teamwork.reward.MinMaxGoal.MinMaxGoal

Class MinMaxGoal

A reward subfunction that is linear in a single feature/action

0) Creating a new goal (i.e., __init__()): goal = MinMaxGoal(entity,direction,type,key)

1) Accessing elements of this goal goal.isMax()

Instance Methods

__init__(self, entity=None, direction=None, goalType=None, key=None, value={})
Constructs a goal object with the specified field values source code

__copy__(self)

source code

generateName(self)
Returns a canonical string representation

source code

boolean

isMax(self)
Returns: true if maximization goal, false if minimization

source code

Key

toKey(self)
Returns: the vector key corresponding to this goal

source code

reward(self, context)
Applies this goal in the specified context

source code

(str,Distribution)

maxElement(self)
Finds the element that has the highest value

source code

__getitem__(self, index)
Accessor: supports access in the form `self[index]'

source code

__setitem__(self, index, value)
Accessor: supports access in the form `self[index]=x'

source code

keys(self)
Returns: all of the element names in this goal

source code

evaluate(self, context)
Returns a new goal instance in the given context by computing the reward and storing it in its value attribute

source code

__add__(self, goal)

source code

__neg__(self)

source code

__sub__(self, goal)

source code

__mul__(self, factor)

source code

__div__(self, factor)

source code

total(self)
Returns sum over the individual elements in an instantiated goal

source code

__str__(self)

source code

__hash__(self)

source code

__xml__(self)

source code

parse(self, element)

source code

init(self, entity=None, direction=None, goalType=None, key=None, value=`{}`)
(Constructor)

source code

Constructs a goal object with the specified field values

The named arguments are stored as corresponding attributes:

Parameters:

entity (str[]) - list of names (for recursive belief access if a state goal, or for relevant actor/object if action goal)
direction - either 'min' or 'max'
goalType - either 'state' (value of a state feature), 'act' (number of occurences of action of given type), 'actActor' (number of occurrences of action of given type by given actor), 'actObject' (number of occurences of action of given type to given object)
key - for a 'state' goal, is the state feature to be min/maximized for an 'act', goal, is the act type to be min/maximized
direction,goalType,key (str)