WO_opt.py

import copy
import os
import pickle
import util
import datetime
from tune_util import get_vacbo_optimizer


# parameter configurations to enumerate

os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'
optimization_config = {
    'eval_budget': 300
}
optimizer_base_config = {
    'noise_level': [0.0002/5, 0.00075/5, 0.000139/5],
    'kernel_var': 1.0,
    'problem_name': 'WO',
    'normalize_input': False
}
VARS_TO_FIX = []
CONTEXTUAL_VARS = ['p1', 'p2', 'p3', 'p4']

save_name_append = f'_with_context'


def run_opt(bo_config, optimizer_type, problem_config):
    opt, opt_total_cost_list, opt_problem = get_vacbo_optimizer(
                    bo_config['problem_name'], optimizer_type, bo_config,
                     vars_to_fix=VARS_TO_FIX,
                    contextual_vars=CONTEXTUAL_VARS,
        problem_config=problem_config)

    opt_obj_list = []
    constraints1_list = []
    constraints2_list = []
    for _ in range(optimization_config['eval_budget']):
        context_vars = opt_problem.get_context()
        y_obj, constr_vals = opt.make_step(context_vars)
        if optimizer_type == 'safe_bo':
            new_cumu_cost = opt.safe_bo.cumu_vio_cost
            opt_problem = opt.safe_bo.opt_problem
        if optimizer_type == 'constrained_bo':
            new_cumu_cost = opt.constrained_bo.cumu_vio_cost
            opt_problem = opt.constrained_bo.opt_problem
        if optimizer_type == 'violation_aware_bo':
            new_cumu_cost = opt.violation_aware_bo.cumu_vio_cost
            opt_problem = opt.violation_aware_bo.opt_problem
        if optimizer_type == 'pdcbo':
            new_cumu_cost = opt.pdbo.cumu_vio_cost
            opt_problem = opt.pdbo.opt_problem
        if optimizer_type == 'no opt':
            new_cumu_cost = opt.cumu_vio_cost
            opt_problem = opt.opt_problem
        if optimizer_type == 'grid search':
            new_cumu_cost = opt.cumu_vio_cost
            opt_problem = opt.opt_problem

        opt_obj_list.append(y_obj[0,0])
        constraints1_list.append(constr_vals[:, 0])
        constraints2_list.append(constr_vals[:, 1])

        print_log = True
        if print_log:
            print(f"For the problem {opt_problem.problem_name}, in \
                  step {_}, we get objective {y_obj} and constraints \
                  {constr_vals}.")

    return opt_obj_list, constraints1_list, constraints2_list, \
        opt_problem.evaluated_points_list

problem_config = util.get_config(
    optimizer_base_config['problem_name'],
    gp_kernel='Gaussian',
    vars_to_fix=VARS_TO_FIX, contextual_vars=CONTEXTUAL_VARS)

# run PDCBO
pdcbo_config = copy.deepcopy(optimizer_base_config)
pdcbo_config.update({
        'eta_0': 1.0,
        'eta_func': lambda t: 1.0,
        'total_eval_num': optimization_config['eval_budget'],
        'init_dual': 0.0,
        'lcb_coef': lambda t: 1.0  # 1e-6
    })

optimizer_type = 'pdcbo'
opt_obj_list_pdcbo, constraints1_list_pdcbo, constraints2_list_pdcbo, \
    evaluated_points_list_pdcbo = run_opt(pdcbo_config, optimizer_type,
                                          problem_config)


cbo_config = copy.deepcopy(optimizer_base_config)
cbo_config.update({
        'total_eval_num': optimization_config['eval_budget'],
        'kernel_type': 'Gaussian'
    })

optimizer_type = 'constrained_bo'
opt_obj_list_cbo, constraints1_list_cbo, constraints2_list_cbo, \
    evaluated_points_list_cbo = run_opt(cbo_config, optimizer_type,
                                        problem_config)

safebo_config = copy.deepcopy(optimizer_base_config)
safebo_config.update({
        'total_eval_num': optimization_config['eval_budget'],
        'kernel_type': 'Gaussian'
    })
optimizer_type = 'safe_bo'
opt_obj_list_safebo, constraints1_list_safebo, constraints2_list_safebo, \
    evaluated_points_list_safebo = run_opt(safebo_config, optimizer_type,
                                           problem_config)
now_time_str = datetime.datetime.now().strftime(
        "%H_%M_%S-%b_%d_%Y")

with open(f'./result/WO_contextual_min_{now_time_str}.pkl',
          'wb') as f:
    pickle.dump([
        opt_obj_list_pdcbo,
        constraints1_list_pdcbo,
        constraints2_list_pdcbo,
        evaluated_points_list_pdcbo,
        opt_obj_list_cbo,
        constraints1_list_cbo,
        constraints2_list_cbo,
        evaluated_points_list_cbo,
        opt_obj_list_safebo,
        constraints1_list_safebo,
        constraints2_list_safebo,
        evaluated_points_list_safebo
    ], f)