def __init__(self,dimensions,min=-1,max=1,noise=0):
self.inputs=[]
self.outputs=[]
self.dimensions=dimensions
self.min=min
self.max=max
self.array_noise=noise
self.mode='direct'
self._all_nodes=None
self.accumulator=Accumulator(dimensions)
def __init__(self,dimensions,min=-1,max=1,noise=0):
self.inputs=[]
self.outputs=[]
self.dimensions=dimensions
self.min=min
self.max=max
self.array_noise=noise
self.mode='direct'
self._all_nodes=None
self.accumulator=Accumulator(dimensions)
class ArrayNode:
_set_array=None
_value=None
_array=None
# _input=None
_output=None
def __init__(self,dimensions,min=-1,max=1,noise=0):
self.inputs=[]
self.outputs=[]
self.dimensions=dimensions
self.min=min
self.max=max
self.array_noise=noise
self.mode='direct'
self._all_nodes=None
self.accumulator=Accumulator(dimensions)
def clone(self):
clone=copy.copy(self)
clone.inputs=[]
clone.outputs=[]
clone.accumulator=Accumulator(self.dimensions)
return clone
def set(self,value,calc_output=True):
if value is None:
self._set_array=None
self._array=None
self._value=None
if calc_output: self._calc_output()
else:
array=numpy.array(self.value_to_array(value))
if len(array.shape)>1: array.shape=array.shape[0]
assert len(array)==self.dimensions
self._set_array=array
if calc_output: self._calc_output()
self._array=None
self._value=None
def array(self):
if self._array is None:
if self._output is None: self._output=numpy.zeros(self.dimensions)
self._array=self._output
return self._array
def value(self):
if self._value is None:
self._value=self.array_to_value(self.array())
return self._value
def _calc_output(self):
if self._output is None: self._output=numpy.zeros(self.dimensions)
else: self._output[:]=0
if self._set_array is not None:
self._output+=self._set_array
else:
self._output+=self.accumulator.value()
if self.array_noise>0:
self._output+=numpy.random.randn(self.dimensions)*self.array_noise
def _clear_inputs(self):
pass
def _transmit_direct_direct(self,conn,dt):
conn.pop2.accumulator.add(conn.apply_func_weight(self._output),conn.tau,dt)
def tick(self,dt=None):
if dt is None: dt=getattr(self,'dt',None)
nodes=self._all_nodes
if nodes is None:
nodes=self.all_nodes()
for n in nodes:
n._all_nodes=nodes
for n in nodes:
n._clear_inputs()
n.accumulator.tick(dt)
for n in nodes:
for conn in n.outputs:
f=getattr(n,'_transmit_%s'%conn.type())
f(conn,dt)
for n in nodes:
n._calc_output()
n._value=None
n._array=None
def connect(self,other,func=None,weight=None,tau=None):
return connect(self,other,func=func,weight=weight,tau=tau)
"""
def all_nodes(self,list=None):
if list is None: list=[]
list.append(self)
for c in self.inputs:
if c.pop1 not in list:
c.pop1.all_nodes(list)
for c in self.outputs:
if c.pop2 not in list:
c.pop2.all_nodes(list)
return list
"""
def all_nodes(self):
all=[]
done=set()
work=set([self])
while len(work)>0:
n=work.pop()
done.add(n)
all.append(n)
for a in n.inputs:
if a.pop1 not in done: work.add(a.pop1)
for a in n.outputs:
if a.pop2 not in done: work.add(a.pop2)
return all
class ArrayNode:
_set_array=None
_value=None
_array=None
# _input=None
_output=None
def __init__(self,dimensions,min=-1,max=1,noise=0):
self.inputs=[]
self.outputs=[]
self.dimensions=dimensions
self.min=min
self.max=max
self.array_noise=noise
self.mode='direct'
self._all_nodes=None
self.accumulator=Accumulator(dimensions)
def set(self,value,calc_output=True):
if value is None:
self._set_array=None
self._array=None
self._value=None
if calc_output: self._calc_output()
else:
array=numpy.array(self.value_to_array(value))
if len(array.shape)>1: array.shape=array.shape[0]
assert len(array)==self.dimensions
self._set_array=array
if calc_output: self._calc_output()
self._array=None
self._value=None
def array(self):
if self._array is None:
if self._output is None: self._output=numpy.zeros(self.dimensions)
self._array=self._output
return self._array
def value(self):
if self._value is None:
self._value=self.array_to_value(self.array())
return self._value
def _calc_output(self):
if self._output is None: self._output=numpy.zeros(self.dimensions)
else: self._output[:]=0
if self._set_array is not None:
self._output+=self._set_array
else:
self._output+=self.accumulator.value()
if self.array_noise>0:
self._output+=numpy.random.randn(self.dimensions)*self.array_noise
def _clear_inputs(self):
pass
def _transmit_direct_direct(self,conn,dt):
conn.pop2.accumulator.add(conn.apply_func_weight(self._output),conn.tau,dt)
def tick(self,dt=None):
if dt is None: dt=getattr(self,'dt',None)
nodes=self._all_nodes
if nodes is None:
nodes=self.all_nodes()
for n in nodes:
n._all_nodes=nodes
for n in nodes:
n._clear_inputs()
n.accumulator.tick(dt)
for n in nodes:
for conn in n.outputs:
f=getattr(n,'_transmit_%s'%conn.type())
f(conn,dt)
for n in nodes:
n._calc_output()
n._value=None
n._array=None
def connect(self,other,func=None,weight=None,tau=None):
return connect(self,other,func=func,weight=weight,tau=tau)
"""
def all_nodes(self,list=None):
if list is None: list=[]
list.append(self)
for c in self.inputs:
if c.pop1 not in list:
c.pop1.all_nodes(list)
for c in self.outputs:
if c.pop2 not in list:
c.pop2.all_nodes(list)
return list
"""
def all_nodes(self):
all=[]
done=set()
work=set([self])
while len(work)>0:
n=work.pop()
done.add(n)
all.append(n)
for a in n.inputs:
if a.pop1 not in done: work.add(a.pop1)
for a in n.outputs:
if a.pop2 not in done: work.add(a.pop2)
return all
def clone(self):
clone = copy.copy(self)
clone.inputs = []
clone.outputs = []
clone.accumulator = Accumulator(self.dimensions)
return clone