Baysian Search

OptunaTPESearch

Solver name is “OptunaTPE”.

class flopt.solvers.optuna_searches.OptunaTPESearch

Tree-structured Parzen Estimator (TPE) Sampling Search of Optuna. https://optuna.readthedocs.io/en/latest/reference/samplers/generated/optuna.samplers.TPESampler.html This solver can solve constrainted problems, but it cannot handle equal constraints well.

Parameters:

• name (str) – name

• consider_prior (bool) –

• consider_magic_clip (bool) –

• consider_endpoints (bool) –

• n_startup_trials (int) –

• n_ei_candidates (int) –

• seed (float) – seed of random generater

Examples

import flopt

x = flopt.Variable("x", lowBound=-1, upBound=1, cat="Continuous")
y = flopt.Variable("y", lowBound=-1, upBound=1, cat="Continuous")

prob = flopt.Problem()
prob += 2*x*x + x*y + y*y + x + y
prob += x + y >= 0

status, log = prob.solve(solver="OptunaTPE", msg=True, timelimit=1)

print("obj =", flopt.Value(prob.obj))
print("x =", flopt.Value(x))
print("y =", flopt.Value(y))

HyperoptSearch

Solver name is “Hyperopt”.

class flopt.solvers.hyperopt_search.HyperoptSearch

TPE Search using Hyperopt (https://hyperopt.github.io/hyperopt/)

Parameters:

• n_trial (int) – number of trials

• show_progressbar (bool) – whether display a progress bar of search

Examples

import flopt

x = flopt.Variable("x", lowBound=-1, upBound=1, cat="Continuous")
y = flopt.Variable("y", lowBound=-1, upBound=1, cat="Continuous")

prob = flopt.Problem()
prob += 2*x*x + x*y + y*y + x + y

status, log = prob.solve(solver="Hyperopt", msg=True, timelimit=1)

print("obj =", flopt.Value(prob.obj))
print("x =", flopt.Value(x))
print("y =", flopt.Value(y))