class Network(object):
'''Initialize and run a dipde simulation.
The Network class handles the initialization of population and connection
objects, and provides a convenience time stepping loop to drive a network
simulation. Typical usage involves the create of populations and
connections, construction of a simulation object, and then call to
simulation.run()
Parameters
----------
population_list : list of ExternalPopulation, InternalPopulation objects
List of populations to include in simulation.
connection_list : list of Connection objects
List of connections to include in simulation.
verbose : bool (default=True)
Setting True prints current time-step at each update evaluation.
'''
def __init__(self,
population_list=[],
connection_list=[],
metadata={},
run_callback=lambda s: None,
update_callback=lambda s: None,
progress=None,
**kwargs):
self.update_callback = update_callback
self.run_callback = run_callback
self.progress = progress
self.population_list = []
for p in population_list:
if isinstance(p, dict):
curr_module, curr_class = p['module'], p['class']
curr_instance = getattr(importlib.import_module(curr_module),
curr_class)(**p)
self.population_list.append(curr_instance)
else:
self.population_list.append(p)
self.connection_list = []
for c in connection_list:
if isinstance(c, dict):
curr_module, curr_class = c['class']
curr_instance = getattr(importlib.import_module(curr_module),
curr_class)(**c)
if curr_instance.nsyn_input != 0:
self.connection_list.append(curr_instance)
else:
if c.nsyn_input != 0:
self.connection_list.append(c)
self.gid_dict = dict(
(population, ii)
for ii, population in enumerate(self.population_list))
for key, val in self.gid_dict.items():
self.gid_dict[val] = key
# Initialize:
self.connection_distribution_collection = ConnectionDistributionCollection(
)
# Each population needs access to dt:
for p in self.population_list:
p.simulation = self
# Each connection needs access to t:
for c in self.connection_list:
c.simulation = self
self.kill_signal = False
@property
def rank(self):
return self.synchronization_harness.rank
def run(self, *args, **kwargs):
try:
self._run(*args, **kwargs)
except Exception as e:
self.shutdown()
print 'Safely Shutting down after exception...'
print traceback.format_exc()
raise e
def _run(self, dt, tf, t0=0., synchronization_harness=None):
'''Main iteration control loop for simulation
The time step selection must be approximately of the same order as dv
for the internal populations, if the 'approx' time stepping method is
selected.
Parameters
----------
t0 : float (default=0.)
Network start time (unit=seconds), passed to initialize call.
tf : float (default=.1)
Network end time (unit=seconds).
dt : float (default=0.001)
Time step (unit=seconds).
'''
if synchronization_harness is None:
self.synchronization_harness = utilities.DefaultSynchronizationHarness(
)
else:
self.synchronization_harness = synchronization_harness
self.firing_rate_organizer = FiringRateOrganizer(
self.synchronization_harness)
self.dt = dt
self.t0 = t0
if not self.progress is None: self.progress.set_network(self)
# Initialize:
start_time = time.time()
self.tf = tf
self.ti = 0
if not self.progress is None: self.progress.initialize()
self.synchronization_harness.initialize(self.ti)
# Initialize populations:
for gid, p in enumerate(self.population_list):
p.initialize()
if self.synchronization_harness.gid_to_rank(gid) == self.rank:
self.firing_rate_organizer.push(self.ti, gid,
p.curr_firing_rate)
self.synchronization_harness.update(
self.ti,
self.firing_rate_organizer.firing_rate_dict_internal.setdefault(
self.ti, {}))
# Initialize connections:
for c in self.connection_list:
c.initialize()
for p in self.population_list:
try:
p.initialize_total_input_dict()
except AttributeError:
pass
self.initialization_time = time.time() - start_time
# Run
start_time = time.time()
while self.t < self.tf:
self.update()
self.run_time = time.time() - start_time
self.synchronization_harness.finalize()
self.shutdown()
self.run_callback(self)
def shutdown(self):
for c in self.connection_list:
c.shutdown()
for p in self.population_list:
p.shutdown()
@property
def t(self):
return self.t0 + self.ti * self.dt
def update(self):
if not self.progress is None: self.progress.update()
self.firing_rate_organizer.drop(self.ti)
self.ti += 1
logger.info('time: %s' % self.t)
for gid, p in enumerate(self.population_list):
if self.synchronization_harness.gid_to_rank(gid) == self.rank:
p.update()
self.firing_rate_organizer.push(self.ti, gid,
p.curr_firing_rate)
self.synchronization_harness.update(
self.ti,
self.firing_rate_organizer.firing_rate_dict_internal.setdefault(
self.ti, {}))
for c in self.connection_list:
c.update()
self.update_callback(self)
if self.kill_signal:
raise StopSimulation()
def to_dict(self, organization='sparse_adjacency_matrix'):
population_list = [p.to_dict() for p in self.population_list]
connection_list = [c.to_dict() for c in self.connection_list]
if organization == 'sparse_adjacency_matrix':
data_dict = {
'population_list': population_list,
'connection_list': connection_list,
'class': self.__class__.__name__,
'module': __name__
}
return data_dict
else:
raise NotImplementedError
def to_json(self, fh=None, **kwargs):
'''Save the contents of the InternalPopultion to json'''
data_dict = self.to_dict()
indent = kwargs.pop('indent', 2)
if fh is None:
return json.dumps(data_dict, indent=indent, **kwargs)
else:
return json.dump(data_dict, fh, indent=indent, **kwargs)
def copy(self):
return Network(**self.to_dict())
def get_curr_firing_rate(self, gid):
return self.firing_rate_organizer.pull(self.ti, gid)
def get_firing_rate(self, gid, t):
return self.population_list[gid].firing_rate(t)
def to_df(self):
pd.options.display.max_columns = 999
df_list = [p.to_df() for p in self.population_list]
return reorder_df_columns(pd.concat(df_list), ['class', 'module'])
def get_total_flux_matrix(self, internal_population, dt):
# Protect memory state of population and network:
population_ind = self.population_list.index(internal_population)
new_network = self.copy()
new_network.dt = dt
new_network.t0 = 0
new_network.ti = 0
new_internal_population = new_network.population_list[population_ind]
new_internal_population.initialize()
new_internal_population.initialize_total_input_dict()
return new_internal_population.get_total_flux_matrix()
@property
def external_population_list(self):
return [
p for p in self.population_list
if isinstance(p, ExternalPopulation)
]
@property
def internal_population_list(self):
return [
p for p in self.population_list
if isinstance(p, InternalPopulation)
]
class Network(object):
'''Initialize and run a dipde simulation.
The Network class handles the initialization of population and connection
objects, and provides a convenience time stepping loop to drive a network
simulation. Typical usage involves the create of populations and
connections, construction of a simulation object, and then call to
simulation.run()
Parameters
----------
population_list : list of ExternalPopulation, InternalPopulation objects
List of populations to include in simulation.
connection_list : list of Connection objects
List of connections to include in simulation.
verbose : bool (default=True)
Setting True prints current time-step at each update evaluation.
'''
def __init__(self,
population_list=[],
connection_list=[],
metadata={},
run_callback=lambda s: None,
update_callback=lambda s: None,
**kwargs):
self.update_callback = update_callback
self.run_callback = run_callback
self.population_list = []
for p in population_list:
if isinstance(p, dict):
curr_module, curr_class = p['module'], p['class']
curr_instance = getattr(importlib.import_module(curr_module),
curr_class)(**p)
self.population_list.append(curr_instance)
else:
self.population_list.append(p)
self.connection_list = []
for c in connection_list:
if isinstance(c, dict):
curr_module, curr_class = c['class']
curr_instance = getattr(importlib.import_module(curr_module),
curr_class)(**c)
if curr_instance.nsyn != 0:
self.connection_list.append(curr_instance)
else:
if c.nsyn != 0:
self.connection_list.append(c)
self.gid_dict = dict(
(population, ii)
for ii, population in enumerate(self.population_list))
for key, val in self.gid_dict.items():
self.gid_dict[val] = key
# Initialize:
self.connection_distribution_collection = ConnectionDistributionCollection(
)
# Each population needs access to dt:
for p in self.population_list:
p.simulation = self
# Each connection needs access to t:
for c in self.connection_list:
c.simulation = self
@property
def rank(self):
return self.distributed_configuration.rank()
def run(self,
dt,
tf,
t0=0.,
distributed_configuration=utilities.NullObject()):
'''Main iteration control loop for simulation
The time step selection must be approximately of the same order as dv
for the internal populations, if the 'approx' time stepping method is
selected.
Parameters
----------
t0 : float (default=0.)
Network start time (unit=seconds), passed to initialize call.
tf : float (default=.1)
Network end time (unit=seconds).
dt : float (default=0.001)
Time step (unit=seconds).
'''
self.distributed_configuration = distributed_configuration
self.firing_rate_organizer = FiringRateOrganizer(
self.distributed_configuration)
self.dt = dt
self.t0 = t0
# Initialize:
start_time = time.time()
self.tf = tf
self.ti = 0
self.distributed_configuration.initialize(self.ti)
# Initialize populations:
for gid, p in enumerate(self.population_list):
p.initialize()
self.firing_rate_organizer.push(self.ti, gid, p.curr_firing_rate)
self.distributed_configuration.update(
self.ti,
self.firing_rate_organizer.firing_rate_dict_internal[self.ti])
for p in self.population_list:
if isinstance(p, InternalPopulation):
p.initialize_total_input_dict()
# Initialize connections:
for c in self.connection_list:
c.initialize()
self.initialization_time = time.time() - start_time
# Run
start_time = time.time()
while self.t < self.tf:
self.update()
self.run_time = time.time() - start_time
self.distributed_configuration.finalize()
self.run_callback(self)
@property
def t(self):
return self.t0 + self.ti * self.dt
def update(self):
self.firing_rate_organizer.drop(self.ti)
self.ti += 1
logger.info('time: %s' % self.t)
for gid, p in enumerate(self.population_list):
if self.distributed_configuration.gid_to_rank(gid) == self.rank:
p.update()
self.firing_rate_organizer.push(self.ti, gid,
p.curr_firing_rate)
self.distributed_configuration.update(
self.ti,
self.firing_rate_organizer.firing_rate_dict_internal[self.ti])
for c in self.connection_list:
c.update()
self.update_callback(self)
def to_dict(self):
data_dict = {
'population_list': [p.to_dict() for p in self.population_list],
'connection_list': [c.to_dict() for c in self.connection_list],
'class': self.__class__.__name__,
'module': __name__
}
return data_dict
def to_json(self, fh=None, **kwargs):
'''Save the contents of the InternalPopultion to json'''
data_dict = self.to_dict()
indent = kwargs.pop('indent', 2)
if fh is None:
return json.dumps(data_dict, indent=indent, **kwargs)
else:
return json.dump(data_dict, fh, indent=indent, **kwargs)
def copy(self):
return Network(**self.to_dict())
def get_curr_firing_rate(self, gid):
return self.firing_rate_organizer.pull(self.ti, gid)
def get_firing_rate(self, gid, t):
return self.population_list[gid].firing_rate(t)
def to_df(self):
pd.options.display.max_columns = 999
df_list = [p.to_df() for p in self.population_list]
return reorder_df_columns(pd.concat(df_list), ['class', 'module'])
class Network(object):
'''Initialize and run a dipde simulation.
The Network class handles the initialization of population and connection
objects, and provides a convenience time stepping loop to drive a network
simulation. Typical usage involves the create of populations and
connections, construction of a simulation object, and then call to
simulation.run()
Parameters
----------
population_list : list of ExternalPopulation, InternalPopulation objects
List of populations to include in simulation.
connection_list : list of Connection objects
List of connections to include in simulation.
verbose : bool (default=True)
Setting True prints current time-step at each update evaluation.
'''
def __init__(self,
population_list=[],
connection_list=[],
metadata={},
run_callback=lambda s: None,
update_callback=lambda s: None,
progress=None,
**kwargs):
self.update_callback = update_callback
self.run_callback = run_callback
self.progress = progress
self.population_list = []
for p in population_list:
if isinstance(p, dict):
curr_module, curr_class = p['module'], p['class']
curr_instance = getattr(importlib.import_module(curr_module), curr_class)(**p)
self.population_list.append(curr_instance)
else:
self.population_list.append(p)
self.connection_list = []
for c in connection_list:
if isinstance(c, dict):
curr_module, curr_class = c['class']
curr_instance = getattr(importlib.import_module(curr_module), curr_class)(**c)
if curr_instance.nsyn_input != 0:
self.connection_list.append(curr_instance)
else:
if c.nsyn_input != 0:
self.connection_list.append(c)
self.gid_dict = dict((population, ii) for ii, population in enumerate(self.population_list))
for key, val in self.gid_dict.items():
self.gid_dict[val] = key
# Initialize:
self.connection_distribution_collection = ConnectionDistributionCollection()
# Each population needs access to dt:
for p in self.population_list:
p.simulation = self
# Each connection needs access to t:
for c in self.connection_list:
c.simulation = self
self.kill_signal = False
@property
def rank(self):
return self.synchronization_harness.rank
def run(self, *args, **kwargs):
try:
self._run(*args, **kwargs)
except Exception as e:
self.shutdown()
print 'Safely Shutting down after exception...'
print traceback.format_exc()
raise e
def _run(self, dt, tf, t0=0., synchronization_harness=None):
'''Main iteration control loop for simulation
The time step selection must be approximately of the same order as dv
for the internal populations, if the 'approx' time stepping method is
selected.
Parameters
----------
t0 : float (default=0.)
Network start time (unit=seconds), passed to initialize call.
tf : float (default=.1)
Network end time (unit=seconds).
dt : float (default=0.001)
Time step (unit=seconds).
'''
if synchronization_harness is None:
self.synchronization_harness = utilities.DefaultSynchronizationHarness()
else:
self.synchronization_harness = synchronization_harness
self.firing_rate_organizer = FiringRateOrganizer(self.synchronization_harness)
self.dt = dt
self.t0 = t0
if not self.progress is None: self.progress.set_network(self)
# Initialize:
start_time = time.time()
self.tf = tf
self.ti = 0
if not self.progress is None: self.progress.initialize()
self.synchronization_harness.initialize(self.ti)
# Initialize populations:
for gid, p in enumerate(self.population_list):
p.initialize()
if self.synchronization_harness.gid_to_rank(gid) == self.rank:
self.firing_rate_organizer.push(self.ti, gid, p.curr_firing_rate)
self.synchronization_harness.update(self.ti, self.firing_rate_organizer.firing_rate_dict_internal.setdefault(self.ti, {}))
# Initialize connections:
for c in self.connection_list:
c.initialize()
for p in self.population_list:
try:
p.initialize_total_input_dict()
except AttributeError:
pass
self.initialization_time = time.time() - start_time
# Run
start_time = time.time()
while self.t < self.tf:
self.update()
self.run_time = time.time() - start_time
self.synchronization_harness.finalize()
self.shutdown()
self.run_callback(self)
def shutdown(self):
for c in self.connection_list:
c.shutdown()
for p in self.population_list:
p.shutdown()
@property
def t(self): return self.t0+self.ti*self.dt
def update(self):
if not self.progress is None: self.progress.update()
self.firing_rate_organizer.drop(self.ti)
self.ti += 1
logger.info( 'time: %s' % self.t)
for gid, p in enumerate(self.population_list):
if self.synchronization_harness.gid_to_rank(gid) == self.rank:
p.update()
self.firing_rate_organizer.push(self.ti, gid, p.curr_firing_rate)
self.synchronization_harness.update(self.ti, self.firing_rate_organizer.firing_rate_dict_internal.setdefault(self.ti, {}))
for c in self.connection_list:
c.update()
self.update_callback(self)
if self.kill_signal:
raise StopSimulation()
def to_dict(self, organization='sparse_adjacency_matrix'):
population_list = [p.to_dict() for p in self.population_list]
connection_list = [c.to_dict() for c in self.connection_list]
if organization == 'sparse_adjacency_matrix':
data_dict = {'population_list':population_list,
'connection_list':connection_list,
'class':self.__class__.__name__,
'module':__name__}
return data_dict
else:
raise NotImplementedError
def to_json(self, fh=None, **kwargs):
'''Save the contents of the InternalPopultion to json'''
data_dict = self.to_dict()
indent = kwargs.pop('indent',2)
if fh is None:
return json.dumps(data_dict, indent=indent, **kwargs)
else:
return json.dump(data_dict, fh, indent=indent, **kwargs)
def copy(self):
return Network(**self.to_dict())
def get_curr_firing_rate(self, gid):
return self.firing_rate_organizer.pull(self.ti,gid)
def get_firing_rate(self, gid, t):
return self.population_list[gid].firing_rate(t)
def to_df(self):
pd.options.display.max_columns = 999
df_list = [p.to_df() for p in self.population_list]
return reorder_df_columns(pd.concat(df_list), ['class', 'module'])
def get_total_flux_matrix(self, internal_population, dt):
# Protect memory state of population and network:
population_ind = self.population_list.index(internal_population)
new_network = self.copy()
new_network.dt = dt
new_network.t0 = 0
new_network.ti = 0
new_internal_population = new_network.population_list[population_ind]
new_internal_population.initialize()
new_internal_population.initialize_total_input_dict()
return new_internal_population.get_total_flux_matrix()
@property
def external_population_list(self):
return [p for p in self.population_list if isinstance(p, ExternalPopulation)]
@property
def internal_population_list(self):
return [p for p in self.population_list if isinstance(p, InternalPopulation)]