From a47a5b0222e050db54e571068e0fff163b5ab7fd Mon Sep 17 00:00:00 2001
From: Antonio Andriella <aandriella@iri.upc.edu>
Date: Wed, 11 Nov 2020 08:13:48 +0100
Subject: [PATCH] working version for experiment of the simulator
---
.gitignore | 4 +
bn_agent_model/agent_assistive_model.bif | 49 ---
bn_agent_model/agent_feedback_model.bif | 77 ----
bn_agent_model/utilities.py | 78 -----
bn_functions.py | 2 +-
bn_models/agent_model_id_1_True.bif | 40 +++
bn_models/agent_model_template.bif | 26 ++
.../persona_model_template.bif | 18 -
bn_models/user_model_id_1_True.bif | 51 +++
bn_other_user_model/user_model.bif | 103 ------
.../__pycache__/utilities.cpython-36.pyc | Bin 2460 -> 0 bytes
bn_persona_model/persona_model.bif | 103 ------
bn_persona_model/user_action_model.bif | 84 -----
bn_persona_model/user_react_time_model.bif | 95 -----
bn_persona_model/utilities.py | 90 -----
.../__pycache__/utilities.cpython-36.pyc | Bin 3939 -> 0 bytes
simulation.py | 330 +++++++++---------
test.py | 35 +-
18 files changed, 304 insertions(+), 881 deletions(-)
delete mode 100644 bn_agent_model/agent_assistive_model.bif
delete mode 100644 bn_agent_model/agent_feedback_model.bif
delete mode 100644 bn_agent_model/utilities.py
create mode 100644 bn_models/agent_model_id_1_True.bif
create mode 100644 bn_models/agent_model_template.bif
rename bn_persona_model/persona_model_test.bif => bn_models/persona_model_template.bif (58%)
create mode 100644 bn_models/user_model_id_1_True.bif
delete mode 100644 bn_other_user_model/user_model.bif
delete mode 100644 bn_persona_model/__pycache__/utilities.cpython-36.pyc
delete mode 100644 bn_persona_model/persona_model.bif
delete mode 100644 bn_persona_model/user_action_model.bif
delete mode 100644 bn_persona_model/user_react_time_model.bif
delete mode 100644 bn_persona_model/utilities.py
delete mode 100644 bn_robot_model/__pycache__/utilities.cpython-36.pyc
diff --git a/.gitignore b/.gitignore
index af56f61..6016778 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,2 +1,6 @@
.DS_Store
.idea/
+__pycache__/
+old_models/
+results/
+
diff --git a/bn_agent_model/agent_assistive_model.bif b/bn_agent_model/agent_assistive_model.bif
deleted file mode 100644
index e3b6291..0000000
--- a/bn_agent_model/agent_assistive_model.bif
+++ /dev/null
@@ -1,49 +0,0 @@
-network agent_assistive_model {
-}
-
-%VARIABLES DEFINITION
-
-variable agent_assistance {
- type discrete [ 6 ] { lev_0, lev_1, lev_2, lev_3, lev_4, lev_5};
-}
-variable attempt {
- type discrete [ 4 ] { att_1, att_2, att_3, att_4 };
-}
-variable game_state {
- type discrete [ 3 ] { beg, mid, end };
-}
-
-
-%INDIVIDUAL PROBABILITIES DEFINITION
-probability ( agent_assistance ) {
- table 0.17, 0.17, 0.17, 0.17, 0.16, 0.16;
-}
-probability ( game_state ) {
- table 0.34, 0.33, 0.33;
-}
-probability ( attempt ) {
- table 0.25, 0.25, 0.25, 0.25;
-}
-
-
-#Conditional Probabilities
-
-
-probability (agent_assistance | game_state, attempt) {
-
-(beg, att_1) 0.5, 0.3, 0.2, 0.0, 0.0, 0.0;
-(beg, att_2) 0.2, 0.3, 0.2, 0.1, 0.1, 0.1;
-(beg, att_3) 0.3, 0.2, 0.2, 0.1, 0.1, 0.1;
-(beg, att_4) 0.3, 0.3, 0.1, 0.1, 0.1, 0.1;
-
-(mid, att_1) 0.5, 0.3, 0.2, 0.0, 0.0, 0.0;
-(mid, att_2) 0.2, 0.3, 0.2, 0.1, 0.1, 0.1;
-(mid, att_3) 0.3, 0.2, 0.2, 0.1, 0.1, 0.1;
-(mid, att_4) 0.3, 0.3, 0.1, 0.1, 0.1, 0.1;
-
-(end, att_1) 0.5, 0.3, 0.2, 0.0, 0.0, 0.0;
-(end, att_2) 0.2, 0.3, 0.2, 0.1, 0.1, 0.1;
-(end, att_3) 0.3, 0.2, 0.2, 0.1, 0.1, 0.1;
-(end, att_4) 0.3, 0.3, 0.1, 0.1, 0.1, 0.1;
-
-}
diff --git a/bn_agent_model/agent_feedback_model.bif b/bn_agent_model/agent_feedback_model.bif
deleted file mode 100644
index 7321377..0000000
--- a/bn_agent_model/agent_feedback_model.bif
+++ /dev/null
@@ -1,77 +0,0 @@
-network agent_feedback_model {
-}
-
-%VARIABLES DEFINITION
-variable reactivity {
- type discrete [3] {slow, medium, fast};
-}
-variable memory {
- type discrete[3] {low, medium, high};
-}
-variable agent_feedback {
- type discrete [ 2 ] { no, yes };
-}
-variable attempt {
- type discrete [ 4 ] { att_1, att_2, att_3, att_4 };
-}
-variable game_state {
- type discrete [ 3 ] { beg, mid, end };
-}
-
-variable user_capability {
- type discrete [ 3 ] { very_mild, mild, severe };
-}
-
-%INDIVIDUAL PROBABILITIES DEFINITION
-probability ( agent_feedback ) {
- table 0.5, 0.5;
-}
-probability ( game_state ) {
- table 0.34, 0.33, 0.33;
-}
-probability ( attempt ) {
- table 0.25, 0.25, 0.25, 0.25;
-}
-probability ( user_capability ) {
- table 0.33, 0.33, 0.34;
-}
-#CPDS 4 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( reactivity ) {
- table 0.34, 0.33, 0.33;
-}
-#CPDS 3 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( memory ) {
- table 0.33, 0.33, 0.34;
-}
-
-#Conditional Probabilities
-
-#CPDS X (very_mild, mild, severe)
-probability (user_capability | memory, reactivity) {
-(low, slow) 0.1, 0.2, 0.7;
-(medium, slow) 0.2, 0.6, 0.2;
-(high, slow) 0.7, 0.2, 0.1;
-
-(low, medium) 0.1, 0.3, 0.6;
-(medium, medium) 0.3, 0.6, 0.1;
-(high, medium) 0.1, 0.4, 0.5;
-
-(low, fast) 0.3, 0.2, 0.5;
-(medium, fast) 0.7, 0.2, 0.1;
-(high, fast) 0.8, 0.1, 0.1;
-}
-
-probability (agent_feedback | user_capability) {
-(very_mild) 0.2, 0.8;
-(mild) 0.5, 0.5;
-(severe) 0.8, 0.2;
-}
-
-probability (game_state | agent_feedback) {
-(no) 0.2, 0.4, 0.4;
-(yes) 0.4, 0.4, 0.2;
-}
-probability (attempt | agent_feedback) {
-(no) 0.1,0.15, 0.25, 0.5;
-(yes) 0.4, 0.3, 0.2, 0.1;
-}
diff --git a/bn_agent_model/utilities.py b/bn_agent_model/utilities.py
deleted file mode 100644
index 56fcb05..0000000
--- a/bn_agent_model/utilities.py
+++ /dev/null
@@ -1,78 +0,0 @@
-import random
-import bn_functions
-
-
-def get_dynamic_variables(evidence_variables_name, evidence_variables_value):
- '''
- This func returns a dict of the form name:value and it defines the "evidences"
- that will be used to query the BN
- Args:
- :evidence_variables_name: the name of the variable
- :evidence_variables_value: the value of the given variable
- Return:
- a dict of the form name:value
- '''
- if len(evidence_variables_name)!=len(evidence_variables_value):
- assert "The variables name numbers is different from the variables value"
- else:
- dynamic_variables = {evidence_variables_name[i]:evidence_variables_value[i] for i in range(len(evidence_variables_name))}
- return dynamic_variables
-
-def infer_prob(user_bn_model, variable_to_infer, evidence_vars_name, evidence_vars_value):
- '''
- Given the model, the variable to infer, and the evidences returns the distribution prob for that variable
- Args:
- user_bn_model:
- variable_to_infer:
- evidence_vars_name:
- evidence_vars_value:
- Returns:
- the probability distribution for varibale_to_infer
- '''
- evidence = get_dynamic_variables(evidence_vars_name, evidence_vars_value)
- dist_prob = bn_functions.get_inference_from_state(user_bn_model,
- variables=variable_to_infer,
- evidence=evidence)
- return dist_prob
-
-def get_stochastic_action(actions_distr_prob):
- '''
- Select one of the actions according to the actions_prob
- Args:
- actions_prob: the probability of the Persona based on the BN to make a correct move, wrong move, timeout
- Return:
- the id of the selected action
- N.B:
- '''
- def compute_distance(values, target):
- '''
- Return the index of the most closest value in values to target
- Args:
- target: the target value
- values: a list of values from 0 to 1
- Return:
- return the index of the value closer to target
- '''
- min_dist = 1
- index = 0
- for i in range(len(values)):
- if abs(target-values[i])<min_dist:
- min_dist = abs(target-values[i])
- index = i
- return index
-
- actions_distr_prob_scaled = [0]*len(actions_distr_prob)
- accum = 0
- for i in range(len(actions_distr_prob)):
- accum += actions_distr_prob[i]
- actions_distr_prob_scaled[i] = accum
-
- rnd_val = random.uniform(0, 1)
- action_id = compute_distance(actions_distr_prob_scaled, rnd_val)
-
- return action_id
-
-
-actions_prob_distr = [0.32, 0.105, 0.035, 0.035, 0.005, 0.36, 0.065, 0.035, 0.035, 0.005]
-action_index = get_stochastic_action(actions_prob_distr)
-print(action_index)
\ No newline at end of file
diff --git a/bn_functions.py b/bn_functions.py
index a8502a0..c528861 100644
--- a/bn_functions.py
+++ b/bn_functions.py
@@ -178,7 +178,7 @@ def get_stochastic_action(actions_distr_prob):
accum += actions_distr_prob[i]
actions_distr_prob_scaled[i] = accum
- rnd_val = random.uniform(0, 1)
+ rnd_val = random.random()
action_id = compute_distance(actions_distr_prob_scaled, rnd_val)
return action_id
diff --git a/bn_models/agent_model_id_1_True.bif b/bn_models/agent_model_id_1_True.bif
new file mode 100644
index 0000000..ebc8cee
--- /dev/null
+++ b/bn_models/agent_model_id_1_True.bif
@@ -0,0 +1,40 @@
+network agent_assistive_model {
+}
+
+%VARIABLES DEFINITION
+
+variable agent_assistance {
+ type discrete [ 6 ] { lev_0, lev_1, lev_2, lev_3, lev_4, lev_5};
+}
+variable attempt {
+ type discrete [ 4 ] { att_1, att_2, att_3, att_4 };
+}
+variable game_state {
+ type discrete [ 3 ] { beg, mid, end };
+}
+
+
+%INDIVIDUAL PROBABILITIES DEFINITION
+probability ( agent_assistance ) {
+ table 0.17, 0.17, 0.17, 0.17, 0.16, 0.16;
+}
+probability ( game_state ) {
+ table 0.34, 0.33, 0.33;
+}
+probability ( attempt ) {
+ table 0.25, 0.25, 0.25, 0.25;
+}
+probability (agent_assistance | game_state, attempt) {
+(beg, att_1) 0.2857142857142857, 0.35714285714285715, 0.14285714285714285, 0.07142857142857142, 0.07142857142857142, 0.07142857142857142;
+(beg, att_2) 0.07142857142857142, 0.21428571428571427, 0.35714285714285715, 0.21428571428571427, 0.07142857142857142, 0.07142857142857142;
+(beg, att_3) 0.06666666666666667, 0.13333333333333333, 0.13333333333333333, 0.3333333333333333, 0.26666666666666666, 0.06666666666666667;
+(beg, att_4) 0.058823529411764705, 0.058823529411764705, 0.11764705882352941, 0.29411764705882354, 0.23529411764705882, 0.23529411764705882;
+(mid, att_1) 0.2857142857142857, 0.35714285714285715, 0.14285714285714285, 0.07142857142857142, 0.07142857142857142, 0.07142857142857142;
+(mid, att_2) 0.07142857142857142, 0.21428571428571427, 0.35714285714285715, 0.21428571428571427, 0.07142857142857142, 0.07142857142857142;
+(mid, att_3) 0.06666666666666667, 0.13333333333333333, 0.13333333333333333, 0.3333333333333333, 0.26666666666666666, 0.06666666666666667;
+(mid, att_4) 0.058823529411764705, 0.058823529411764705, 0.11764705882352941, 0.29411764705882354, 0.23529411764705882, 0.23529411764705882;
+(end, att_1) 0.2857142857142857, 0.35714285714285715, 0.14285714285714285, 0.07142857142857142, 0.07142857142857142, 0.07142857142857142;
+(end, att_2) 0.07142857142857142, 0.21428571428571427, 0.35714285714285715, 0.21428571428571427, 0.07142857142857142, 0.07142857142857142;
+(end, att_3) 0.06666666666666667, 0.13333333333333333, 0.13333333333333333, 0.3333333333333333, 0.26666666666666666, 0.06666666666666667;
+(end, att_4) 0.058823529411764705, 0.058823529411764705, 0.11764705882352941, 0.29411764705882354, 0.23529411764705882, 0.23529411764705882;
+}
\ No newline at end of file
diff --git a/bn_models/agent_model_template.bif b/bn_models/agent_model_template.bif
new file mode 100644
index 0000000..1ca11f0
--- /dev/null
+++ b/bn_models/agent_model_template.bif
@@ -0,0 +1,26 @@
+network agent_assistive_model {
+}
+
+%VARIABLES DEFINITION
+
+variable agent_assistance {
+ type discrete [ 6 ] { lev_0, lev_1, lev_2, lev_3, lev_4, lev_5};
+}
+variable attempt {
+ type discrete [ 4 ] { att_1, att_2, att_3, att_4 };
+}
+variable game_state {
+ type discrete [ 3 ] { beg, mid, end };
+}
+
+
+%INDIVIDUAL PROBABILITIES DEFINITION
+probability ( agent_assistance ) {
+ table 0.17, 0.17, 0.17, 0.17, 0.16, 0.16;
+}
+probability ( game_state ) {
+ table 0.34, 0.33, 0.33;
+}
+probability ( attempt ) {
+ table 0.25, 0.25, 0.25, 0.25;
+}
diff --git a/bn_persona_model/persona_model_test.bif b/bn_models/persona_model_template.bif
similarity index 58%
rename from bn_persona_model/persona_model_test.bif
rename to bn_models/persona_model_template.bif
index 7db034e..d8227ea 100644
--- a/bn_persona_model/persona_model_test.bif
+++ b/bn_models/persona_model_template.bif
@@ -31,21 +31,3 @@ probability ( attempt ) {
probability ( user_action ) {
table 0.33, 0.33, 0.34;
}
-probability (game_state | user_action) {
- (correct) 0.30, 0.30, 0.4;
- (wrong) 0.35, 0.35, 0.3;
- (timeout) 0.33, 0.33, 0.34;
-}
-probability (attempt | user_action) {
- (correct) 0.25, 0.25, 0.25, 0.25;
- (wrong) 0.4, 0.3, 0.2, 0.1;
- (timeout) 0.4, 0.3, 0.2, 0.1;
-}
-probability (user_action | agent_assistance) {
-(lev_0) 0.05 0.85 0.1;
-(lev_1) 0.1 0.8 0.1;
-(lev_2) 0.2 0.7 0.1;
-(lev_3) 0.33 0.57 0.1;
-(lev_4) 0.9 0.1 0.0;
-(lev_5) 1.0 0.0 0.0;
-}
diff --git a/bn_models/user_model_id_1_True.bif b/bn_models/user_model_id_1_True.bif
new file mode 100644
index 0000000..3f1ae72
--- /dev/null
+++ b/bn_models/user_model_id_1_True.bif
@@ -0,0 +1,51 @@
+network persona_model {
+}
+
+%VARIABLES DEFINITION
+
+variable agent_assistance {
+ type discrete [ 6 ] { lev_0, lev_1, lev_2, lev_3, lev_4, lev_5 };
+}
+variable attempt {
+ type discrete [ 4 ] { att_1, att_2, att_3, att_4 };
+}
+variable game_state {
+ type discrete [ 3 ] { beg, mid, end };
+}
+
+variable user_action {
+ type discrete [ 3 ] { correct, wrong, timeout };
+}
+
+%INDIVIDUAL PROBABILITIES DEFINITION
+
+probability ( agent_assistance ) {
+ table 0.17, 0.16, 0.16, 0.17, 0.17, 0.17;
+}
+probability ( game_state) {
+ table 0.34, 0.33, 0.33;
+}
+probability ( attempt ) {
+ table 0.25, 0.25, 0.25, 0.25;
+}
+probability ( user_action ) {
+ table 0.33, 0.33, 0.34;
+}
+probability (game_state | user_action) {
+(correct) 0.2222222222222222,0.3333333333333333,0.4444444444444444;
+(wrong) 0.5,0.3333333333333333,0.16666666666666666;
+(timeout) 0.5,0.3333333333333333,0.16666666666666666;
+}
+probability (attempt | user_action) {
+(correct) 0.15384615384615385,0.23076923076923078,0.3076923076923077,0.3076923076923077;
+(wrong) 0.42857142857142855,0.2857142857142857,0.14285714285714285,0.14285714285714285;
+(timeout) 0.42857142857142855,0.2857142857142857,0.14285714285714285,0.14285714285714285;
+}
+probability (user_action | agent_assistance) {
+(lev_0) 0.4,0.3,0.3;
+(lev_1) 0.6,0.2,0.2;
+(lev_2) 0.6,0.2,0.2;
+(lev_3) 0.8,0.1,0.1;
+(lev_4) 1.0,0.0,0.0;
+(lev_5) 1.0,0.0,0.0;
+}
\ No newline at end of file
diff --git a/bn_other_user_model/user_model.bif b/bn_other_user_model/user_model.bif
deleted file mode 100644
index 79daa7f..0000000
--- a/bn_other_user_model/user_model.bif
+++ /dev/null
@@ -1,103 +0,0 @@
-network persona_model_4 {
-}
-
-%VARIABLES DEFINITION
-variable reactivity {
- type discrete [3] {slow, medium, fast};
-}
-variable memory {
- type discrete[3] {low, medium, high};
-}
-variable attention {
- type discrete[3] {low, medium, high};
-}
-variable robot_assistance {
- type discrete [ 5 ] { lev_0, lev_1, lev_2, lev_3, lev_4 };
-}
-variable attempt {
- type discrete [ 4 ] { att_1, att_2, att_3, att_4 };
-}
-variable game_state {
- type discrete [ 3 ] { beg, mid, end };
-}
-variable robot_feedback {
- type discrete [ 2 ] { yes, no };
-}
-variable user_action {
- type discrete [ 3 ] { correct, wrong, timeout };
-}
-
-%INDIVIDUAL PROBABILITIES DEFINITION
-probability ( robot_assistance ) {
- table 0.2, 0.2, 0.2, 0.2, 0.2;
-}
-probability ( game_state ) {
- table 0.34, 0.33, 0.33;
-}
-probability ( attempt ) {
- table 0.25, 0.25, 0.25, 0.25;
-}
-probability ( user_action ) {
- table 0.33, 0.33, 0.34;
-}
-#CPDS 4 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( reactivity ) {
- table 0.34, 0.33, 0.33;
-}
-#CPDS 3 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( memory ) {
- table 0.33, 0.33, 0.34;
-}
-#CPDS 1 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( attention ) {
- table 0.33, 0.33, 0.34;
-}
-probability ( robot_feedback ) {
- table 0.5, 0.5;
-}
-probability ( reactivity | attention ) {
- (low) 0.5, 0.4, 0.1;
- (medium) 0.3, 0.5, 0.2;
- (high) 0.1, 0.2, 0.7;
-}
-#CPDS 7
-probability (user_action | memory, reactivity) {
-(low, slow) 0.2, 0.5, 0.3;
-(low, medium) 0.3, 0.5, 0.2;
-(low, fast) 0.4, 0.5, 0.1;
-(medium, slow) 0.5, 0.3, 0.2;
-(medium, medium) 0.55, 0.35, 0.1;
-(medium, fast) 0.6, 0.4, 0.0;
-(high, slow) 0.5, 0.4, 0.1;
-(high, medium) 0.6, 0.3, 0.1;
-(high, fast) 0.8, 0.2, 0.0;
-}
-#CPDS 5
-probability (robot_feedback | user_action) {
- (correct) 0.5, 0.5;
- (wrong) 0.5, 0.5;
- (timeout) 0.5, 0.5;
-}
-#CPDS 6
-probability (robot_assistance | user_action) {
- (correct) 0.05 0.1 0.15 0.3 0.4;
- (wrong) 0.4 0.2 0.2 0.1 0.1;
- (timeout) 0.4 0.2 0.2 0.1 0.1;
-}
-#CPDS 2
-probability (game_state | user_action) {
- (correct) 0.2, 0.4, 0.4;
- (wrong) 0.4, 0.4, 0.2;
- (timeout) 0.6, 0.3, 0.1;
-}
-#CPDS 0
-probability (attempt | user_action) {
- (correct) 0.1, 0.2, 0.3, 0.4;
- (wrong) 0.7, 0.2, 0.1, 0.0;
- (timeout) 0.6, 0.3, 0.1, 0.0;
-}
-#CPDS 5
-probability (robot_assistance | robot_feedback) {
- (yes) 0.5 0.3 0.1 0.1 0.0;
- (no) 0.0 0.1 0.1 0.3 0.5;
-}
\ No newline at end of file
diff --git a/bn_persona_model/__pycache__/utilities.cpython-36.pyc b/bn_persona_model/__pycache__/utilities.cpython-36.pyc
deleted file mode 100644
index 9ae85612e5c19ba5c07e4ca739b2c68190e8318e..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001
literal 2460
zcmZ`*UvJws5GN(smfgfjv-B^DLc@Ro#*C!vh7JgdAWhp>qgytha4`t7NV}RPDkCXx
zDU_F`K>M<%z3vO_o9r|2bx-{S1NyW(O0gX$lc76`Jd$_3-~H&X*4I1FZ+sj6-6iB7
zvh+C6K7dz!4T2FyQ!=bH9a82n_bnMZ%wrAEE^D$DXpi}<4Z6YT5$Sb4z)UiTaV30Q
zc)ePyeQ0R%o}5vT^o&kjd6!YOr|EB01n<Z@I(F<XzIH0RGx`jyPo2^^0zY7VZ}8^e
z``0{}MKXy-DOZ7rG9HK`8*!<EL<KAvkGbSR2V<FMfxcioi_(I}6#T#koovAS3B2kq
zh=i2n$E{QPo|g0uIEwNQ=V$MT{6!>M_Z;KwJTNqYMtAN*h(4H&#kbwYPWbqA_Xh=@
z?M9JcGT~_&?LN!nA_E82?$ey{-NQln0xWV7?T$ovTsI4yq=`<r+BrV?*m=y7SjTyG
zyx&G&(Um57)0=P|vJ==niNpEsJCD;mj#9P1vzXO$jhk{|x=$2%qmx`bmok^8AtN#2
zJy+r}$R-GL{nCPU<O>_wLc<$Z?v4v79|WPBwBy{Pavj<+E{zXrG<gcIdJCciluaFZ
zpJA2(W8aWdTDn%d;w6kaFnUk|_NK&Yv~Wt#T6p3SSa@JDsx44g77eQ#V*9F<OIkXm
zTY6<<+yGQHqE~>4P5=*rYg~vJ&nUnXFEoH`lFYcUSc(g2!KHMhc>qZbqRKWO+wlU<
zw{Kg5|EdFPW*2@UCu+~Oa2qrWtC@SjgZ`KOhw~vFscFb(c?uB7#a=MtlYWpTtRHZ}
z=2meIe#AP=@)-}~ybzkp#qyEXJUiBx2Y!*cnEcNqon$;ORwjMNbs@!KWvr)qI~5ei
z3{C-$0Ni1Tm;^eXpF$m;^;lxbeXs%<;R6D=f>qO8TRIEVy%@dOSc$}JUlq97x*V6;
zSc%8(R|$*Xtvb6QCMvXv+5orIJrHEWrylhIz?%!{Vvq00TQDA5h(iPG@fy4q(4Rv?
zlb<MpdIB6`A?B2gvRSs6^Jj~>zq_Y?>02U`*TK5iGMh3V<vNU%O27#bbJMDWHk}C&
zOXZsBR%T<)*(i#qrkU|9mnWu;#sZ;YI+7!#W{ExlernGd!T5ZZFb+#Y_ZpDQIlz^<
zjYXUY%a}0_kut0OdTq0A69eWM&{S$}t@>Ff*xb12S=}v>Rn68Sf^#s|39sGnR@tA&
zuwlO!ftc${@KE6~?6sQ6ZH!^b${Ia{-PJY-;xr+Z&eCs<$}gaQehnZeAu4SOUqS;I
z15@bSL8c+h1??iwfc7xGct@Q%rwvG~Go|tob55zaZ@)bG0(I%$BEVNKm<uBw^I$!^
z9UE`{wdHWcQ>fh_7d)tGqoM+JxsvJu8<U-Tkw=z?_iX9m2}L8kJWkFRudkP26e%cH
z;597b!vPj+7EOW0prS3C1jr`+z!p>8u8YWSeH9lZuwcb}$_8cG0MTe?40aAH>KIrC
zkno3t$W0Ic#3tN5EYD1<5Xm@~S+6-<TMlQ~uD-&A4R4-TecfeodN>$vS9QPAd-J^V
zSNdC0Ft{hvD*>c*9>0z>+@#g41MpJa20_}k?xFJ69G^B_pRR%QsN8|EUQ_l#$xl(>
mO=sD`;>(UNZDl#nGLQclDym5NJ`$qpfB;@`{f_^c-~A7Jf3Xe#
diff --git a/bn_persona_model/persona_model.bif b/bn_persona_model/persona_model.bif
deleted file mode 100644
index ef80625..0000000
--- a/bn_persona_model/persona_model.bif
+++ /dev/null
@@ -1,103 +0,0 @@
-network persona_model {
-}
-
-%VARIABLES DEFINITION
-variable reactivity {
- type discrete [3] {slow, medium, fast};
-}
-variable memory {
- type discrete[3] {low, medium, high};
-}
-variable attention {
- type discrete[3] {low, medium, high};
-}
-variable robot_assistance {
- type discrete [ 5 ] { lev_0, lev_1, lev_2, lev_3, lev_4 };
-}
-variable attempt {
- type discrete [ 4 ] { att_1, att_2, att_3, att_4 };
-}
-variable game_state {
- type discrete [ 3 ] { beg, mid, end };
-}
-variable robot_feedback {
- type discrete [ 2 ] { yes, no };
-}
-variable user_action {
- type discrete [ 3 ] { correct, wrong, timeout };
-}
-
-%INDIVIDUAL PROBABILITIES DEFINITION
-probability ( robot_assistance ) {
- table 0.2, 0.2, 0.2, 0.2, 0.2;
-}
-probability ( game_state ) {
- table 0.34, 0.33, 0.33;
-}
-probability ( attempt ) {
- table 0.25, 0.25, 0.25, 0.25;
-}
-probability ( user_action ) {
- table 0.33, 0.33, 0.34;
-}
-#CPDS 4 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( reactivity ) {
- table 0.34, 0.33, 0.33;
-}
-#CPDS 3 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( memory ) {
- table 0.33, 0.33, 0.34;
-}
-#CPDS 1 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( attention ) {
- table 0.33, 0.33, 0.34;
-}
-probability ( robot_feedback ) {
- table 0.5, 0.5;
-}
-probability ( reactivity | attention ) {
- (low) 0.7, 0.2, 0.1;
- (medium) 0.2, 0.5, 0.3;
- (high) 0.1, 0.2, 0.7;
-}
-#CPDS 7
-probability (user_action | memory, reactivity) {
-(low, slow) 0.2, 0.5, 0.3;
-(low, medium) 0.3, 0.5, 0.2;
-(low, fast) 0.4, 0.5, 0.1;
-(medium, slow) 0.5, 0.3, 0.2;
-(medium, medium) 0.55, 0.35, 0.1;
-(medium, fast) 0.6, 0.4, 0.0;
-(high, slow) 0.5, 0.4, 0.1;
-(high, medium) 0.6, 0.3, 0.1;
-(high, fast) 0.8, 0.2, 0.0;
-}
-#CPDS 5
-probability (robot_feedback | user_action) {
- (correct) 0.8, 0.2;
- (wrong) 0.5, 0.5;
- (timeout) 0.2, 0.8;
-}
-#CPDS 6
-probability (robot_assistance | user_action) {
- (correct) 0.05 0.1 0.15 0.3 0.4;
- (wrong) 0.4 0.3 0.2 0.05 0.05;
- (timeout) 0.4 0.4 0.1 0.05 0.05;
-}
-#CPDS 2
-probability (game_state | user_action) {
- (correct) 0.2, 0.4, 0.4;
- (wrong) 0.4, 0.4, 0.2;
- (timeout) 0.6, 0.3, 0.1;
-}
-#CPDS 0
-probability (attempt | user_action) {
- (correct) 0.1, 0.2, 0.3, 0.4;
- (wrong) 0.5, 0.3, 0.15, 0.05;
- (timeout) 0.4, 0.3, 0.2, 0.1;
-}
-#CPDS 5
-probability (robot_assistance | robot_feedback) {
- (yes) 0.5 0.3 0.1 0.1 0.0;
- (no) 0.0 0.1 0.1 0.3 0.5;
-}
\ No newline at end of file
diff --git a/bn_persona_model/user_action_model.bif b/bn_persona_model/user_action_model.bif
deleted file mode 100644
index 24c07dd..0000000
--- a/bn_persona_model/user_action_model.bif
+++ /dev/null
@@ -1,84 +0,0 @@
-network persona_model {
-}
-
-%VARIABLES DEFINITION
-variable user_reactivity {
- type discrete [3] {slow, medium, fast};
-}
-variable user_memory {
- type discrete[3] {low, medium, high};
-}
-variable agent_assistance {
- type discrete [ 6 ] { lev_0, lev_1, lev_2, lev_3, lev_4, lev_5 };
-}
-variable attempt {
- type discrete [ 4 ] { att_1, att_2, att_3, att_4 };
-}
-variable game_state {
- type discrete [ 3 ] { beg, mid, end };
-}
-variable agent_feedback {
- type discrete [ 2 ] { yes, no };
-}
-variable user_action {
- type discrete [ 3 ] { correct, wrong, timeout };
-}
-
-
-%INDIVIDUAL PROBABILITIES DEFINITION
-probability ( agent_assistance ) {
- table 0.16, 0.16, 0.17, 0.16, 0.17, 0.16;
-}
-probability ( game_state ) {
- table 0.34, 0.33, 0.33;
-}
-probability ( attempt ) {
- table 0.25, 0.25, 0.25, 0.25;
-}
-probability ( user_action ) {
- table 0.33, 0.33, 0.34;
-}
-#CPDS 4 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( user_reactivity ) {
- table 0.34, 0.33, 0.33;
-}
-#CPDS 3 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( user_memory ) {
- table 0.33, 0.33, 0.34;
-}
-probability ( agent_feedback ) {
- table 0.5, 0.5;
-}
-
-
-probability (user_action | user_memory, user_reactivity) {
-(low, slow) 0.2, 0.2, 0.6;
-(low, medium) 0.3, 0.2, 0.5;
-(low, fast) 0.4, 0.4, 0.2;
-(medium, slow) 0.3, 0.1, 0.6;
-(medium, medium) 0.4, 0.2, 0.4;
-(medium, fast) 0.6, 0.2, 0.2;
-(high, slow) 0.7, 0.1, 0.2;
-(high, medium) 0.8, 0.1, 0.1;
-(high, fast) 0.8, 0.2, 0.0;
-}
-probability (agent_feedback | user_action) {
- (correct) 0.8, 0.2;
- (wrong) 0.5, 0.5;
- (timeout) 0.2, 0.8;
-}
-probability (agent_assistance | user_action) {
- (correct) 0.05 0.05 0.1 0.1 0.35 0.35;
- (wrong) 0.45 0.3 0.2 0.05 0.0 0.0;
- (timeout) 0.4 0.4 0.2 0.0 0.0 0.0;
-}
-probability (game_state | user_action) {
- (correct) 0.2, 0.4, 0.4;
- (wrong) 0.4, 0.4, 0.2;
- (timeout) 0.6, 0.3, 0.1;
-}
-probability (attempt | user_action) {
- (correct) 0.1, 0.2, 0.3, 0.4;
- (wrong) 0.5, 0.3, 0.15, 0.05;
- (timeout) 0.4, 0.3, 0.2, 0.1;
-}
\ No newline at end of file
diff --git a/bn_persona_model/user_react_time_model.bif b/bn_persona_model/user_react_time_model.bif
deleted file mode 100644
index 1c2bc5d..0000000
--- a/bn_persona_model/user_react_time_model.bif
+++ /dev/null
@@ -1,95 +0,0 @@
-network persona_model {
-}
-
-%VARIABLES DEFINITION
-variable reactivity {
- type discrete [3] {slow, medium, fast};
-}
-variable memory {
- type discrete[3] {low, medium, high};
-}
-variable attention {
- type discrete[3] {low, medium, high};
-}
-variable agent_assistance {
- type discrete [ 6 ] { lev_0, lev_1, lev_2, lev_3, lev_4, lev_5 };
-}
-variable attempt {
- type discrete [ 4 ] { att_1, att_2, att_3, att_4 };
-}
-variable game_state {
- type discrete [ 3 ] { beg, mid, end };
-}
-variable agent_feedback {
- type discrete [ 2 ] { yes, no };
-}
-variable user_react_time {
- type discrete [ 3 ] { slow, normal, fast};
-}
-
-%INDIVIDUAL PROBABILITIES DEFINITION
-variable agent_assistance {
- type discrete [ 6 ] { lev_0, lev_1, lev_2, lev_3, lev_4, lev_5 };
-}
-probability ( game_state ) {
- table 0.34, 0.33, 0.33;
-}
-probability ( attempt ) {
- table 0.25, 0.25, 0.25, 0.25;
-}
-#CPDS 4 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( reactivity ) {
- table 0.34, 0.33, 0.33;
-}
-#CPDS 3 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( memory ) {
- table 0.33, 0.33, 0.34;
-}
-#CPDS 1 #SPECIFICALLY FOR THE GIVEN PATIENT
-probability ( attention ) {
- table 0.33, 0.33, 0.34;
-}
-probability ( agent_feedback ) {
- table 0.5, 0.5;
-}
-probability (user_react_time) {
- table 0.33, 0.33, 0.34
-}
-probability ( reactivity | attention ) {
- (low) 0.7, 0.2, 0.1;
- (medium) 0.2, 0.5, 0.3;
- (high) 0.1, 0.2, 0.7;
-}
-
-
-probability (user_react_time | memory, reactivity) {
-(low, slow) 0.2, 0.5, 0.3;
-(low, medium) 0.3, 0.5, 0.2;
-(low, fast) 0.4, 0.5, 0.1;
-(medium, slow) 0.5, 0.3, 0.2;
-(medium, medium) 0.55, 0.35, 0.1;
-(medium, fast) 0.6, 0.4, 0.0;
-(high, slow) 0.5, 0.4, 0.1;
-(high, medium) 0.6, 0.3, 0.1;
-(high, fast) 0.8, 0.2, 0.0;
-}
-probability (agent_feedback | user_react_time) {
- (slow) 0.8, 0.2;
- (normal) 0.5, 0.5;
- (fast) 0.2, 0.8;
-}
-probability (agent_assistance | user_react_time) {
- (slow) 0.05 0.05 0.1 0.1 0.35 0.35;
- (normal) 0.45 0.3 0.2 0.05 0.0 0.0;
- (fast) 0.4 0.4 0.2 0.0 0.0 0.0;
-}
-probability (game_state | user_react_time) {
- (slow) 0.2, 0.4, 0.4;
- (normal) 0.4, 0.4, 0.2;
- (fast) 0.6, 0.3, 0.1;
-}
-probability (attempt | user_react_time) {
- (slow) 0.1, 0.2, 0.3, 0.4;
- (normal) 0.5, 0.3, 0.15, 0.05;
- (fast) 0.4, 0.3, 0.2, 0.1;
-}
diff --git a/bn_persona_model/utilities.py b/bn_persona_model/utilities.py
deleted file mode 100644
index e44dbec..0000000
--- a/bn_persona_model/utilities.py
+++ /dev/null
@@ -1,90 +0,0 @@
-import random
-import bn_functions
-
-
-def get_dynamic_variables(variables_name, variables_value):
- if len(variables_name)!=len(variables_value):
- assert "The variables name numbers is different from the variables value"
- else:
- dynamic_variables = {variables_name[i]:variables_value[i] for i in range(len(variables_name))}
- return dynamic_variables
-
-
-
-
-def compute_next_state(user_action, task_evolution, attempt_counter, correct_move_counter,
- wrong_move_counter, timeout_counter
- ):
- '''
- The function computes given the current state and action of the user, the next state
- Args:
- user_action: 0,1,2
- task_evolution: beg, mid, end
- correct_move_counter:
- attempt_counter:
- wrong_move_counter:
- timeout_counter:
- Return:
- the counters updated according to the user_action
- '''
- if user_action == 0:
- attempt_counter = 0
- task_evolution += 1
- correct_move_counter += 1
- # if the user made a wrong move and still did not reach the maximum number of attempts
- elif user_action == 1 and attempt_counter < 3:
- attempt_counter += 1
- wrong_move_counter += 1
- # if the user did not move any token and still did not reach the maximum number of attempts
- elif user_action == 2 and attempt_counter < 3:
- attempt_counter += 1
- timeout_counter += 1
- # the robot or therapist makes the correct move on the patient's behalf
- else:
- attempt_counter = 0
- task_evolution += 1
- correct_move_counter += 1
-
- return task_evolution, attempt_counter, correct_move_counter, wrong_move_counter, timeout_counter
-
-def get_user_action_prob(user_bn_model, robot_assistance_action, robot_feedback_action,
- attempt_counter, game_state_counter, user_memory, user_attention, user_reactivity):
- user_actions_prob = bn_functions.get_inference_from_state(user_bn_model,
- variables=['user_action'],
- evidence={'robot_assistance': robot_assistance_action,
- 'attempt': attempt_counter,
- 'game_state': game_state_counter,
- 'robot_feedback': robot_feedback_action,
- 'memory': user_memory,
- 'attention': user_attention,
- 'reactivity': user_reactivity})
- return user_actions_prob
-
-
-def get_stochatic_action(actions_prob):
- '''
- Select one of the actions according to the actions_prob
- Args:
- actions_prob: the probability of the Persona based on the BN to make a correct move, wrong move, timeout
- Return:
- the id of the selected action
- N.B:
- '''
- action_id = None
- correct_action_from_BN = actions_prob[0]
- wrong_action_from_BN = actions_prob[1]
- timeout_action_from_BN = actions_prob[2]
-
- rnd_val = random.uniform(0,1)
- #if user_prob is lower than the correct action prob then is the correct one
- if rnd_val<=correct_action_from_BN:
- action_id = 0
- #if rnd is larger than the correct action prob and lower than wrong
- # action prob then is the wrong one
- elif rnd_val>correct_action_from_BN \
- and rnd_val<(correct_action_from_BN+wrong_action_from_BN):
- action_id = 1
- #timeout
- else:
- action_id = 2
- return action_id
diff --git a/bn_robot_model/__pycache__/utilities.cpython-36.pyc b/bn_robot_model/__pycache__/utilities.cpython-36.pyc
deleted file mode 100644
index 28581fb5fa7c725dab25c9a4c7b5ca743e6f6c3a..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001
literal 3939
zcmaJ^UvJw+66aruRutQp25Ft91$$_V6E2AJ=dK3`MVq(*`sNGLJD?B_SWw!PLYpGh
zCFRtC{FL;upP>B${Yv}VCx3-L^*6huNYV09<}AsXo!yz8ncw)y&6~a3NB?HOd|+9>
zTaC-b^HbdUV-Rj}JF!M)v`5aj7I(P&&f@M{XXJ8^`|qrg$2+`>(dYJe)*!gVJnMKu
zw7MKTJjI>=4MbQYn_DA?+t8r3L4(ocmzp{}fTkT4938Wl**GiS{Pdq!aUSBYcqPI~
z66N$16=4yb#_24K(;@>|Eb#>blIE~^#zpdA98bfUNQI0Fp{Mskx_F!yG9E9BI7`C?
zCPd>niHjIA(@fS{!e~Mwxt=v4L+GNB?r(7C`yhpNVz2DC_R3-ox3GEpc;Kq;I88((
zQ`L*-i%b^m`Ll183o>w&my2XN+HY)?P2)uTX2~w}{qgMZzqye4VHBlY#v)0g!{^y#
zITvY>AASSN4&j>@kjT>LaGWx&=Md%?BOWZyRlr!B#sy<Lv`f8p=gnU#mp-VtlMdK)
z1?8=cqLU({PMnp^?U}vuSKi85;lzIRxcjqv(OGp)+>7q23uC?LN87TDW8vFpa+VL*
zZ{%&-M;s+=vf%kJB)aM44dEpWXLB*n<oPh1%4{A!3URzot_n#(r*jq;|3`4Rh-5Sm
zXVF}+yg(4JNw!Q2A%|f&7PI@|Jm&X9k@5}o<*cYE#C%cIQ|j&fV=ezu6ib<I8mt4T
z4y*Gpn>Jz!)+~fXoXv-@tz`W(OvIT;Or(+1^z}L}UWEuq8Z(4<6i$WU<7jfaUSA!U
zAVN|gsH1cuSS2&?RJXK2^=cQWj&ZN*)gDAh?wmz3j>d_|RUpn{E}&}QNf)7@cH(p@
zBz;ltG%i%<EJ~Il9|V%ZXVh<=u)5JWLbaz~>V?#;dLxx%)n3cf%Ku?0<oW2M(lnEV
zj6BNd49rpLR565tSG_kAMO)@382K)UwbyqXn|^(#W52n#PJ}I!)`?EJNcQWT_#O}3
zzMzb74y+3oql?kQ=wb9JU#OzJbtpSl?x`)m;Le4Ueyra-`GUrk_r7&%$rs!O2bB-{
z9ORH6{k54x--$#_iZDw>6(40GLP45jl2g?~7*Vw=8QSo5P+ZNyrr7Yh_<Tu~GaN^`
z;8<Hr9Uap`s8+50OtU1(Uei8G&6^qUXh8AI9<Fz^Elp_qF;+6kVb;LPM&Wwq@q?pU
z_pyYVtV>O(rXg-S5=AnbtL`$5DX9is9h9oyG+FJKHx_es2Og9#hn1gnL>(Qg+plGo
z&e~t^6!Baj_uAjPGUW_u(2l}=3aDn>evWPCz*&??&kpG4^qs)&xPjdV_3SqX8^PF|
zdpzh;Ay27%G1X&)YRiP4`;7gtjFOVmWsGMN#v~Pcc|hb-A{4+4LTMt&Z9eW?SA47=
zc)@}GI-$>WQPz04b=>#i?7|{=qZnUV7&{nUj9rWl#-OkZ=LE6M?E~}(2i6Ht0@Md2
z`pAc!EqvsR=6AsFZQ=VRzYG42E&NW&55Vv1)pxL3YX{8NoL-BAwM)*876+@AoPLXg
z^%zhPx!R5XSLNzvo?WeMC4xPq9C~~W;IwLA6ZC47DT88N<C}`xQdRjMh7Sol9)(;i
zfN;7~nuaxmgs-9;dlI?>qJm$8An=J=3x?K0_fe)|Xyh(5L{v$uxn$jP!GW$iY1^r;
zn>GUda;g#K8q{=+NyY21gk^(&%A=)}ueIxpmmeY=F9~ceDM$w$)t_W@bbLY=J|z1Y
zdc0Zr=A;X?Q=l&7iwy0=05jx!H0`FmL(lJnsIvL0aZ~LYFIWB>?YG@9x~Cy1!&<C5
zHprBvVRTS)btp7gAJwc%J!`N(soC{@v_K_2`m}VLLAdL-@X1xF_EtW-DpD=d+|uBx
z*fm>g6&Q8P%+&23GiUxB4%h9!wWli}YK~)nV0Y}hxXqKM?p=Jr&yi0sW8It5xzn-h
zdh`wtD<JWyBR{V2_+QqAP2DL*H~k)52i&JCpfRwPM&dZA4m5h{7m)BEF)k%&G!niZ
z{q*)W%`2OFV06O%)Yq8pY_9{l<uTxb9K8SN{^K>66;Xc5#95Z;e?5lRl!OYP<~Zi6
zNvbihioB|mT`M>EwS7{BVjGjTW~%9r^D?6ej$0i^!#usTNzl7EmP;2zr>c$#WU1)|
zjW5*hru8NbZ*MbJ-MVUv+H2WxCGM(<kcuA*(;{ZrPks-?+6`<%-=4EyQ^G!=u^fUw
t=$KB|PnYw>xje)tjfpxCwM1h%4Xe8648Amo{1tU122t(6-TgZF`=3~pOHu#;
diff --git a/simulation.py b/simulation.py
index 27e1eb7..15bd225 100644
--- a/simulation.py
+++ b/simulation.py
@@ -3,6 +3,7 @@ import os
import bnlearn
import numpy as np
import random
+import copy
#import classes and modules
from bn_variables import Agent_Assistance, Agent_Feedback, User_Action, User_React_time, Game_State, Attempt
import bn_functions
@@ -156,10 +157,6 @@ def compute_next_state(user_action, task_progress_counter, attempt_counter, corr
task_progress_counter +=1
print("Reach the end of the episode")
- # TODO call the function to compute the state of the game (beg, mid, end)
-
-
-
next_state = (game_state_counter, attempt_counter, user_action)
@@ -177,20 +174,19 @@ def select_agent_action(agent_action, epsilon):
if random.random()>epsilon:
return np.argmax(agent_action)
else:
- agent_action[np.argmax(agent_action)]=0
- return np.argmax(agent_action)
+ agent_action_rm_best = agent_action[:]
+ agent_action_rm_best[np.argmax(agent_action)] = 0
+ return np.argmax(agent_action_rm_best)
def simulation(bn_model_user_action,
bn_model_agent_behaviour,
var_user_action_target_action,
var_agent_behaviour_target_action,
game_state_bn_name, attempt_bn_name,
- agent_assistance_bn_name, agent_feedback_bn_name,
- user_pref_assistance,
- agent_behaviour,
+ agent_assistance_bn_name,
agent_policy,
state_space, action_space,
- epochs=50, task_complexity=5, max_attempt_per_object=4, alpha_learning=0):
+ epoch=50, run = 50, task_complexity=5, max_attempt_per_object=4, alpha_learning=0):
'''
Args:
@@ -201,170 +197,174 @@ def simulation(bn_model_user_action,
'''
- # user_action_per_robot_feedback_robot_assistance = [[[0 for i in range(User_Action.counter.value)]
- # for j in range(Agent_Assistance.counter.value)]
- # for l in range(Agent_Feedback.counter.value)
- # ]
- # attempt_counter_per_user_action = [[0 for i in range(Attempt.counter.value)] for j in
- # range(User_Action.counter.value)]
- # game_state_counter_per_user_action = [[0 for i in range(Game_State.counter.value)] for j in
- # range(User_Action.counter.value)]
+ user_action_per_agent_assistance = [[0 for i in range(User_Action.counter.value)]
+ for j in range(Agent_Assistance.counter.value)]
+ attempt_counter_per_user_action = [[0 for i in range(Attempt.counter.value)] for j in
+ range(User_Action.counter.value)]
+ game_state_counter_per_user_action = [[0 for i in range(Game_State.counter.value)] for j in
+ range(User_Action.counter.value)]
#output variables:
- n_correct_per_episode = [0]*epochs
- n_wrong_per_episode = [0]*epochs
- n_timeout_per_episode = [0]*epochs
- n_max_attempt_per_episode = [0]*epochs
- game_performance_episode = [0]*epochs
- n_assistance_lev_per_episode = [[0 for i in range(Agent_Assistance.counter.value)] for j in range(epochs)]
- n_feedback_per_episode = [[0 for i in range(Agent_Feedback.counter.value)] for j in range(epochs)]
- n_react_time_per_episode = [[0 for i in range(User_React_time.counter.value)] for j in range(epochs)]
+ n_correct_per_episode_epoch = [0]*epoch
+ n_wrong_per_episode_epoch = [0]*epoch
+ n_timeout_per_episode_epoch = [0]*epoch
+ n_max_attempt_per_episode_epoch = [0]*epoch
+ game_performance_episode_epoch = [0]*epoch
+ n_assistance_lev_per_episode_epoch = [[0 for i in range(Agent_Assistance.counter.value)] for j in range(epoch)]
#data structure to memorise a sequence of episode
episodes = []
ep = Episode()
+ bn_model_user_action_ref = copy.deepcopy(bn_model_user_action)
- for e in range(epochs):
+ for e in range(epoch):
print("##########################################################")
- print("EPISODE ",e)
+ print("EPISODE ", e)
print("##########################################################")
-
- '''Simulation framework'''
- #counters
- game_state_counter = 0
- attempt_counter = 1
- iter_counter = 0
- correct_move_counter = 0
- wrong_move_counter = 0
- timeout_counter = 0
- max_attempt_counter = 0
-
- #The following variables are used to update the BN at the end of the episode
- # user_action_dynamic_variables = {
- # 'attempt': attempt_counter_per_user_action,
- # 'game_state': game_state_counter_per_user_action,
- # 'user_action': user_action_per_robot_feedback_robot_assistance
- # }
-
- #data structure to memorise the sequence of states (state, action, next_state)
- episode = []
- selected_user_action = 0
- task_progress_counter = 0
- #####################SIMULATE ONE EPISODE#########################################
- while(task_progress_counter<=task_complexity):
-
- current_state = (game_state_counter, attempt_counter, selected_user_action)
- current_state_index = ep.state_from_point_to_index(state_space, current_state)
- # if agent_policy==[]:
- # selected_agent_feedback_action, selected_agent_assistance_action = \
- # generate_agent_assistance(preferred_assistance=user_pref_assistance,
- # agent_behaviour=agent_behaviour,
- # current_state=current_state_index,
- # state_space=state_space,
- # action_space=action_space
- # )
- # else:
- # #TODO agent_policy is a list of 12 items
- # # select the one with the highest probability 1-epsilon of the times and one of the others epsilon times
- #
- # selected_agent_feedback_action, selected_agent_assistance_action = ep.state_from_index_to_point(action_space, select_agent_action(agent_policy[current_state_index], epsilon=0.1))
-
- vars_agent_evidence = {game_state_bn_name: game_state_counter,
- attempt_bn_name: attempt_counter - 1,
- }
-
- query_agent_behaviour_prob = bn_functions.infer_prob_from_state(user_bn_model=bn_model_agent_behaviour,
- infer_variable=var_agent_behaviour_target_action,
- evidence_variables=vars_agent_evidence)
-
- selected_agent_behaviour_action = bn_functions.get_stochastic_action(query_agent_behaviour_prob.values)
- #selected_agent_behaviour_action = np.argmax(query_agent_behaviour_prob.values)
-
- #counters for plots
- n_assistance_lev_per_episode[e][selected_agent_behaviour_action] += 1
- print("agent_assistance {}, attempt {}, game {}".format(selected_agent_behaviour_action, attempt_counter, game_state_counter))
-
- ##########################QUERY FOR THE USER ACTION AND REACT TIME#####################################
- #return the user action in this state based on the Persona profile
- vars_user_evidence = { game_state_bn_name: game_state_counter,
- attempt_bn_name: attempt_counter - 1,
- agent_assistance_bn_name: selected_agent_behaviour_action,
- }
-
- query_user_action_prob = bn_functions.infer_prob_from_state(user_bn_model=bn_model_user_action,
- infer_variable=var_user_action_target_action,
- evidence_variables=vars_user_evidence)
-
- selected_user_action = bn_functions.get_stochastic_action(query_user_action_prob.values)
-
- # # updates counters for simulation
- # # remap user_action index
- # if selected_user_action == 0:
- # selected_user_action = 1
- # elif selected_user_action == 1:
- # selected_user_action = -1
- # else:
- # selected_user_action = 0
-
- #updates counters for simulation
- iter_counter += 1
- next_state, task_progress_counter, game_state_counter, attempt_counter, correct_move_counter, \
- wrong_move_counter, timeout_counter, max_attempt_counter = compute_next_state(selected_user_action,
- task_progress_counter,
- attempt_counter,
- correct_move_counter, wrong_move_counter,
- timeout_counter, max_attempt_counter,
- max_attempt_per_object)
-
- # update counters
- # if game_state_counter <= 2:
- # user_action_per_robot_feedback_robot_assistance[selected_agent_feedback_action][selected_agent_assistance_action][selected_user_action] += 1
- # attempt_counter_per_user_action[selected_user_action][attempt_counter - 1] += 1
- # game_state_counter_per_user_action[selected_user_action][game_state_counter] += 1
-
- # store the (state, action, next_state)
- episode.append((ep.state_from_point_to_index(state_space, current_state),
- selected_agent_behaviour_action,
- ep.state_from_point_to_index(state_space, next_state)))
-
- print("current_state ", current_state, " next_state ", next_state)
- ####################################END of EPISODE#######################################
- print("game_state_counter {}, iter_counter {}, correct_counter {}, wrong_counter {}, "
- "timeout_counter {}, max_attempt {}".format(game_state_counter, iter_counter, correct_move_counter,
- wrong_move_counter, timeout_counter, max_attempt_counter))
-
- #save episode
- episodes.append(Episode(episode))
-
-
-
- #update user models
- # bn_model_user_action = bn_functions.update_cpds_tables(bn_model_user_action, user_action_dynamic_variables, alpha_learning)
- #
-
- #reset counter
- user_action_per_robot_feedback_robot_assistance = [[[0 for i in range(User_Action.counter.value)]
- for j in range(Agent_Assistance.counter.value)]
- for l in range(Agent_Feedback.counter.value)
- ]
- attempt_counter_per_user_action = [[0 for i in range(Attempt.counter.value)] for j in
- range(User_Action.counter.value)]
- game_state_counter_per_user_action = [[0 for i in range(Game_State.counter.value)] for j in
- range(User_Action.counter.value)]
-
- #for plots
- n_correct_per_episode[e] = correct_move_counter
- n_wrong_per_episode[e] = wrong_move_counter
- n_timeout_per_episode[e] = timeout_counter
- n_max_attempt_per_episode[e] = max_attempt_counter
- game_performance_episode[e] = [n_correct_per_episode[e],
- n_wrong_per_episode[e],
- n_timeout_per_episode[e],
- n_max_attempt_per_episode[e]]
-
-
- return game_performance_episode, n_react_time_per_episode, n_assistance_lev_per_episode, n_feedback_per_episode, episodes
+ bn_model_user_action = copy.deepcopy(bn_model_user_action_ref)
+
+ n_correct_per_episode_run = [0] * run
+ n_wrong_per_episode_run = [0] * run
+ n_timeout_per_episode_run = [0] * run
+ n_max_attempt_per_episode_run = [0] * run
+ game_performance_episode_run = [0] * run
+ n_assistance_lev_per_episode_run = [[0 for i in range(Agent_Assistance.counter.value)] for j in range(run)]
+
+ for r in range(run):
+
+ '''Simulation framework'''
+ #counters
+ game_state_counter = 0
+ attempt_counter = 1
+ iter_counter = 0
+ correct_move_counter = 0
+ wrong_move_counter = 0
+ timeout_counter = 0
+ max_attempt_counter = 0
+
+ #The following variables are used to update the BN at the end of the episode
+ user_action_dynamic_variables = {
+ 'attempt': attempt_counter_per_user_action,
+ 'game_state': game_state_counter_per_user_action,
+ 'user_action': user_action_per_agent_assistance
+ }
+
+
+
+ #data structure to memorise the sequence of states (state, action, next_state)
+ episode = []
+ selected_user_action = 0
+ task_progress_counter = 0
+ #####################SIMULATE ONE EPISODE#########################################
+ while(task_progress_counter<=task_complexity):
+
+ current_state = (game_state_counter, attempt_counter, selected_user_action)
+ current_state_index = ep.state_from_point_to_index(state_space, current_state)
+ if agent_policy==[]:
+ vars_agent_evidence = {game_state_bn_name: game_state_counter,
+ attempt_bn_name: attempt_counter - 1,
+ }
+
+ query_agent_behaviour_prob = bn_functions.infer_prob_from_state(user_bn_model=bn_model_agent_behaviour,
+ infer_variable=var_agent_behaviour_target_action,
+ evidence_variables=vars_agent_evidence)
+
+ #selected_agent_behaviour_action = bn_functions.get_stochastic_action(query_agent_behaviour_prob.values)
+ selected_agent_behaviour_action = select_agent_action(query_agent_behaviour_prob.values, epsilon=0.2)
+ else:
+ selected_agent_behaviour_action = select_agent_action(agent_policy[current_state_index], epsilon=0.2)
+ #selected_agent_behaviour_action = bn_functions.get_stochastic_action(agent_policy[current_state_index])
+ #selected_agent_behaviour_action =np.argmax(agent_policy[current_state_index])
+
+ #counters for plots
+ n_assistance_lev_per_episode_run[r][selected_agent_behaviour_action] += 1
+ print("agent_assistance {}, attempt {}, game {}".format(selected_agent_behaviour_action, attempt_counter, game_state_counter))
+
+ ##########################QUERY FOR THE USER ACTION AND REACT TIME#####################################
+ #return the user action in this state based on the Persona profile
+ vars_user_evidence = { game_state_bn_name: game_state_counter,
+ attempt_bn_name: attempt_counter - 1,
+ agent_assistance_bn_name: selected_agent_behaviour_action,
+ }
+
+ query_user_action_prob = bn_functions.infer_prob_from_state(user_bn_model=bn_model_user_action,
+ infer_variable=var_user_action_target_action,
+ evidence_variables=vars_user_evidence)
+
+ selected_user_action = bn_functions.get_stochastic_action(query_user_action_prob.values)
+ #selected_user_action = np.argmax(query_user_action_prob.values)
+
+ #updates counters for simulation
+ iter_counter += 1
+ next_state, task_progress_counter, game_state_counter, attempt_counter, correct_move_counter, \
+ wrong_move_counter, timeout_counter, max_attempt_counter = compute_next_state(selected_user_action,
+ task_progress_counter,
+ attempt_counter,
+ correct_move_counter, wrong_move_counter,
+ timeout_counter, max_attempt_counter,
+ max_attempt_per_object)
+
+ # update counters
+ if game_state_counter <= 2:
+ user_action_per_agent_assistance[selected_agent_behaviour_action][selected_user_action] += 1
+ attempt_counter_per_user_action[selected_user_action][attempt_counter - 1] += 1
+ game_state_counter_per_user_action[selected_user_action][game_state_counter] += 1
+
+ # store the (state, action, next_state)
+ episode.append((ep.state_from_point_to_index(state_space, current_state),
+ selected_agent_behaviour_action,
+ ep.state_from_point_to_index(state_space, next_state)))
+
+ print("current_state ", current_state, " user_action:", selected_user_action, " next_state ", next_state)
+ ####################################END of EPISODE#######################################
+ print("game_state_counter {}, iter_counter {}, correct_counter {}, wrong_counter {}, "
+ "timeout_counter {}, max_attempt {}".format(game_state_counter, iter_counter, correct_move_counter,
+ wrong_move_counter, timeout_counter, max_attempt_counter))
+
+ #save episode
+ episodes.append(Episode(episode))
+
+ #update user models
+ # bn_model_user_action = bn_functions.update_cpds_tables(bn_model_user_action, user_action_dynamic_variables, alpha_learning)
+
+ #reset counter
+ user_action_per_agent_assistance = [[0 for i in range(User_Action.counter.value)]
+ for j in range(Agent_Assistance.counter.value)]
+ attempt_counter_per_user_action = [[0 for i in range(Attempt.counter.value)] for j in
+ range(User_Action.counter.value)]
+ game_state_counter_per_user_action = [[0 for i in range(Game_State.counter.value)] for j in
+ range(User_Action.counter.value)]
+
+ #for plots
+ n_correct_per_episode_run[r] = correct_move_counter
+ n_wrong_per_episode_run[r] = wrong_move_counter
+ n_timeout_per_episode_run[r] = timeout_counter
+ n_max_attempt_per_episode_run[r] = max_attempt_counter
+ game_performance_episode_run[r] = [n_correct_per_episode_run[r],
+ n_wrong_per_episode_run[r],
+ n_timeout_per_episode_run[r],
+ n_max_attempt_per_episode_run[r]]
+
+ #compute average of the values for one epoch and store it
+ n_correct_per_episode_epoch[e] = sum(n_correct_per_episode_run)/run
+ n_wrong_per_episode_epoch[e] = sum(n_wrong_per_episode_run)/run
+ n_timeout_per_episode_epoch[e] = sum(n_timeout_per_episode_run)/run
+ n_max_attempt_per_episode_epoch[e] = sum(n_max_attempt_per_episode_run)/run
+ game_performance_episode_epoch[e] = list(map(lambda x: sum(x)/run, zip(*game_performance_episode_run)))
+ n_assistance_lev_per_episode_epoch[e] = list(map(lambda x: sum(x)/run, zip(*n_assistance_lev_per_episode_run)))
+
+ #reset variables
+ n_correct_per_episode_run = [0] * run
+ n_wrong_per_episode_run = [0] * run
+ n_timeout_per_episode_run = [0] * run
+ n_max_attempt_per_episode_run = [0] * run
+ game_performance_episode_run = [0] * run
+ n_assistance_lev_per_episode_run = [[0 for i in range(Agent_Assistance.counter.value)] for j in range(run)]
+
+
+
+ return game_performance_episode_epoch, n_assistance_lev_per_episode_epoch, episodes
@@ -380,7 +380,7 @@ def simulation(bn_model_user_action,
# epochs = 20
# scaling_factor = 1
# # initialise the agent
-# bn_model_user_action = bnlearn.import_DAG('/home/pal/Documents/Framework/bn_generative_model/bn_persona_model/persona_model_test.bif')
+# bn_model_user_action = bnlearn.import_DAG('/home/pal/Documents/Framework/bn_generative_model/bn_persona_model/persona_model_template.bif')
#
# # initialise memory, attention and reactivity variables
# persona_memory = 0;
diff --git a/test.py b/test.py
index a5ec2a8..b5fd6ab 100644
--- a/test.py
+++ b/test.py
@@ -25,29 +25,28 @@ def import_data_from_csv(csv_filename, dag_filename):
-DAG_shared = import_data_from_csv(csv_filename='bn_persona_model/cognitive_game.csv', dag_filename='bn_persona_model/persona_model_test.bif')
+#DAG_shared = import_data_from_csv(csv_filename='/data/test.csv', dag_filename='bn_persona_model/persona_model_template.bif')
-# DAG = bn.import_DAG('bn_persona_model/persona_model_test.bif')
+DAG_agent = bn.import_DAG('old_models/bn_agent_model/agent_test.bif')
+DAG_user = bn.import_DAG('old_models/bn_persona_model/persona_test.bif')
+
# #G = bn.plot(DAG)
#
-# q_origin = bn.inference.fit(DAG, variables=[ 'user_action'], evidence={
+# q_origin = bn.inference.fit(DAG_agent, variables=[ 'agent_assistance'], evidence={
# 'game_state':0,
-# 'attempt':0,
-# 'agent_feedback':0,
-# 'agent_assistance':0,
+# 'attempt':3,
+# #'agent_assistance':2,
# })
-# q_shared = bn.inference.fit(DAG_shared, variables=[ 'user_action'], evidence={
-# 'game_state':0,
-# 'attempt':0,
-# 'agent_feedback':1,
-# 'user_memory': 2,
-# 'user_reactivity':2,
-# 'agent_assistance':0,
-# })
-#
-# print("Q origin: ", q_origin.values, " Q shared ", q_shared.values)
+q_shared = bn.inference.fit(DAG_user, variables=[ 'user_action'], evidence={
+ 'game_state':0,
+ 'attempt':3,
+ 'agent_assistance':0,
+})
+
+
+print( " Q shared ", q_shared.values)
# df = pd.read_csv('bn_persona_model/cognitive_game.csv')
# df = bn.sampling(DAG, n=10000)
# #model_sl = bn.structure_learning.fit(df, methodtype='hc', scoretype='bic')
@@ -72,10 +71,10 @@ DAG_shared = import_data_from_csv(csv_filename='bn_persona_model/cognitive_game.
# print("GS:", gs, " ATT:", att, " AA", aas, " AF", af)
#
# df.head()
-# DAG = bn.import_DAG('bn_persona_model/persona_model_test.bif', CPD=False)
+# DAG = bn.import_DAG('bn_persona_model/persona_model_template.bif', CPD=False)
# bn.plot(DAG)
# DAG_update = bn.parameter_learning.fit(DAG, df)
-# DAG_true = bn.import_DAG('bn_persona_model/persona_model_test.bif', CPD=True)
+# DAG_true = bn.import_DAG('bn_persona_model/persona_model_template.bif', CPD=True)
# q1 = bn.inference.fit(DAG_update, variables=['user_action'], evidence={
# 'game_state': 0,
# 'attempt':2,
--
GitLab