ICLV estimation: Model components are inconsistent across workers. Try seting apollo_control$nCores=1
Posted: 14 May 2021, 07:54
Hi Stephane and David,
I am trying to estimate a hybrid choice model, and the multithreading process is not completed. Using 1 core, the model runs. When I estimate a Mixed-logit on the same data, I do not encounter this issue. Please see the code of the hybrid model below :
Thank you for your help!
Richard
I am trying to estimate a hybrid choice model, and the multithreading process is not completed. Using 1 core, the model runs. When I estimate a Mixed-logit on the same data, I do not encounter this issue. Please see the code of the hybrid model below :
Code: Select all
# ################################################################# #
#### LOAD LIBRARY AND DEFINE CORE SETTINGS ####
# ################################################################# #
### Clear memory
rm(list = ls())
### Load Apollo library
library(apollo)
### Initialise code
apollo_initialise()
### Set core controls
apollo_control = list(
modelName ="HCM_10.3_Additve_spec",
modelDescr ="HCM on HK mode choice" ,
indivID ="ID",
mixing = TRUE,
nCores = 5,
seed = 1
)
# ################################################################# #
#### LOAD DATA AND APPLY ANY TRANSFORMATIONS ####
# ################################################################# #
database = read.csv("Transp_data.csv",header=TRUE)
# ################################################################# #
#### DEFINE MODEL PARAMETERS ####
# ################################################################# #
### Vector of parameters, including any that are kept fixed in estimation
apollo_beta = c(asc_MMB=8.07867264304393,
asc_TX=14.1425505024284,
asc_MB=8.37124321068007,
asc_NO=0,
LV_alig_2_sig=0.15,
LV_alig_3_sig=0.1,
LV_cont_2_sig=0.1,
LV_cont_3_sig=0.3,
LV_cont_4_sig=0.2,
LV_tt_1_sig=-0.00690069481267281,
LV_tt_2_sig=-0.0526678409831866,
LV_tc_sig=-0.0153376953475673,
LV_tt_t_sig=-0.0526678409831866,
LV_tc_t_sig=-0.0153376953475673,
gamma_Repeat_vis_LV_Transport_perception=-0.330606290602541,
gamma_Long_stay_LV_Transport_perception=-0.419229905652882,
gamma_att_long_stay=-1.04105157135528,
gamma_att_Repeat_vis=1.44802784555093,
gamma_ratt_2_long_stay=5.16792949094382,
gamma_ratt_2_Repeat_vis=0.264547788636091,
gamma_ratt_3_long_stay=0.910518293398389,
gamma_ratt_3_Repeat_vis=-2.28861227368236,
gamma_ratt_4_long_stay=-0.90320923093451,
gamma_ratt_4_Repeat_vis=-1.21661259009885,
gamma_tt_1_long_stay=0.0310079716670544,
gamma_tt_1_Repeat_vis=0.00305837384742729,
gamma_tt_2_long_stay=0.0531481122826535,
gamma_tt_2_Repeat_vis=0.0496424976439684,
gamma_tc_long_stay=0.0152064786451245,
gamma_tc_Repeat_vis=0.0148539003396019,
gamma_tt_t_long_stay=0.000873118038410258,
gamma_tt_t_Repeat_vis=-0.0606427779063918,
gamma_tc_t_long_stay=-0.000955604312717446,
gamma_tc_t_Repeat_vis=0.00804111092832115,
gamma_cont_2_long_stay=-1.05719156097221,
gamma_cont_2_Repeat_vis=-0.2910588084329,
gamma_cont_3_long_stay=-1.06520463049246,
gamma_cont_3_Repeat_vis=-0.418864919754063,
gamma_cont_4_long_stay=-0.853521890329236,
gamma_cont_4_Repeat_vis=-0.243132925148644,
gamma_alig_2_long_stay=-1.94525411755059,
gamma_alig_2_Repeat_vis=-1.73449015629004,
gamma_alig_3_long_stay=-1.2861021856139,
gamma_alig_3_Repeat_vis=-1.44450104944019,
zeta_lv_perc_1=1,
tau_lv_perc_1_1=-3,
tau_lv_perc_1_2=3,
zeta_lv_perc_3=1.73160727845793,
tau_lv_perc_3_1=-7.06980942882699,
tau_lv_perc_3_2=-3.93173332063141,
tau_lv_perc_3_3=-2.27856774291353,
tau_lv_perc_3_4=1.10291753123815,
att_mu=-2.13329277915583,
att_sig=-1.60246131551844,
ratt_1=0,
ratt_2_mu=-7.04535402682224,
ratt_2_sig=4.25099356938291,
ratt_3_mu=4.91948650647052,
ratt_3_sig=2.72711149986421,
ratt_4_mu=4.7457409526998,
ratt_4_sig=3.97937621097142,
tt_1_mu=-0.0834060960217425,
tt_1_sig=0.0298362540846351,
tt_2_mu=-0.106817424568177,
tt_2_sig=0.0827947571407638,
tc_mu=-0.0828335208169694,
tc_sig=-0.0237097011316197,
tt_t_mu=-0.0974516470649308,
tt_t_sig=0.117595735340974,
tc_t_mu=-0.0677111031816054,
tc_t_sig=-0.0136903559499059,
alig_1=0,
alig_2_mu=3.74916031794965,
alig_2_sig=-0.113769248730282,
alig_3_mu=2.63595392254908,
alig_3_sig=0.0508798403249827,
cont_1=0,
cont_2_mu=1.12854769250774,
cont_2_sig=0.663869560033205,
cont_3_mu=0.507381018899945,
cont_3_sig=1.15584140932361,
cont_4_mu=1.55075862278489,
cont_4_sig=-0.00244538458101513)
### Vector with names (in quotes) of parameters to be kept fixed at their starting value in apollo_beta, use apollo_beta_fixed = c() if none
apollo_fixed = c("asc_NO","ratt_1","alig_1","cont_1")
### Read in starting values for at least some parameters from existing model output file
apollo_beta=apollo_readBeta(apollo_beta,apollo_fixed,"HCM_10.2_additve_spec",overwriteFixed = FALSE)
# ################################################################# #
#### DEFINE RANDOM COMPONENTS ####
# ################################################################# #
### Set parameters for generating draws
apollo_draws = list(
interDrawsType = "pmc",
interNDraws = 100,
interUnifDraws = c(),
interNormDraws = c("draws_LV_Tr_perc",
"draws_tt_1","draws_tt_2","draws_tc","draws_tc_t","draws_tt_t",
"draws_att","draws_ratt_2","draws_ratt_3","draws_ratt_4",
"draws_alig_2","draws_alig_3",
"draws_cont_2","draws_cont_3","draws_cont_4"),
#"draws_asc_MMB","draws_asc_MB","draws_asc_TX"),
intraDrawsType = "halton",
intraNDraws = 0,
intraUnifDraws = c(),
intraNormDraws = c()
)
### Create random parameters
apollo_randCoeff = function(apollo_beta, apollo_inputs){
randcoeff = list()
randcoeff[["LV_Tr_perc"]] = draws_LV_Tr_perc + gamma_Repeat_vis_LV_Transport_perception*Repeat_vis +
gamma_Long_stay_LV_Transport_perception*four_nights_long
randcoeff[["tt_1"]] = tt_1_mu + tt_1_sig * draws_tt_1+gamma_tt_1_long_stay*four_nights_long +gamma_tt_1_Repeat_vis*Repeat_vis
randcoeff[["tt_2"]] = tt_2_mu + tt_2_sig * draws_tt_2+gamma_tt_2_long_stay*four_nights_long +gamma_tt_2_Repeat_vis*Repeat_vis
randcoeff[["tc"]] = tc_mu + tc_sig * draws_tc+gamma_tc_long_stay*four_nights_long + gamma_tc_Repeat_vis*Repeat_vis
randcoeff[["tt_t"]] = tt_t_mu + tt_t_sig * draws_tt_t+gamma_tt_t_long_stay*four_nights_long +gamma_tt_t_Repeat_vis*Repeat_vis
randcoeff[["tc_t"]] = tc_t_mu + tc_t_sig * draws_tc_t+gamma_tc_t_long_stay*four_nights_long +gamma_tc_t_Repeat_vis*Repeat_vis
randcoeff[["att"]] = att_mu + att_sig * draws_att+gamma_att_long_stay*four_nights_long + gamma_att_Repeat_vis*Repeat_vis
randcoeff[["ratt_2"]] = ratt_2_mu + ratt_2_sig * draws_ratt_2+gamma_ratt_2_long_stay*four_nights_long + gamma_ratt_2_Repeat_vis*Repeat_vis
randcoeff[["ratt_3"]] = ratt_3_mu + ratt_3_sig * draws_ratt_3+gamma_ratt_3_long_stay*four_nights_long + gamma_ratt_3_Repeat_vis*Repeat_vis
randcoeff[["ratt_4"]] = ratt_4_mu + ratt_4_sig * draws_ratt_4+gamma_ratt_4_long_stay*four_nights_long + gamma_ratt_4_Repeat_vis*Repeat_vis
randcoeff[["cont_2"]] = cont_2_mu + cont_2_sig * draws_cont_2+gamma_cont_2_long_stay*four_nights_long + gamma_cont_2_Repeat_vis*Repeat_vis
randcoeff[["cont_3"]] = cont_3_mu + cont_3_sig * draws_cont_3+gamma_cont_3_long_stay*four_nights_long + gamma_cont_3_Repeat_vis*Repeat_vis
randcoeff[["cont_4"]] = cont_4_mu + cont_4_sig * draws_cont_4+gamma_cont_4_long_stay*four_nights_long + gamma_cont_4_Repeat_vis*Repeat_vis
randcoeff[["alig_2"]] = alig_2_mu + alig_2_sig * draws_alig_2+gamma_alig_2_long_stay*four_nights_long + gamma_alig_2_Repeat_vis*Repeat_vis
randcoeff[["alig_3"]] = alig_3_mu + alig_3_sig * draws_alig_3+gamma_alig_3_long_stay*four_nights_long + gamma_alig_3_Repeat_vis*Repeat_vis
return(randcoeff)
}
# ################################################################# #
#### GROUP AND VALIDATE INPUTS ####
# ################################################################# #
apollo_inputs = apollo_validateInputs()
# ################################################################# #
#### DEFINE MODEL AND LIKELIHOOD FUNCTION ####
# ################################################################# #
apollo_probabilities=function(apollo_beta, apollo_inputs, functionality="estimate"){
### Function initialisation: do not change the following three commands
### Attach inputs and detach after function exit
apollo_attach(apollo_beta, apollo_inputs)
on.exit(apollo_detach(apollo_beta, apollo_inputs))
### Create list of probabilities P
P = list()
## PERCVULNEV
OL_settings1 = list(outcomeOrdered=Q49_1.1,
V=zeta_lv_perc_1*LV_Tr_perc,
tau=c(tau_lv_perc_1_1,tau_lv_perc_1_2),
rows=(Sequence==1),
componentName = "indic_1")
OL_settings3 = list(outcomeOrdered=Q49_3,
V=zeta_lv_perc_3*LV_Tr_perc,
tau=c(tau_lv_perc_3_1,tau_lv_perc_3_2,tau_lv_perc_3_3,tau_lv_perc_3_4),
rows=(Sequence==1),
componentName = "indic_3")
P[["indic_1"]] = apollo_ol(OL_settings1, functionality)
P[["indic_3"]] = apollo_ol(OL_settings3, functionality)
### Creating interactions with latent variable
tt_1_val=tt_1+LV_tt_1_sig*LV_Tr_perc
tt_2_val=tt_2+LV_tt_2_sig*LV_Tr_perc
tt_t_val=tt_t+LV_tt_t_sig*LV_Tr_perc
### List of utilities: these must use the same names as in mnl_settings, order is irrelevant
V = list()
V[['MMB']] = asc_MMB + tt_1_val*mm.tt1 + tt_2_val*mm.tt2 + tc_val*mm.tc + cont_1*(mm.cont==0) + cont_2_val*(mm.cont==1) + cont_3_val*(mm.cont==2) +cont_4_val*(mm.cont==3) + alig_1*(mm.aligm==0) + alig_2_val*(mm.aligm==1) + alig_3_val*(mm.aligm==2)
V[['MBUS']] = asc_MB + tt_1_val*mb.tt1 + tt_2_val*mb.tt2 + tc_val*mb.tc + cont_1*(mb.cont==0) + cont_2_val*(mb.cont==1) + cont_3_val*(mb.cont==2) +cont_4_val*(mb.cont==3) + alig_1*(mb.aligb==0) + alig_2_val*(mb.aligb==1) + alig_3_val*(mb.aligb==2)
V[['TAXI']] = asc_TX + tt_t_val*tx.ttt + tc_t_val*tx.ttc
V[['NO']] = asc_NO + att*mm.attr + ratt_1*(mm.rattr==0) + ratt_2*(mm.rattr==1) + ratt_3*(mm.rattr==2) + ratt_4*(mm.rattr==3)
### Define settings for MNL model component
mnl_settings = list(
alternatives = c(MMB=1, MBUS=2, TAXI=3, NO=4),
avail = list(MMB=1, MBUS=1, TAXI=1, NO=1),
choiceVar = Choice1,
V = V,
componentName = "Mode_choice"
)
### Compute probabilities for MNL model component
P[["Mode_choice"]] = apollo_mnl(mnl_settings, functionality)
### Likelihood of the whole model
P = apollo_combineModels(P, apollo_inputs, functionality)
### Take product across observation for same individual
P = apollo_panelProd(P, apollo_inputs, functionality)
### Average across inter-individual draws
P = apollo_avgInterDraws(P, apollo_inputs, functionality)
### Prepare and return outputs of function
P = apollo_prepareProb(P, apollo_inputs, functionality)
return(P)
}
# ################################################################# #
#### MODEL ESTIMATION ####
# ################################################################# #
model = apollo_estimate(apollo_beta, apollo_fixed,apollo_probabilities, apollo_inputs) #estimate_settings = list(maxIterations=500,hessianRoutine="maxLik"))
Richard