class CellRider(object):
def __init__(self, cfg, w = None):
self.cfg = cfg
self.cfg.update(defaultcfg, overwrite = False, protect_structure = True)
self.bounds = self.cfg.effect.s_bounds
assert self.bounds is not None
self.celltree = CellTree(self.bounds, cfg.cell.crit_size, cfg, w = w)
def __len__(self):
return len(self.celltree)
def add_effect(self, effect, goal = None, prediction = None, competence = None, pred_error = None):
self.celltree.add(effect, goal = goal, prediction = prediction, competence = competence, pred_error = pred_error)
def next_goal(self):
target_cell = None
dice = random.random()
if dice < self.cfg.effect.random_explo:
target_cell = random.sample(self.celltree.active_cells, 1)[0]
# elif dice < self.cfg.effect.random_explo + self.cfg.effect.unknown_explo:
# target_cell = self._priority_choice()
if target_cell is None:
target_cell = self._priority_choice()
if target_cell is None:
cell_uid = [cell.uid for cell in self.celltree.active_cells]
interests = [cell.interest() for cell in self.celltree.active_cells]
if sum(interests) > 0:
idx = toolbox.roulette_wheel(interests)
priority_uid = [cell.uid for cell in self.celltree.active_cells]
target_cell = self.celltree.cells[priority_uid[idx]]
if target_cell is not None:
return target_cell.random_point()
else:
return None
def _priority_choice(self):
target_cell = None
priority_uid = [cell.uid for cell in self.celltree.active_cells]
priorities = [self._priority(cell) for cell in self.celltree.active_cells]
if sum(priorities) > 0:
idx = toolbox.roulette_wheel(priorities)
target_cell = self.celltree.cells[priority_uid[idx]]
return target_cell
def _priority(self, cell):
if len(cell.ecell) > 0:
return max(3 - len(cell.gcell), 0)
else:
return 0
def __init__(self, cfg, w=None):
self.cfg = cfg
self.cfg.update(defaultcfg, overwrite=False, protect_structure=True)
self.bounds = self.cfg.effect.s_bounds
assert self.bounds is not None
self.celltree = CellTree(self.bounds, cfg.cell.crit_size, cfg, w=w)
def __init__(self, cfg, w = None):
self.cfg = cfg
self.cfg.update(defaultcfg, overwrite = False, protect_structure = True)
self.bounds = self.cfg.effect.s_bounds
assert self.bounds is not None
self.celltree = CellTree(self.bounds, cfg.cell.crit_size, cfg, w = w)
class CellRider(object):
def __init__(self, cfg, w=None):
self.cfg = cfg
self.cfg.update(defaultcfg, overwrite=False, protect_structure=True)
self.bounds = self.cfg.effect.s_bounds
assert self.bounds is not None
self.celltree = CellTree(self.bounds, cfg.cell.crit_size, cfg, w=w)
def __len__(self):
return len(self.celltree)
def add_effect(self,
effect,
goal=None,
prediction=None,
competence=None,
pred_error=None):
self.celltree.add(effect,
goal=goal,
prediction=prediction,
competence=competence,
pred_error=pred_error)
def next_goal(self):
target_cell = None
dice = random.random()
if dice < self.cfg.effect.random_explo:
target_cell = random.sample(self.celltree.active_cells, 1)[0]
# elif dice < self.cfg.effect.random_explo + self.cfg.effect.unknown_explo:
# target_cell = self._priority_choice()
if target_cell is None:
target_cell = self._priority_choice()
if target_cell is None:
cell_uid = [cell.uid for cell in self.celltree.active_cells]
interests = [
cell.interest() for cell in self.celltree.active_cells
]
if sum(interests) > 0:
idx = toolbox.roulette_wheel(interests)
priority_uid = [
cell.uid for cell in self.celltree.active_cells
]
target_cell = self.celltree.cells[priority_uid[idx]]
if target_cell is not None:
return target_cell.random_point()
else:
return None
def _priority_choice(self):
target_cell = None
priority_uid = [cell.uid for cell in self.celltree.active_cells]
priorities = [
self._priority(cell) for cell in self.celltree.active_cells
]
if sum(priorities) > 0:
idx = toolbox.roulette_wheel(priorities)
target_cell = self.celltree.cells[priority_uid[idx]]
return target_cell
def _priority(self, cell):
if len(cell.ecell) > 0:
return max(3 - len(cell.gcell), 0)
else:
return 0
