Merge pull request #7 from Intelligent-Adaptive-Interventions-Lab/initcontext

add context init imputer api

Author: VictorS67

mooclet_engine/engine/migrations/0025_auto_20220425_1940.py

@@ -0,0 +1,33 @@
+# Generated by Django 2.2.6 on 2022-04-25 19:40
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('engine', '0024_auto_20191011_1713'),
+    ]
+
+    operations = [
+        migrations.AddField(
+            model_name='variable',
+            name='max_value',
+            field=models.FloatField(default=1.0),
+        ),
+        migrations.AddField(
+            model_name='variable',
+            name='min_value',
+            field=models.FloatField(default=0.0),
+        ),
+        migrations.AddField(
+            model_name='variable',
+            name='sample_thres',
+            field=models.PositiveIntegerField(default=15),
+        ),
+        migrations.AddField(
+            model_name='variable',
+            name='value_type',
+            field=models.CharField(choices=[('BIN', 'binary'), ('ORD', 'ordinary'), ('CONT', 'continuous')], default='BIN', max_length=100),
+        ),
+    ]

mooclet_engine/engine/models.py

@@ -81,9 +81,21 @@ class Learner(models.Model):
 
 
 class Variable(models.Model):
+    BINARY = 'BIN'
+    ORDINARY = 'ORD'
+    CONTINUOUS = 'CONT'
+    VARIABLE_TYPE = (
+        (BINARY, 'binary'), 
+        (ORDINARY, 'ordinary'), 
+        (CONTINUOUS, 'continuous')
+    )
     name = models.CharField(max_length=100, unique=True)
     environment = models.ForeignKey(Environment,blank=True,null=True, default=None, on_delete=models.SET_NULL)
     variable_id = models.PositiveIntegerField(blank=True,null=True)
+    min_value = models.FloatField(default=0.0)
+    max_value = models.FloatField(default=1.0)
+    value_type = models.CharField(max_length=100, choices=VARIABLE_TYPE, default=BINARY)
+    sample_thres = models.PositiveIntegerField(default=15)
 
     def __str__(self):
         return self.name

mooclet_engine/engine/views.py

@@ -211,3 +211,58 @@ def get(self, request):
             imputer['imputer'] = np.random.choice([0, 1], 1, p=[1-binary_t_prop, binary_t_prop])
 
         return Response(imputer)
+
+
+class getContextualImputer(APIView):
+    def get(self, request):
+        req = json.loads(request.body)
+        has_learner = request.args.get("learner")
+        has_mooclet = request.args.get("mooclet")
+        has_policy = request.args.get("policy")
+        has_contexts = request.args.get("contexts")
+
+        if has_learner is None or has_mooclet is None or has_policy is None:
+            return Response({"error": "invalid"}, status=500)
+        
+        mooclet = Mooclet.objects.get(pk=req["mooclet"])
+        learner = Learner.objects.get(name=req["learner"])
+        policy = Policy.objects.get(pk=req["policy"])
+
+        if has_contexts is None:
+            contextual_vars = req["contexts"]
+        else:
+            mooclet_params = PolicyParameters.objects.get(mooclet=mooclet, policy=policy)
+            parameters = mooclet_params.parameters
+            contextual_vars = list(filter(lambda context: context != "version", parameters["contextual_variables"]))
+
+        imputer = {}
+        for context_var in contextual_vars:
+            variable = Variable.objects.filter(name=context_var)
+            val_type = variable.value_type
+            val_min = variable.min_value
+            val_max = variable.max_value
+            sample_thres = variable.sample_thres
+            values = Value.objects.filter(variable__name=context_var, mooclet=mooclet)
+            num_values = values.count()
+
+            if num_values == 0 or num_values < sample_thres:
+                if val_type != "continuous":
+                    sample = np.random.choice(np.arange(val_min, val_max + 1))
+                else:
+                    sample = np.random.uniform(val_min, val_max)
+                    sample = (sample - val_min) / (val_max - val_min)
+            else:
+                val_lst = list(values.value_list("value"))
+                sample = np.random.choice(val_lst)
+            
+            Value.objects.create(
+                variable=variable, 
+                value=sample, 
+                text="Init Context",
+                learner=learner,
+                mooclet=mooclet
+            )
+
+            imputer["imputer"] = sample
+
+        return Response(imputer)