def gen_conf_to_optimize(save_path="experimentation/optimize/multiconf/",
main_act="KT6kc",
ref_stud="0",
hierarUtile=0):
if ref_stud in ["2"]:
hierarUtile = 1
filter1_vals, stepUp_vals, upZPD_vals, deact_vals, prom_coef_vals, thresHDeact_vals = all_values_zpdes(
hierarUtile)
print filter1_vals
print stepUp_vals
print upZPD_vals
print deact_vals
print prom_coef_vals
print thresHDeact_vals
multi_confs = {
"ZpdesSsbg": {
"ZpdesSsb": {
"filter1": filter1_vals,
"stepUpdate": stepUp_vals,
"upZPDval": upZPD_vals,
"deactZPDval": deact_vals,
"promote_coeff": prom_coef_vals,
"thresHDeact": thresHDeact_vals
}
}
}
base_conf = func.load_json("ZPDES_base", "params_files/ZPDES")
base_conf["graph"]["file_name"] = "graph_{}_{}".format(main_act, ref_stud)
base_conf["graph"]["main_act"] = main_act
base_conf["graph"].update(
func.load_json(base_conf["graph"]["file_name"],
base_conf["graph"]["path"]))
print base_conf
zpdes_confs = mp.multi_conf(base_conf=base_conf,
multi_params=multi_confs,
combine=1)
#conf_ids = mp.generate_diff_config_id(zpdes_confs)
#zpdes_confs = {conf_ids[x] : zpdes_confs[x] for x in range(len(zpdes_confs))}
uid = str(uuid.uuid1())
jstr = json.dumps(zpdes_confs)
k_lib.config.datafile.create_directories([save_path])
save_path = "{}{}_{}_all_confs.json".format(save_path, main_act, ref_stud)
k_lib.functions.write_in_file(save_path, jstr)
return zpdes_confs
def perturbated_population(zpdes_confs,
ref_xp="perturbation",
nb_step=100,
base_path_to_save="experimentation/data/"):
population_conf = func.load_json(params_file="perturbation_KT6kc",
directory="params_files/studModel")
population = k_lib.student.Population(params=population_conf)
nb_stud = population.nb_students
wkgs = {}
for ref, conf in zpdes_confs.items():
zpdes = k_lib.seq_manager.ZpdesHssbg(params=conf)
wss = []
for k in range(nb_stud):
wss.append(k_lib.experimentation.WorkingSession(
student=copy.deepcopy(population.students[k])),
seq_manager=copy.deepcopy(zpdes))
wkgs["zpdes_{}".format(ref)] = [
k_lib.experimentation.WorkingGroup(WorkingSessions=wss)
]
xp = k_lib.experimentation.Experiment(WorkingGroups=wkgs,
ref_expe=ref_xp,
path_to_save=base_path_to_save,
seq_manager_list=wkgs.keys(),
nb_step=nb_step,
population={
"nb_students": nb_stud,
"model": "KT_student"
})
return xp
def gen_stud_confs(nb_students=1000,
perturbated=0,
params_file="stud_KT6kc_0"):
studconf = func.load_json(params_file, "params_files/studModel")
if perturbated == 1:
pass
else:
stud_confs = [copy.deepcopy(studconf) for x in range(nb_students)]
return stud_confs
def multi_kt_xp(ref_xp="KT6kc",
path_to_save="experimentation/data/",
nb_step=100,
nb_stud=100,
files_to_load=None,
ref_bis="",
disruption_pop_file="perturbation_KT6kc",
disruption=0,
refs_opti=["0", "1", "2", "3"],
refs_stud=["0", "1", "2", "3"],
save_xp_data=0,
mixed_pop=1):
if type(refs_stud) != list:
refs_stud = [refs_stud]
ref_bis = "{}opti{}_stud{}".format(ref_bis, "".join(refs_opti),
"".join(refs_stud))
if files_to_load is None:
files_to_load = {}
def_files_to_load = {}
def_files_to_load["pomdp"] = []
def_files_to_load["stud"] = []
for ref in refs_opti:
def_files_to_load["pomdp"].append({
"file_name":
"POMDP_{}_{}".format(ref_xp, ref),
"path":
"data/pomdp"
})
for ref in refs_stud:
def_files_to_load["stud"].append("stud_{}_{}".format(ref_xp, ref))
def_files_to_load["zpdes"] = "ZPDES_base"
#def_files_to_load["riarit"] = "RIARIT_{}".format(ref_xp)
def_files_to_load["random"] = "RANDOM_base"
for key, val in def_files_to_load.items():
if key not in files_to_load.keys():
files_to_load[key] = val
#pomdP = k_lib.config.datafile.load_file("KT_expe_2","data/pomdp")
zpdes_conf_opti = {}
zpdes_params_opti = {
"0": [0.2, 0.05, 4, 0.6, 0.7, 0.8, 0.5, 0.5],
"1": [0.1, 0.05, 4, 0.7, 0.6, 0.8, 0.5, 0.5],
"2": [0.5, 0.05, 4, 0.8, 0.7, 1.5, 0.5, 0.5],
"3": [0.4, 0.05, 4, 0.8, 0.6, 1.2, 0.5, 0.5],
"4": [0.4, 0.05, 4, 0.8, 0.6, 1.2, 0.5, 0.5],
}
for ref in refs_opti:
zpdes_conf_opti[ref] = {
"algo_name": "ZpdesHssbg",
"graph": {
"file_name": "graph_{}_{}".format(ref_xp, ref),
"path": "graph/",
"main_act": "{}".format(ref_xp)
},
"ZpdesSsbg": {
"ZpdesSsb": {
"filter1": zpdes_params_opti[ref][0], # 0.4,# 0.30,
"uniformval": zpdes_params_opti[ref][1], # 0.05,
"stepUpdate": zpdes_params_opti[ref][2], # 4,# 6,
"upZPDval": zpdes_params_opti[ref][3], # 0.6,
"deactZPDval": zpdes_params_opti[ref][4], # 0.4,# 0.5,
"promote_coeff": zpdes_params_opti[ref][5], # 1.7,
"thresHProm": zpdes_params_opti[ref][6], # 0.5,
"thresHDeact": zpdes_params_opti[ref][7]
}
}
}
params_pop = []
for ref in refs_stud:
p_pop = func.load_json("{}_{}".format(disruption_pop_file, ref),
"params_files/studModel")
p_pop["nb_students"] = nb_stud
params_pop.append(p_pop)
if disruption == 1:
population = k_lib.student.Population(params=params_pop)
elif mixed_pop:
stud_list = []
for ii in range(len(refs_stud)):
for i in range(nb_stud):
stud_list.append(
k_lib.student.KTstudent(
params_file=files_to_load["stud"][ii],
directory="params_files/studModel"))
population = k_lib.student.Population(stud_list=stud_list)
else:
stud_list = {}
for ii in range(len(refs_stud)):
stud_list[refs_stud[ii]] = []
for i in range(nb_stud):
stud_list[refs_stud[ii]].append(
k_lib.student.KTstudent(
params_file=files_to_load["stud"][ii],
directory="params_files/studModel"))
population = k_lib.student.Population(stud_list=stud_list)
#stud = k_lib.student.KTstudent(params=population.base_model)
# nb_stud = nb_stud # population.nb_students
pomdPs = []
zpdesOpts = []
zpdesHs = []
zpdes_params = func.load_json(file_name=files_to_load["zpdes"],
dir_path="params_files/ZPDES")
# zpdesH = k_lib.seq_manager.ZpdesHssbg(zpdes_params) # params=zpdes_params)
for i in range(len(refs_opti)):
pomdPs.append(
k_lib.seq_manager.POMDP(load_p=files_to_load["pomdp"][i]))
zpdesOpts.append(
k_lib.seq_manager.ZpdesHssbg(
zpdes_conf_opti[refs_opti[i]])) # params=zpdes_params)
zpdes_params = change_graph_params(
zpdes_params, "graph_{}_{}".format(ref_xp, refs_opti[i]), ref_xp)
zpdesHs.append(k_lib.seq_manager.ZpdesHssbg(zpdes_params))
#riarit = k_lib.seq_manager.RiaritHssbg(params_file=files_to_load["riarit"],directory="params_files/RIARIT")
random_params = func.load_json(file_name=files_to_load["random"],
dir_path="params_files/RANDOM")
random_params = change_graph_params(random_params,
"{}_graph".format(ref_xp), ref_xp)
#random = k_lib.seq_manager.RandomSequence(random_params)
ws_tab_zpdes = [[] for i in range(len(refs_opti))]
ws_tab_zpdesOpt = [[] for i in range(len(refs_opti))]
ws_tab_pomdp = [[] for i in range(len(refs_opti))]
#ws_tab_riarit = []
#ws_tab_random = []
for i in range(len(population.students)):
stud = population.students[i]
for ii in range(len(refs_opti)):
ws_tab_zpdes[ii].append(
k_lib.experimentation.WorkingSession(
student=cPickle.loads(cPickle.dumps(stud, -1)),
seq_manager=cPickle.loads(cPickle.dumps(zpdesHs[ii], -1))))
ws_tab_zpdesOpt[ii].append(
k_lib.experimentation.WorkingSession(
student=cPickle.loads(cPickle.dumps(stud, -1)),
seq_manager=cPickle.loads(cPickle.dumps(zpdesOpts[ii],
-1))))
ws_tab_pomdp[ii].append(
k_lib.experimentation.WorkingSession(
student=cPickle.loads(cPickle.dumps(stud, -1)),
seq_manager=cPickle.loads(cPickle.dumps(pomdPs[ii], -1))))
#ws_tab_riarit.append(k_lib.experimentation.WorkingSession(student=copy.deepcopy(stud), seq_manager = copy.deepcopy(riarit)))
#ws_tab_random.append(k_lib.experimentation.WorkingSession(student=copy.deepcopy(stud), seq_manager=copy.deepcopy(random)))
wG_zpdes = [
k_lib.experimentation.WorkingGroup(WorkingSessions=x)
for x in ws_tab_zpdes
]
wG_zpdes2 = [
k_lib.experimentation.WorkingGroup(WorkingSessions=x)
for x in ws_tab_zpdesOpt
]
wG_pomdp = [
k_lib.experimentation.WorkingGroup(WorkingSessions=x)
for x in ws_tab_pomdp
]
#wG_riarit = k_lib.experimentation.WorkingGroup(WorkingSessions = ws_tab_riarit)
#wG_random = k_lib.experimentation.WorkingGroup(WorkingSessions=ws_tab_random)
wkgs = {
"POMDP": wG_pomdp,
"ZpdesH": wG_zpdes,
"Zpdes*": wG_zpdes2
#"Random": [wG_random]
} # , "RIARIT": [wG_riarit]} # {"ZPDES": [wG_zpdes]} #
ref_xp_bis = "{}{}".format(ref_xp, ref_bis)
params = {
"ref_expe": ref_xp_bis,
"path_to_save": path_to_save,
"seq_manager_list": wkgs.keys(), # "RIARIT"
"nb_step": nb_step,
"population": {
"nb_students": nb_stud * len(refs_stud),
"model": "KT_student",
"ref_stud": "stud_{}_{}".format(ref_xp, "".join(refs_stud))
}
}
xp = k_lib.experimentation.Experiment(WorkingGroups=wkgs, params=params)
xp.run(nb_step)
ref_sub_group = [copy.deepcopy(refs_opti)]
#draw_xp_kc_curve(xp, ref_sub_group=ref_sub_group, subgroup_treat=True)
#draw_xp_kc_curve(xp)
# values_vect = ["V{}".format(i + 1) for i in range(len(xp.KC))]
# draw_xp_graph(xp, type_ex=values_vect, nb_ex_type=[1] * len(xp.KC), ref_sub_group=ref_sub_group)
# draw_xp_histo(xp, type_ex=values_vect, nb_ex_type=[1] * len(xp.KC),)
#pvals = calcul_pvals(xp)
#draw_pvals(xp, pvals)
#all_mean_data = draw_xp_graph(xp,ref_xp)
#cost = calcul_xp_cost(xp)
if save_xp_data:
xp.save()
return xp
def kt_expe(ref_xp="KT6kc",
path_to_save="experimentation/data/",
nb_step=100,
nb_stud=100,
files_to_load=None,
ref_bis="",
disruption_pop_file="perturbation_KT6kc",
disruption=0,
ref_pomdp="_2",
ref_stud="_2",
save_xp_data=0):
if files_to_load is None:
files_to_load = {}
def_files_to_load = {}
def_files_to_load["pomdp"] = {
"file_name": "POMDP_{}{}".format(ref_xp, ref_pomdp),
"path": "data/pomdp"
}
def_files_to_load["stud"] = "stud_{}{}".format(ref_xp, ref_stud)
def_files_to_load["zpdes"] = "ZPDES_base"
#def_files_to_load["riarit"] = "RIARIT_{}".format(ref_xp)
def_files_to_load["random"] = "RANDOM_base"
for key, val in def_files_to_load.items():
if key not in files_to_load.keys():
files_to_load[key] = val
pomdP = k_lib.seq_manager.POMDP(load_p=files_to_load["pomdp"])
#pomdP = k_lib.config.datafile.load_file("KT_expe_2","data/pomdp")
zpdes_conf_opti = {}
zpdes_params_0opti = {
"algo_name": "ZpdesHssbg",
"graph": {
"file_name": "{}_graph".format(ref_xp),
"path": "graph/",
"main_act": "{}".format(ref_xp)
},
"ZpdesSsbg": {
"ZpdesSsb": {
"filter1": 0.2, # 0.4,# 0.30,
"uniformval": 0.05, # 0.05,
"stepUpdate": 4, # 6,# 6,
"upZPDval": 0.6, # 0.6,
"deactZPDval": 0.7, # 0.4,# 0.5,
"promote_coeff": 0.8, # 1.7,
"thresHProm": 0.5, # 0.5,
"h_promote_coeff": 0.25, # 0.25,
"size_window": 3, # 3,
"spe_promo": 0 # 0
}
}
}
zpdes_params_1opti = {
"algo_name": "ZpdesHssbg",
"graph": {
"file_name": "{}_graph".format(ref_xp),
"path": "graph/",
"main_act": "{}".format(ref_xp)
},
"ZpdesSsbg": {
"ZpdesSsb": {
"filter1": 0.5, # 0.4,# 0.30,
"uniformval": 0.05, # 0.05,
"stepUpdate": 4, # 6,# 6,
"upZPDval": 0.7, # 0.6,
"deactZPDval": 0.6, # 0.4,# 0.5,
"promote_coeff": 0.6, # 1.7,
"thresHProm": 0.5, # 0.5,
"h_promote_coeff": 0.25, # 0.25,
"size_window": 3, # 3,
"spe_promo": 0 # 0
}
}
}
zpdes_params_2opti = {
"algo_name": "ZpdesHssbg",
"graph": {
"file_name": "{}_graph".format(ref_xp),
"path": "graph/",
"main_act": "{}".format(ref_xp)
},
"ZpdesSsbg": {
"ZpdesSsb": {
"filter1": 0.1, # 0.4,# 0.30,
"uniformval": 0.05, # 0.05,
"stepUpdate": 4, # 6,# 6,
"upZPDval": 0.6, # 0.6,
"deactZPDval": 0.7, # 0.4,# 0.5,
"promote_coeff": 1.6, # 1.7,
"thresHProm": 0.5, # 0.5,
"h_promote_coeff": 0.25, # 0.25,
"size_window": 3, # 3,
"spe_promo": 0 # 0
}
}
}
zpdes_conf_opti["_0"] = zpdes_params_0opti
zpdes_conf_opti["_1"] = zpdes_params_1opti
zpdes_conf_opti["_2"] = zpdes_params_2opti
#population_conf = func.load_json()
population = k_lib.student.Population(params_file=disruption_pop_file,
directory="params_files/studModel")
#stud = k_lib.student.KTstudent(params=population.base_model)
stud = k_lib.student.KTstudent(params_file=files_to_load["stud"],
directory="params_files/studModel")
nb_stud = nb_stud # population.nb_students
zpdes_params = func.load_json(file_name=files_to_load["zpdes"],
dir_path="params_files/ZPDES")
zpdes_params = change_graph_params(zpdes_params, "{}_graph".format(ref_xp),
ref_xp)
zpdes = k_lib.seq_manager.ZpdesHssbg(zpdes_params) # params=zpdes_params)
zpdesOpt = k_lib.seq_manager.ZpdesHssbg(
zpdes_conf_opti[ref_pomdp]) # params=zpdes_params)
#riarit = k_lib.seq_manager.RiaritHssbg(params_file=files_to_load["riarit"],directory="params_files/RIARIT")
random_params = func.load_json(file_name=files_to_load["random"],
dir_path="params_files/RANDOM")
random_params = change_graph_params(random_params,
"{}_graph".format(ref_xp), ref_xp)
random = k_lib.seq_manager.RandomSequence(random_params)
ws_tab_zpdes = []
ws_tab_zpdesOpt = []
#ws_tab_riarit = []
ws_tab_random = []
ws_tab_pomdp = []
for i in range(nb_stud):
if disruption == 1:
stud = population.students[i]
ws_tab_zpdes.append(
k_lib.experimentation.WorkingSession(
student=copy.deepcopy(stud), seq_manager=copy.deepcopy(zpdes)))
ws_tab_zpdesOpt.append(
k_lib.experimentation.WorkingSession(
student=copy.deepcopy(stud),
seq_manager=copy.deepcopy(zpdesOpt)))
ws_tab_pomdp.append(
k_lib.experimentation.WorkingSession(
student=copy.deepcopy(stud), seq_manager=copy.deepcopy(pomdP)))
#ws_tab_riarit.append(k_lib.experimentation.WorkingSession(student=copy.deepcopy(stud), seq_manager = copy.deepcopy(riarit)))
ws_tab_random.append(
k_lib.experimentation.WorkingSession(
student=copy.deepcopy(stud),
seq_manager=copy.deepcopy(random)))
wG_zpdes = k_lib.experimentation.WorkingGroup(WorkingSessions=ws_tab_zpdes)
wG_zpdesOpt = k_lib.experimentation.WorkingGroup(
WorkingSessions=ws_tab_zpdesOpt)
wG_pomdp = k_lib.experimentation.WorkingGroup(WorkingSessions=ws_tab_pomdp)
#wG_riarit = k_lib.experimentation.WorkingGroup(WorkingSessions = ws_tab_riarit)
#wG_random = k_lib.experimentation.WorkingGroup(WorkingSessions=ws_tab_random)
wkgs = {
"POMDP": [wG_pomdp],
"ZPDES": [wG_zpdes],
#"Random": [wG_random],
"ZPDES2": [wG_zpdesOpt]
} # , "RIARIT": [wG_riarit]} # {"ZPDES": [wG_zpdes]} #
ref_xp_bis = "{}{}".format(ref_xp, ref_bis)
params = {
"ref_expe": ref_xp_bis,
"path_to_save": path_to_save,
"seq_manager_list": wkgs.keys(), # "RIARIT"
"nb_step": nb_step,
"population": {
"nb_students": nb_stud,
"model": "KT_student",
}
}
xp = k_lib.experimentation.Experiment(WorkingGroups=wkgs, params=params)
xp.run(nb_step)
values_vect = ["V{}".format(i + 1) for i in range(len(xp.KC))]
draw_xp_graph(xp, type_ex=values_vect, nb_ex_type=[1] * len(xp.KC))
#all_mean_data = draw_xp_graph(xp,ref_xp)
#cost = calcul_xp_cost(xp)
if save_xp_data:
xp.save()
return xp