def newPosition(self, globalBestPosition, rng):
"""See comments in base class."""
# First, update the velocity. The new velocity is given as:
# v = (inertia * v) + (cogRate * r1 * (localBest-pos))
# + (socRate * r2 * (globalBest-pos))
#
# where r1 and r2 are random numbers between 0 and 1.0
lb = float(Configuration.get("nupic.hypersearch.randomLowerBound"))
ub = float(Configuration.get("nupic.hypersearch.randomUpperBound"))
self._velocity = (self._velocity * self._inertia +
rng.uniform(lb, ub) * self._cogRate *
(self._bestPosition - self.getPosition()))
if globalBestPosition is not None:
self._velocity += rng.uniform(lb, ub) * self._socRate * (
globalBestPosition - self.getPosition())
# update position based on velocity
self._position += self._velocity
# Clip it
self._position = max(self.min, self._position)
self._position = min(self.max, self._position)
# Return it
return self.getPosition()
def newPosition(self, globalBestPosition, rng):
"""See comments in base class."""
# First, update the velocity. The new velocity is given as:
# v = (inertia * v) + (cogRate * r1 * (localBest-pos))
# + (socRate * r2 * (globalBest-pos))
#
# where r1 and r2 are random numbers between 0 and 1.0
lb=float(Configuration.get("nupic.hypersearch.randomLowerBound"))
ub=float(Configuration.get("nupic.hypersearch.randomUpperBound"))
self._velocity = (self._velocity * self._inertia + rng.uniform(lb, ub) *
self._cogRate * (self._bestPosition - self.getPosition()))
if globalBestPosition is not None:
self._velocity += rng.uniform(lb, ub) * self._socRate * (
globalBestPosition - self.getPosition())
# update position based on velocity
self._position += self._velocity
# Clip it
self._position = max(self.min, self._position)
self._position = min(self.max, self._position)
# Return it
return self.getPosition()
def __init__(self, milestones=None, logLevel=None):
# Set class constants.
self.MATURITY_WINDOW = int(
Configuration.get("nupic.hypersearch.swarmMaturityWindow"))
self.MAX_GENERATIONS = int(
Configuration.get("nupic.hypersearch.swarmMaxGenerations"))
if self.MAX_GENERATIONS < 0:
self.MAX_GENERATIONS = None
# Set up instsance variables.
self._isTerminationEnabled = bool(
int(Configuration.get('nupic.hypersearch.enableSwarmTermination')))
self.swarmBests = dict()
self.swarmScores = dict()
self.terminatedSwarms = set([])
self._logger = logging.getLogger(".".join([
'com.numenta', self.__class__.__module__, self.__class__.__name__
]))
if milestones is not None:
self.milestones = milestones
else:
self.milestones = copy.deepcopy(self._DEFAULT_MILESTONES)
def __init__(self,
min,
max,
stepSize=None,
inertia=None,
cogRate=None,
socRate=None):
"""Construct a variable that permutes over floating point values using
the Particle Swarm Optimization (PSO) algorithm. See descriptions of
PSO (i.e. http://en.wikipedia.org/wiki/Particle_swarm_optimization)
for references to the inertia, cogRate, and socRate parameters.
Parameters:
-----------------------------------------------------------------------
min: min allowed value of position
max: max allowed value of position
stepSize: if not None, the position must be at min + N * stepSize,
where N is an integer
inertia: The inertia for the particle.
cogRate: This parameter controls how much the particle is affected
by its distance from it's local best position
socRate: This parameter controls how much the particle is affected
by its distance from the global best position
"""
super(PermuteFloat, self).__init__()
self.min = min
self.max = max
self.stepSize = stepSize
# The particle's initial position and velocity.
self._position = (self.max + self.min) / 2.0
self._velocity = (self.max - self.min) / 5.0
# The inertia, cognitive, and social components of the particle
self._inertia = (float(Configuration.get("nupic.hypersearch.inertia"))
if inertia is None else inertia)
self._cogRate = (float(Configuration.get("nupic.hypersearch.cogRate"))
if cogRate is None else cogRate)
self._socRate = (float(Configuration.get("nupic.hypersearch.socRate"))
if socRate is None else socRate)
# The particle's local best position and the best global position.
self._bestPosition = self.getPosition()
self._bestResult = None
def __init__(self, min, max, stepSize=None, inertia=None, cogRate=None,
socRate=None):
"""Construct a variable that permutes over floating point values using
the Particle Swarm Optimization (PSO) algorithm. See descriptions of
PSO (i.e. http://en.wikipedia.org/wiki/Particle_swarm_optimization)
for references to the inertia, cogRate, and socRate parameters.
Parameters:
-----------------------------------------------------------------------
min: min allowed value of position
max: max allowed value of position
stepSize: if not None, the position must be at min + N * stepSize,
where N is an integer
inertia: The inertia for the particle.
cogRate: This parameter controls how much the particle is affected
by its distance from it's local best position
socRate: This parameter controls how much the particle is affected
by its distance from the global best position
"""
super(PermuteFloat, self).__init__()
self.min = min
self.max = max
self.stepSize = stepSize
# The particle's initial position and velocity.
self._position = (self.max + self.min) / 2.0
self._velocity = (self.max - self.min) / 5.0
# The inertia, cognitive, and social components of the particle
self._inertia = (float(Configuration.get("nupic.hypersearch.inertia"))
if inertia is None else inertia)
self._cogRate = (float(Configuration.get("nupic.hypersearch.cogRate"))
if cogRate is None else cogRate)
self._socRate = (float(Configuration.get("nupic.hypersearch.socRate"))
if socRate is None else socRate)
# The particle's local best position and the best global position.
self._bestPosition = self.getPosition()
self._bestResult = None
def __init__(self, milestones=None, logLevel=None):
# Set class constants.
self.MATURITY_WINDOW = int(Configuration.get("nupic.hypersearch.swarmMaturityWindow"))
self.MAX_GENERATIONS = int(Configuration.get("nupic.hypersearch.swarmMaxGenerations"))
if self.MAX_GENERATIONS < 0:
self.MAX_GENERATIONS = None
# Set up instsance variables.
self._isTerminationEnabled = bool(int(Configuration.get("nupic.hypersearch.enableSwarmTermination")))
self.swarmBests = dict()
self.swarmScores = dict()
self.terminatedSwarms = set([])
self._logger = logging.getLogger(".".join(["com.numenta", self.__class__.__module__, self.__class__.__name__]))
if milestones is not None:
self.milestones = milestones
else:
self.milestones = copy.deepcopy(self._DEFAULT_MILESTONES)
