candidate from the next idle worker; each tell feeds the score back to the specific worker that produced that candidate.

Parameters:

space (dict[str, dict]) – The search space for one method.
method (str) – Which scipy optimizer to use.
n_workers (int, optional) – Number of concurrent ScipyAskTell threads to run.
exponential_param_power (float, optional) – Power for ParamFloatExp.
kwargs – Extra keyword arguments forwarded to the scipy optimizer.

params¶

_ndim¶

_method = 'differential_evolution'¶

_scipy_opts¶

_n_workers = 1¶

_trial_counter = 0¶

_trial_to_worker¶

_workers = []¶

_worker_idx = 0¶

_make_worker()¶: Create and start a fresh ScipyAskTell instance.

_next_worker()¶: Pick the next worker in round-robin order, restarting any that have finished.

ask()¶

Return (trial_number, params_dict).

Picks the next idle worker (round-robin), blocking until it has a candidate ready. If a worker has converged, it is automatically restarted.

tell(trial_number, score)¶: Report a score back to the specific worker that produced this trial.

stop()¶: Stop all background threads.

class cotengra.experimental.hyper_scipy.ScipyOptLib¶

Bases: cotengra.experimental.hyper.HyperOptLib

Hyper-optimization using scipy gradient-free optimizers with an LCB method selector.

setup(methods, space, optimizer=None, method='differential_evolution', method_exploration=1.0, method_temperature=1.0, exponential_param_power=None, **scipy_opts)¶

Initialize per-method scipy optimizers.

Parameters:

methods (list[str]) – The contraction methods to optimize over.
space (dict[str, dict[str, dict]]) – The search space.
optimizer (HyperOptimizer, optional) – The parent optimizer instance.
method (str, optional) – Which scipy global optimizer to use. One of 'differential_evolution', 'dual_annealing', 'direct', 'shgo'.
method_exploration (float, optional) – Exploration parameter for the LCB method selector.
method_temperature (float, optional) – Temperature parameter for the LCB method selector.
exponential_param_power (float, optional) – Power for ParamFloatExp.
scipy_opts – Extra keyword arguments forwarded to the scipy optimizer.

get_setting()¶

report_result(setting, trial, score)¶

cleanup()¶: Stop all background optimizer threads.