def __init__(self, actions, mapname, contextD, context_rewards, **kwargs):
"""Initialize the environment variables.
:param actions: actions available to the system
:type actions: list of tuples (action_name,action_vector)
:param mapname: filename for map file
:param contextD: dimension of vector representing context
:param context_rewards: mapping from region labels to rewards for being
in that region (each entry represents one context)
:type context_rewards: dict {"regionlabel":rewardval,...}
:param **kwargs: see PlaceCellEnvironment.__init__
"""
PlaceCellEnvironment.__init__(self,
actions,
mapname,
name="ContextEnvironment",
**kwargs)
self.rewards = context_rewards
# generate vectors representing each context
self.contexts = {} # mapping from region label to context vector
for i, r in enumerate(self.rewards):
self.contexts[r] = list(MU.I(contextD)[i])
self.context = self.contexts[random.choice(self.contexts.keys())]
# randomly pick a new context every context_delay seconds
self.context_delay = 60
self.context_update = self.context_delay
self.create_origin("placewcontext",
lambda: self.place_activations + self.context)
self.create_origin("context", lambda: self.context)
def __init__(self, *args, **kwargs):
"""Initialize environment variables.
:param name: name for environment
:param *args: see PlaceCellEnvironment.__init__
:param **kwargs: see PlaceCellEnvironment.__init__
"""
PlaceCellEnvironment.__init__(self, name="DeliveryEnvironment", *args,
**kwargs)
# reward value when no reward condition is met
self.defaultreward = -0.05
self.contexts = {"in_hand": [1, 0], "out_hand": [0, 1]}
self.in_hand = False
self.create_origin("placewcontext",
lambda: (self.place_activations +
self.contexts["in_hand"] if self.in_hand
else self.place_activations +
self.contexts["out_hand"]))
self.create_origin("context", lambda: (self.contexts["in_hand"]
if self.in_hand else
self.contexts["out_hand"]))
def __init__(self, actions, mapname, contextD, context_rewards, **kwargs):
"""Initialize the environment variables.
:param actions: actions available to the system
:type actions: list of tuples (action_name,action_vector)
:param mapname: filename for map file
:param contextD: dimension of vector representing context
:param context_rewards: mapping from region labels to rewards for being in that
region (each entry represents one context)
:type context_rewards: dict {"regionlabel":rewardval,...}
:param **kwargs: see PlaceCellEnvironment.__init__
"""
PlaceCellEnvironment.__init__(self, actions, mapname, name="ContextEnvironment", **kwargs)
self.rewards = context_rewards
# generate vectors representing each context
self.contexts = {} # mapping from region label to context vector
for i, r in enumerate(self.rewards):
# self.contexts[r] = list(RandomHypersphereVG().genVectors(1, contextD)[0])
self.contexts[r] = list(MU.I(contextD)[i])
self.context = self.contexts[random.choice(self.contexts.keys())]
# randomly pick a new context every context_delay seconds
self.context_delay = 60
self.context_update = self.context_delay
self.create_origin("placewcontext", lambda: self.place_activations + self.context)
self.create_origin("context", lambda: self.context)
def __init__(self, *args, **kwargs):
"""Initialize environment variables.
:param name: name for environment
:param *args: see PlaceCellEnvironment.__init__
:param **kwargs: see PlaceCellEnvironment.__init__
"""
PlaceCellEnvironment.__init__(self,
name="DeliveryEnvironment",
*args,
**kwargs)
# reward value when no reward condition is met
self.defaultreward = -0.05
self.contexts = {"in_hand": [1, 0], "out_hand": [0, 1]}
self.in_hand = False
self.create_origin(
"placewcontext", lambda:
(self.place_activations + self.contexts["in_hand"] if self.in_hand
else self.place_activations + self.contexts["out_hand"]))
self.create_origin(
"context", lambda:
(self.contexts["in_hand"]
if self.in_hand else self.contexts["out_hand"]))
