def __init__(self, geometry, name=None, copied=False):
"""
:param geometry: population geometry as tuple. It must correspond to the image size and be fixed through the whole simulation.
:param name: unique name of the population (optional).
"""
# Check geometry
if isinstance(geometry, int) or len(geometry) == 1:
Global._error(
'The geometry of an ImagePopulation should be 2D (grayscale) or 3D (color).'
)
if len(geometry) == 3 and (geometry[2] != 3 and geometry[2] != 1):
Global._error(
'The third dimension of an ImagePopulation should be either 1 (grayscale) or 3 (color).'
)
if len(geometry) == 3 and geometry[2] == 1:
geometry = (int(geometry[0]), int(geometry[1]))
# Create the population
Population.__init__(self,
geometry=geometry,
name=name,
neuron=Neuron(parameters="r = 0.0"),
copied=copied)
def __init__(self, spike_times, name=None):
if not isinstance(spike_times, list):
Global._error('in SpikeSourceArray, spike_times must be a Python list.')
exit(0)
if isinstance(spike_times[0], list): # several neurons
nb_neurons = len(spike_times)
else: # a single Neuron
nb_neurons = 1
spike_times = [ spike_times ]
# Create a fake neuron just to be sure the description has the correct parameters
neuron = Neuron(
parameters="""
spike_times = 0.0
""",
equations="",
spike=" t == spike_times",
reset="",
name="Spike source",
description="Spikes source array."
)
Population.__init__(self, geometry=nb_neurons, neuron=neuron, name=name)
# Do some sorting to save C++ complexity
times = []
for neur_times in spike_times:
times.append(sorted(list(set(neur_times)))) # suppress doublons and sort
self.init['spike_times'] = times
def __init__(self, geometry, name=None):
"""
*Parameters*:
* *geometry*: population geometry as tuple. It must correspond to the image size and be fixed through the whole simulation.
* If the geometry is 2D, it corresponds to the (height, width) of the image. Only the luminance of the pixels will be represented (grayscale image).
* If the geometry is 3D, the third dimension can be either 1 (grayscale) or 3 (color).
If the third dimension is 3, each will correspond to the RGB values of the pixels.
.. warning::
Due to the indexing system of Numpy, a 640*480 image should be fed into a (480, 640) or (480, 640, 3) population.
* *name*: unique name of the population (optional).
"""
# Check geometry
if isinstance(geometry, int) or len(geometry)==1:
Global._error('The geometry of an ImagePopulation should be 2D (grayscale) or 3D (color).')
exit(0)
if len(geometry)==3 and (geometry[2]!=3 and geometry[2]!=1):
Global._error('The third dimension of an ImagePopulation should be either 1 (grayscale) or 3 (color).')
exit(0)
if len(geometry)==3 and geometry[2]==1:
geometry = (geometry[0], geometry[1])
# Create the population
Population.__init__(self, geometry = geometry, name=name, neuron = Neuron(parameters="r = 0.0") )
def __setattr__(self, name, value):
" Method called when setting an attribute."
Population.__setattr__(self, name, value)
if name in ['rates', 'corr', 'tau'] and hasattr(self, 'initialized'):
# Correction of mu and sigma everytime r, c or tau is changed
self.mu, self.sigma = self._rectify(self.rates, self.corr,
self.tau)
def __init__(self, geometry, name=None, copied=False):
"""
*Parameters*:
* *geometry*: population geometry as tuple. It must correspond to the image size and be fixed through the whole simulation.
* *name*: unique name of the population (optional).
About the geometry:
* If the geometry is 2D, it corresponds to the (height, width) of the image. Only the luminance of the pixels will be represented (grayscale image).
* If the geometry is 3D, the third dimension can be either 1 (grayscale) or 3 (color).
If the third dimension is 3, each will correspond to the RGB values of the pixels.
.. warning::
Due to the indexing system of Numpy, a 640*480 image should be fed into a (480, 640) or (480, 640, 3) population.
"""
# Check geometry
if isinstance(geometry, int) or len(geometry)==1:
Global._error('The geometry of an ImagePopulation should be 2D (grayscale) or 3D (color).')
if len(geometry)==3 and (geometry[2]!=3 and geometry[2]!=1):
Global._error('The third dimension of an ImagePopulation should be either 1 (grayscale) or 3 (color).')
if len(geometry)==3 and geometry[2]==1:
geometry = (int(geometry[0]), int(geometry[1]))
# Create the population
Population.__init__(self, geometry = geometry, name=name, neuron = Neuron(parameters="r = 0.0"), copied=copied)
def __init__(self, population, name=None, scaling=1.0, refractory=None, copied=False):
"""
:param population: the Population to convert. Its neuron type must be spiking.
:param name: the (optional) name of the hybrid population.
:param scaling: the scaling of the firing rate. Defines what a rate ``r`` of 1.0 means in Hz (default: 1.0).
:param refractory: a refractory period in ms to ensure the ISI is not too high (default: None)
"""
self.population = population
self.refractory_init = refractory
if not self.population.neuron_type.description['type'] == 'rate':
Global._error('the population ' + self.population.name + ' must contain rate-coded neurons.')
# Create the description, but it will not be used for generation
Population.__init__(
self,
geometry = self.population.geometry,
name=name,
neuron = Neuron(
parameters="""
scaling = %(scaling)s : population
""" % {'scaling': scaling} ,
equations="""
p = Uniform(0.0, 1.0)
rates = p
""",
spike="rates>p",
refractory=refractory
) ,
copied=self.copied
)
self._specific = True
def __init__(self, population, name=None, scaling=1.0, refractory=None):
"""
*Parameters*:
* **population**: the Population to convert. Its neuron type must be spiking.
* **name**: the (optional) name of the hybrid population.
* **scaling**: the scaling of the firing rate. Defines what a rate ``r`` of 1.0 means in Hz (default: 1.0).
* **refractory**: a refractory period in ms to ensure the ISI is not too high (default: None)
"""
self.population = population
if not self.population.neuron_type.description['type'] == 'rate':
Global._error('the population ' + self.population.name + ' must contain rate-coded neurons.')
# Create the description, but it will not be used for generation
Population.__init__(
self,
geometry = self.population.geometry,
name=name,
neuron = Neuron(
parameters="""
scaling = %(scaling)s : population
""" % {'scaling': scaling} ,
equations="""
p = Uniform(0.0, 1.0)
rates = p
""",
spike="rates>p",
refractory=refractory
)
)
self._specific = True
def __init__(self, spike_times, name=None):
if not isinstance(spike_times, list):
Global._error('In a SpikeSourceArray, spike_times must be a Python list.')
if isinstance(spike_times[0], list): # several neurons
nb_neurons = len(spike_times)
else: # a single Neuron
nb_neurons = 1
spike_times = [ spike_times ]
# Create a fake neuron just to be sure the description has the correct parameters
neuron = Neuron(
parameters="""
spike_times = 0.0 : int
""",
equations="",
spike=" t == spike_times",
reset="",
name="Spike source",
description="Spikes source array."
)
Population.__init__(self, geometry=nb_neurons, neuron=neuron, name=name)
self.init['spike_times'] = spike_times
def __setattr__(self, name, value):
if name == 'spike_times':
if self.initialized:
self.cyInstance.set_spike_times(value)
else:
object.__setattr__(self, name, value)
else:
Population.__setattr__(self, name, value)
def __init__(self, geometry, neuron, name=None, copied=False):
"""
Initialization, receive default arguments of Population objects.
"""
Population.__init__(self,
geometry=geometry,
neuron=neuron,
name=name,
stop_condition=None,
storage_order='post_to_pre',
copied=copied)
def __setattr__(self, name, value):
if name == 'spike_times':
if not isinstance(value[0], list): # several neurons
value = [ value ]
if not len(value) == self.size:
Global._error('SpikeSourceArray: the size of the spike_times attribute must match the number of neurons in the population.')
self.init['spike_times'] = value # when reset is called
if self.initialized:
self.cyInstance.set_spike_times(self._sort_spikes(value))
else:
Population.__setattr__(self, name, value)
def __setattr__(self, name, value):
if name == 'spike_times':
if not isinstance(value[0], list): # several neurons
value = [value]
if not len(value) == self.size:
Global._error(
'SpikeSourceArray: the size of the spike_times attribute must match the number of neurons in the population.'
)
self.init['spike_times'] = value # when reset is called
if self.initialized:
self.cyInstance.set_spike_times(self._sort_spikes(value))
else:
Population.__setattr__(self, name, value)
def __getattr__(self, name):
if name == 'schedule':
if self.initialized:
return Global.config['dt'] * self.cyInstance.get_schedule()
else:
return self.init['schedule']
elif name == 'rates':
if self.initialized:
if len(self.geometry) > 1:
# unflatten the data
flat_values = self.cyInstance.get_rates()
values = np.zeros( tuple( [len(self.schedule)] + list(self.geometry) ) )
for x in range(len(self.schedule)):
values[x] = np.reshape( flat_values[x], self.geometry)
return values
else:
return self.cyInstance.get_rates()
else:
return self.init['rates']
elif name == 'period':
if self.initialized:
return self.cyInstance.get_period() * Global.config['dt']
else:
return self.init['period']
else:
return Population.__getattribute__(self, name)
def __getattr__(self, name):
if name == 'schedule':
if self.initialized:
return Global.config['dt'] * self.cyInstance.get_schedule()
else:
return self.init['schedule']
elif name == 'rates':
if self.initialized:
if len(self.geometry) > 1:
# unflatten the data
flat_values = self.cyInstance.get_rates()
values = np.zeros(
tuple([len(self.schedule)] + list(self.geometry)))
for x in range(len(self.schedule)):
values[x] = np.reshape(flat_values[x], self.geometry)
return values
else:
return self.cyInstance.get_rates()
else:
return self.init['rates']
elif name == 'period':
if self.initialized:
return self.cyInstance.get_period() * Global.config['dt']
else:
return self.init['period']
else:
return Population.__getattribute__(self, name)
def __getattr__(self, name):
if name == 'spike_times':
if self.initialized:
return [ [Global.config['dt']*time for time in neur] for neur in self.cyInstance.get_spike_times()]
else:
return self.init['spike_times']
else:
return Population.__getattribute__(self, name)
def __getattr__(self, name):
if name == 'spike_times':
if self.initialized:
return self.cyInstance.get_spike_times()
else:
return object.__getattribute__(self, name)
else:
return Population.__getattribute__(self, name)
def __getattr__(self, name):
if name == 'spike_times':
if self.initialized:
return [[Global.config['dt'] * time for time in neur]
for neur in self.cyInstance.get_spike_times()]
else:
return self.init['spike_times']
else:
return Population.__getattribute__(self, name)
def __init__(self, geometry, rates, corr, tau, name=None, refractory=None):
"""
*Parameters*:
* **geometry**: population geometry as tuple.
* **rates**: rate in Hz of the population (must be a positive float)
* **corr**: total correlation strength (float in [0, 1])
* **tau**: correlation time constant in ms.
* **name**: unique name of the population (optional).
* **refractory**: refractory period in ms (careful: may break the correlation)
"""
# Store parameters
self.rates = float(rates)
self.corr = corr
# Correction of mu and sigma
mu, sigma = self._rectify(self.rates, self.corr, tau)
# Create the neuron
corr_neuron = Neuron(
parameters = """
tau = %(tau)s : population
mu = %(mu)s : population
sigma = %(sigma)s : population
""" % {'tau': tau, 'mu': mu, 'sigma': sigma},
equations = """
x += dt*(mu - x)/tau + sqrt(dt/tau) * sigma * Normal(0., 1.) : population, init=%(mu)s
p = Uniform(0.0, 1.0) * 1000.0 / dt
""" % {'mu': mu},
spike = "p < x",
refractory=refractory,
name="HomogeneousCorrelated",
description="Homogeneous correlated spike trains."
)
Population.__init__(self, geometry=geometry, neuron=corr_neuron, name=name)
def __setattr__(self, name, value):
if name == 'schedule':
if self.initialized:
self.cyInstance.set_schedule( np.array(value) / Global.config['dt'] )
else:
self.init['schedule'] = value
elif name == 'rates':
if self.initialized:
if len(value.shape) > 2:
# we need to flatten the provided data
flat_values = value.reshape( (value.shape[0], self.size) )
self.cyInstance.set_rates( flat_values )
else:
self.cyInstance.set_rates( value )
else:
self.init['rates'] = value
elif name == "period":
if self.initialized:
self.cyInstance.set_period(int(value /Global.config['dt']))
else:
self.init['period'] = value
else:
Population.__setattr__(self, name, value)
def __setattr__(self, name, value):
if name == 'schedule':
if self.initialized:
self.cyInstance.set_schedule(
np.array(value) / Global.config['dt'])
else:
self.init['schedule'] = value
elif name == 'rates':
if self.initialized:
if len(value.shape) > 2:
# we need to flatten the provided data
flat_values = value.reshape((value.shape[0], self.size))
self.cyInstance.set_rates(flat_values)
else:
self.cyInstance.set_rates(value)
else:
self.init['rates'] = value
elif name == "period":
if self.initialized:
self.cyInstance.set_period(int(value / Global.config['dt']))
else:
self.init['period'] = value
else:
Population.__setattr__(self, name, value)
def __init__(self, geometry, name=None, rates=None, target=None, parameters=None, refractory=None):
"""
*Parameters*:
* *geometry*: population geometry as tuple.
* *name*: unique name of the population (optional).
* *rates*: mean firing rate of each neuron. It can be a single value (e.g. 10.0) or an equation (as string).
* *parameters*: additional parameters which can be used in the *rates* equation.
* *refractory*: refractory period in ms.
"""
if rates == None and target==None:
Global._error('A PoissonPopulation must define either rates or target.')
exit(0)
if target is not None: # hybrid population
# Create the neuron
poisson_neuron = Neuron(
parameters = """
%(params)s
""" % {'params': parameters if parameters else ''},
equations = """
rates = sum(%(target)s)
p = Uniform(0.0, 1.0) * 1000.0 / dt
_sum_%(target)s = 0.0
""" % {'target': target},
spike = """
p < rates
""",
refractory=refractory,
name="Hybrid",
description="Hybrid spiking neuron emitting spikes according to a Poisson distribution at a frequency determined by the weighted sum of inputs."
)
elif isinstance(rates, str):
# Create the neuron
poisson_neuron = Neuron(
parameters = """
%(params)s
""" % {'params': parameters if parameters else ''},
equations = """
rates = %(rates)s
p = Uniform(0.0, 1.0) * 1000.0 / dt
_sum_exc = 0.0
""" % {'rates': rates},
spike = """
p < rates
""",
refractory=refractory,
name="Poisson",
description="Spiking neuron with spikes emitted according to a Poisson distribution."
)
elif isinstance(rates, np.ndarray):
poisson_neuron = Neuron(
parameters = """
rates = 10.0
""",
equations = """
p = Uniform(0.0, 1.0) * 1000.0 / dt
""",
spike = """
p < rates
""",
refractory=refractory,
name="Poisson",
description="Spiking neuron with spikes emitted according to a Poisson distribution."
)
else:
poisson_neuron = Neuron(
parameters = """
rates = %(rates)s
""" % {'rates': rates},
equations = """
p = Uniform(0.0, 1.0) * 1000.0 / dt
""",
spike = """
p < rates
""",
refractory=refractory,
name="Poisson",
description="Spiking neuron with spikes emitted according to a Poisson distribution."
)
Population.__init__(self, geometry=geometry, neuron=poisson_neuron, name=name)
if isinstance(rates, np.ndarray):
self.rates = rates
def __init__(self, geometry, neuron, name=None, copied=False):
"""
Initialization, receive default arguments of Population objects.
"""
Population.__init__(self, geometry=geometry, neuron=neuron, name=name, stop_condition=None, storage_order='post_to_pre', copied=copied)
def _create_adaptive(self):
# Create the description, but it will not be used for generation
Population.__init__(
self,
geometry = self.population.geometry,
name=self.name,
neuron = Neuron(
parameters="""
window = %(window)s : population
scaling = %(scaling)s : population
smooth = %(smooth)s : population
""" % {'window': self.window, 'scaling': self.scaling, 'smooth': self.smooth} ,
equations="r = 0.0"
)
)
# Generate specific code
omp_code = "#pragma omp parallel for private(pop%(id)s_nb, pop%(id)s_out)" if Global.config['num_threads'] > 1 else ""
code = """#pragma once
#include "pop%(id_pre)s.hpp"
extern PopStruct%(id_pre)s pop%(id_pre)s;
struct PopStruct%(id)s{
// Number of neurons
int size;
bool _active;
// Local parameter window
double window ;
std::vector< double > ad_window ;
// Global parameter scaling
double scaling ;
// Global parameter smooth
double smooth ;
// Local variable r
std::vector< double > r ;
std::vector< std::vector< double > > recorded_r ;
bool record_r ;
// Store the last spikes
std::vector< std::vector<long int> > last_spikes;
std::vector< double > isi;
int get_size() { return size; }
void set_size(int value) { size = value; }
bool is_active() { return _active; }
void set_active(bool value) { _active = value; }
double get_window() { return window; }
void set_window(double value) { window = value; }
double get_scaling() { return scaling; }
void set_scaling(double value) { scaling = value; }
double get_smooth() { return smooth; }
void set_smooth(double value) { smooth = value; }
std::vector< double > get_r() { return r; }
void set_r(std::vector< double > value) { r = value; }
double get_single_r(int rk) { return r[rk]; }
void set_single_r(int rk, double value) { r[rk] = value; }
void init_population() {
size = %(size)s;
_active = true;
last_spikes = std::vector< std::vector<long int> >(size, std::vector<long int>());
ad_window = std::vector< double >(size, window);
isi = std::vector< double >(size, 10000.0);
}
void update_rng() {
}
void update() {
// Updating the local variables of Spike2Rate population %(id)s
if(_active){
int pop%(id)s_nb, pop%(id)s_out;
%(omp_code)s
for(int i = 0; i < size; i++){
// Increase when spiking
if (pop%(id_pre)s.last_spike[i] == t-1){
if(last_spikes[i].size() > 0)
isi[i] = double(t-1 - last_spikes[i][last_spikes[i].size()-1]);
else
isi[i] = 10000.0;
last_spikes[i].push_back(t-1);
}
pop%(id)s_nb = 0;
pop%(id)s_out = -1;
for(int j=0; j < last_spikes[i].size(); j++){
if(last_spikes[i][j] >= t -1 - (long int)(ad_window[i]/dt) ){
pop%(id)s_nb++;
}
else{
pop%(id)s_out = j;
}
}
if (pop%(id)s_out > -1){
last_spikes[i].erase(last_spikes[i].begin(), last_spikes[i].begin()+pop%(id)s_out);
}
r[i] += dt*(1000.0/scaling / ad_window[i] * double(pop%(id)s_nb) - r[i] ) / smooth;
//ad_window[i] = clip(5.0*isi[i], 20.0*dt, window) ;
ad_window[i] += dt * (clip(5.0*isi[i], 20.0*dt, window) - ad_window[i])/100.0;
if (i==0)
std::cout << ad_window[i] << " " << isi[i] << std::endl;
}
}
}
};
""" % {'id' : self.id, 'id_pre': self.population.id, 'omp_code': omp_code, 'size': self.size}
return code
def _create_isi(self):
# Create the description, but it will not be used for generation
Population.__init__(
self,
geometry = self.population.geometry,
name=self.name,
neuron = Neuron(
parameters="""
cut = %(cut)s : population
scaling = %(scaling)s : population
smooth = %(smooth)s : population
""" % {'cut': self.cut, 'scaling': self.scaling, 'smooth': self.smooth} ,
equations="r = 0.0"
)
)
# Generate specific code
omp_code = "#pragma omp parallel for" if Global.config['num_threads'] > 1 else ""
code = """#pragma once
#include "pop%(id_pre)s.hpp"
extern PopStruct%(id_pre)s pop%(id_pre)s;
struct PopStruct%(id)s{
// Number of neurons
int size;
bool _active;
// Global parameter cut
double cut ;
// Global parameter scaling
double scaling ;
// Global parameter smooth
double smooth ;
// Local variable r
std::vector< double > r ;
// Store the last spike
std::vector< long int > last_spike;
std::vector< double > isi;
std::vector< double > support;
int get_size() { return size; }
void set_size(int value) { size = value; }
bool is_active() { return _active; }
void set_active(bool value) { _active = value; }
double get_cut() { return cut; }
void set_cut(double value) { cut = value; }
double get_scaling() { return scaling; }
void set_scaling(double value) { scaling = value; }
double get_smooth() { return smooth; }
void set_smooth(double value) { smooth = value; }
std::vector< double > get_r() { return r; }
void set_r(std::vector< double > value) { r = value; }
double get_single_r(int rk) { return r[rk]; }
void set_single_r(int rk, double value) { r[rk] = value; }
void init_population() {
size = %(size)s;
_active = true;
last_spike = std::vector< long int >(size, -10000L);
isi = std::vector< double >(size, 10000.0);
support = std::vector< double >(size, 10000.0);
}
void update_rng() {
}
void update() {
// Updating the local variables of Spike2Rate population %(id)s
if(_active){
%(omp_code)s
for(int i = 0; i < size; i++){
// Increase when spiking
if (pop%(id_pre)s.last_spike[i] == t-1){
isi[i] = double(t-1 - last_spike[i]);
support[i] = double(t-1 - last_spike[i]);
last_spike[i] = t-1;
}
else if( double(t - last_spike[i]) <= isi[i]){
// do nothing
}
else if( double(t - last_spike[i]) <= cut*isi[i]){
support[i] += 1.0 ;
}
else{
support[i] = 10000.0 ;
}
r[i] += dt*(1000.0/scaling/support[i]/dt - r[i])/smooth;
}
}
}
};
""" % {'id' : self.id, 'id_pre': self.population.id, 'omp_code': omp_code, 'size': self.size }
return code
def _create_adaptive(self):
# Create the description, but it will not be used for generation
Population.__init__(self,
geometry=self.population.geometry,
name=self.name,
neuron=Neuron(parameters="""
window = %(window)s : population
scaling = %(scaling)s : population
smooth = %(smooth)s : population
""" % {
'window': self.window,
'scaling': self.scaling,
'smooth': self.smooth
},
equations="r = 0.0"),
copied=self.copied)
# Generate specific code
omp_code = "#pragma omp parallel for private(pop%(id)s_nb, pop%(id)s_out)" if Global.config[
'num_threads'] > 1 else ""
code = """#pragma once
#include "pop%(id_pre)s.hpp"
extern PopStruct%(id_pre)s pop%(id_pre)s;
struct PopStruct%(id)s{
// Number of neurons
int size;
bool _active;
// Local parameter window
%(float_prec)s window ;
std::vector< %(float_prec)s > ad_window ;
// Global parameter scaling
%(float_prec)s scaling ;
// Global parameter smooth
%(float_prec)s smooth ;
// Local variable r
std::vector< %(float_prec)s > r ;
std::vector< std::vector< %(float_prec)s > > recorded_r ;
bool record_r ;
// Store the last spikes
std::vector< std::vector<long int> > last_spikes;
std::vector< %(float_prec)s > isi;
int get_size() { return size; }
void set_size(int value) { size = value; }
bool is_active() { return _active; }
void set_active(bool value) { _active = value; }
%(float_prec)s get_window() { return window; }
void set_window(%(float_prec)s value) { window = value; }
%(float_prec)s get_scaling() { return scaling; }
void set_scaling(%(float_prec)s value) { scaling = value; }
%(float_prec)s get_smooth() { return smooth; }
void set_smooth(%(float_prec)s value) { smooth = value; }
std::vector< %(float_prec)s > get_r() { return r; }
void set_r(std::vector< %(float_prec)s > value) { r = value; }
%(float_prec)s get_single_r(int rk) { return r[rk]; }
void set_single_r(int rk, %(float_prec)s value) { r[rk] = value; }
void init_population() {
size = %(size)s;
_active = true;
last_spikes = std::vector< std::vector<long int> >(size, std::vector<long int>());
ad_window = std::vector< %(float_prec)s >(size, window);
isi = std::vector< %(float_prec)s >(size, 10000.0);
}
void update_rng() {
}
void update() {
// Updating the local variables of Spike2Rate population %(id)s
if(_active){
int pop%(id)s_nb, pop%(id)s_out;
%(omp_code)s
for(int i = 0; i < size; i++){
// Increase when spiking
if (pop%(id_pre)s.last_spike[i] == t-1){
if(last_spikes[i].size() > 0)
isi[i] = %(float_prec)s(t-1 - last_spikes[i][last_spikes[i].size()-1]);
else
isi[i] = 10000.0;
last_spikes[i].push_back(t-1);
}
pop%(id)s_nb = 0;
pop%(id)s_out = -1;
for(int j=0; j < last_spikes[i].size(); j++){
if(last_spikes[i][j] >= t -1 - (long int)(ad_window[i]/dt) ){
pop%(id)s_nb++;
}
else{
pop%(id)s_out = j;
}
}
if (pop%(id)s_out > -1){
last_spikes[i].erase(last_spikes[i].begin(), last_spikes[i].begin()+pop%(id)s_out);
}
r[i] += dt*(1000.0/scaling / ad_window[i] * %(float_prec)s(pop%(id)s_nb) - r[i] ) / smooth;
//ad_window[i] = clip(5.0*isi[i], 20.0*dt, window) ;
ad_window[i] += dt * (clip(5.0*isi[i], 20.0*dt, window) - ad_window[i])/100.0;
if (i==0)
std::cout << ad_window[i] << " " << isi[i] << std::endl;
}
}
}
};
""" % {
'id': self.id,
'id_pre': self.population.id,
'omp_code': omp_code,
'size': self.size,
'float_prec': Global.config['precision']
}
return code
def __init__(self,
populations=None,
bold_model=balloon_RN,
mapping={'I_CBF': 'r'},
scale_factor=None,
normalize_input=None,
recorded_variables=None,
start=False,
net_id=0,
copied=False):
"""
:param populations: list of recorded populations.
:param bold_model: computational model for BOLD signal defined as a BoldModel class/object (see ANNarchy.extensions.bold.PredefinedModels for more predefined examples). Default is `balloon_RN`.
:param mapping: mapping dictionary between the inputs of the BOLD model (`I_CBF` for single inputs, `I_CBF` and `I_CMRO2` for double inputs in the provided examples) and the variables of the input populations. By default, `{'I_CBF': 'r'}` maps the firing rate `r` of the input population(s) to the variable `I_CBF` of the BOLD model.
:param scale_factor: list of float values to allow a weighting of signals between populations. By default, the input signal is weighted by the ratio of the population size to all populations within the recorded region.
:param normalize_input: list of integer values which represent a optional baseline per population. The input signals will require an additional normalization using a baseline value. A value different from 0 represents the time period for determing this baseline in milliseconds (biological time).
:param recorded_variables: which variables of the BOLD model should be recorded? (by default, the output variable of the BOLD model is added, e.g. ["BOLD"] for the provided examples).
"""
self.net_id = net_id
# instantiate if necessary, please note
# that population will make a deepcopy on this objects
if inspect.isclass(bold_model):
bold_model = bold_model()
# for reporting
bold_model._model_instantiated = True
# The usage of [] as default arguments in the __init__ call lead to strange side effects.
# We decided therefore to use None as default and create the lists locally.
if populations is None:
Global._error(
"Either a population or a list of populations must be provided to the BOLD monitor (populations=...)"
)
if scale_factor is None:
scale_factor = []
if normalize_input is None:
normalize_input = []
if recorded_variables is None:
recorded_variables = []
# argument check
if not (isinstance(populations, list)):
populations = [populations]
if not (isinstance(scale_factor, list)):
scale_factor = [scale_factor] * len(populations)
if not (isinstance(normalize_input, list)):
normalize_input = [normalize_input] * len(populations)
if isinstance(recorded_variables, str):
recorded_variables = [recorded_variables]
if len(scale_factor) > 0:
if len(populations) != len(scale_factor):
Global._error(
"BoldMonitor: Length of scale_factor must be equal to number of populations"
)
if len(normalize_input) > 0:
if len(populations) != len(normalize_input):
Global._error(
"BoldMonitor: Length of normalize_input must be equal to number of populations"
)
# Check mapping
for target, input_var in mapping.items():
if not target in bold_model._inputs:
Global._error("BoldMonitor: the key " + target +
" of mapping is not part of the BOLD model.")
# Check recorded variables
if len(recorded_variables) == 0:
recorded_variables = bold_model._output
else:
# Add the output variables (and remove doublons)
l1 = bold_model._output
l2 = [recorded_variables] if isinstance(
recorded_variables, str) else recorded_variables
recorded_variables = list(set(l2 + l1))
recorded_variables.sort()
if not copied:
# Add the container to the object management
Global._network[0]['extensions'].append(self)
self.id = len(Global._network[self.net_id]['extensions'])
# create the population
self._bold_pop = Population(1,
neuron=bold_model,
name=bold_model.name)
self._bold_pop.enabled = start
# create the monitor
self._monitor = Monitor(self._bold_pop,
recorded_variables,
start=start)
# create the projection(s)
self._acc_proj = []
if len(scale_factor) == 0:
pop_overall_size = 0
for _, pop in enumerate(populations):
pop_overall_size += pop.size
# the conductance is normalized between [0 .. 1]. This scale factor
# should balance different population sizes
for _, pop in enumerate(populations):
scale_factor_conductance = float(
pop.size) / float(pop_overall_size)
scale_factor.append(scale_factor_conductance)
if len(normalize_input) == 0:
normalize_input = [0] * len(populations)
# TODO: can we check if users used NormProjections? If not, this will crash ...
for target, input_var in mapping.items():
for pop, scale, normalize in zip(populations, scale_factor,
normalize_input):
tmp_proj = AccProjection(pre=pop,
post=self._bold_pop,
target=target,
variable=input_var,
scale_factor=scale,
normalize_input=normalize)
tmp_proj.connect_all_to_all(weights=1.0)
self._acc_proj.append(tmp_proj)
else:
# Add the container to the object management
Global._network[net_id]['extensions'].append(self)
self.id = len(Global._network[self.net_id]['extensions'])
# instances are assigned by the copying instance
self._bold_pop = None
self._monitor = None
self._acc_proj = []
self.name = "bold_monitor"
# store arguments for copy
self._populations = populations
self._mapping = mapping
self._recorded_variables = recorded_variables
self._bold_model = bold_model
self._start = start
# Finalize initialization
self._initialized = True if not copied else False
def __init__(self, geometry, name=None, rates=None, target=None, parameters=None, refractory=None):
"""
*Parameters*:
* **geometry**: population geometry as tuple.
* **name**: unique name of the population (optional).
* **rates**: mean firing rate of each neuron. It can be a single value (e.g. 10.0) or an equation (as string).
* **target**: the mean firing rate will be the weighted sum of inputs having this target name (e.g. "exc").
* **parameters**: additional parameters which can be used in the *rates* equation.
* **refractory**: refractory period in ms.
"""
if rates is None and target is None:
Global._error('A PoissonPopulation must define either rates or target.')
if target is not None: # hybrid population
# Create the neuron
poisson_neuron = Neuron(
parameters = """
%(params)s
""" % {'params': parameters if parameters else ''},
equations = """
rates = sum(%(target)s)
p = Uniform(0.0, 1.0) * 1000.0 / dt
_sum_%(target)s = 0.0
""" % {'target': target},
spike = """
p < rates
""",
refractory=refractory,
name="Hybrid",
description="Hybrid spiking neuron emitting spikes according to a Poisson distribution at a frequency determined by the weighted sum of inputs."
)
elif isinstance(rates, str):
# Create the neuron
poisson_neuron = Neuron(
parameters = """
%(params)s
""" % {'params': parameters if parameters else ''},
equations = """
rates = %(rates)s
p = Uniform(0.0, 1.0) * 1000.0 / dt
_sum_exc = 0.0
""" % {'rates': rates},
spike = """
p < rates
""",
refractory=refractory,
name="Poisson",
description="Spiking neuron with spikes emitted according to a Poisson distribution."
)
elif isinstance(rates, np.ndarray):
poisson_neuron = Neuron(
parameters = """
rates = 10.0
""",
equations = """
p = Uniform(0.0, 1.0) * 1000.0 / dt
""",
spike = """
p < rates
""",
refractory=refractory,
name="Poisson",
description="Spiking neuron with spikes emitted according to a Poisson distribution."
)
else:
poisson_neuron = Neuron(
parameters = """
rates = %(rates)s
""" % {'rates': rates},
equations = """
p = Uniform(0.0, 1.0) * 1000.0 / dt
""",
spike = """
p < rates
""",
refractory=refractory,
name="Poisson",
description="Spiking neuron with spikes emitted according to a Poisson distribution."
)
Population.__init__(self, geometry=geometry, neuron=poisson_neuron, name=name)
if isinstance(rates, np.ndarray):
self.rates = rates
def __setattr__(self, name, value):
" Method called when setting an attribute."
Population.__setattr__(self, name, value)
if name in ['rates', 'corr', 'tau'] and hasattr(self, 'initialized'):
# Correction of mu and sigma everytime r, c or tau is changed
self.mu, self.sigma = self._rectify(self.rates, self.corr, self.tau)
def _create_isi(self):
# Create the description, but it will not be used for generation
Population.__init__(self,
geometry=self.population.geometry,
name=self.name,
neuron=Neuron(parameters="""
cut = %(cut)s : population
scaling = %(scaling)s : population
smooth = %(smooth)s : population
""" % {
'cut': self.cut,
'scaling': self.scaling,
'smooth': self.smooth
},
equations="r = 0.0"),
copied=self.copied)
# Generate specific code
omp_code = "#pragma omp parallel for" if Global.config[
'num_threads'] > 1 else ""
code = """#pragma once
#include "pop%(id_pre)s.hpp"
extern PopStruct%(id_pre)s pop%(id_pre)s;
struct PopStruct%(id)s{
// Number of neurons
int size;
bool _active;
// Global parameter cut
%(float_prec)s cut;
// Global parameter scaling
%(float_prec)s scaling;
// Global parameter smooth
%(float_prec)s smooth ;
// Local variable r
std::vector< %(float_prec)s > r ;
// Store the last spike
std::vector< long int > last_spike;
std::vector< %(float_prec)s > isi;
std::vector< %(float_prec)s > support;
int get_size() { return size; }
void set_size(int value) { size = value; }
bool is_active() { return _active; }
void set_active(bool value) { _active = value; }
%(float_prec)s get_cut() { return cut; }
void set_cut(%(float_prec)s value) { cut = value; }
%(float_prec)s get_scaling() { return scaling; }
void set_scaling(%(float_prec)s value) { scaling = value; }
%(float_prec)s get_smooth() { return smooth; }
void set_smooth(%(float_prec)s value) { smooth = value; }
std::vector< %(float_prec)s > get_r() { return r; }
void set_r(std::vector< %(float_prec)s > value) { r = value; }
%(float_prec)s get_single_r(int rk) { return r[rk]; }
void set_single_r(int rk, %(float_prec)s value) { r[rk] = value; }
void init_population() {
size = %(size)s;
_active = true;
last_spike = std::vector< long int >(size, -10000L);
isi = std::vector< %(float_prec)s >(size, 10000.0);
support = std::vector< %(float_prec)s >(size, 10000.0);
}
void update_rng() {
}
void update() {
// Updating the local variables of Spike2Rate population %(id)s
if(_active){
%(omp_code)s
for(int i = 0; i < size; i++){
// Increase when spiking
if (pop%(id_pre)s.last_spike[i] == t-1){
isi[i] = %(float_prec)s(t-1 - last_spike[i]);
support[i] = %(float_prec)s(t-1 - last_spike[i]);
last_spike[i] = t-1;
}
else if( %(float_prec)s(t - last_spike[i]) <= isi[i]){
// do nothing
}
else if( %(float_prec)s(t - last_spike[i]) <= cut*isi[i]){
support[i] += 1.0 ;
}
else{
support[i] = 10000.0 ;
}
r[i] += dt*(1000.0/scaling/support[i]/dt - r[i])/smooth;
}
}
}
};
""" % {
'id': self.id,
'id_pre': self.population.id,
'omp_code': omp_code,
'size': self.size,
'float_prec': Global.config['precision']
}
return code
def _create_window(self):
# Create the description, but it will not be used for generation
Population.__init__(
self,
geometry = self.population.geometry,
name=self.name,
neuron = Neuron(
parameters="""
window = %(window)s : population
scaling = %(scaling)s : population
smooth = %(smooth)s : population
""" % {'window': self.window, 'scaling': self.scaling, 'smooth': self.smooth} ,
equations="r = 0.0"
) ,
copied=self.copied
)
# Generate specific code
omp_code = "#pragma omp parallel for" if Global.config['num_threads'] > 1 else ""
code = """#pragma once
#include "pop%(id_pre)s.hpp"
extern PopStruct%(id_pre)s pop%(id_pre)s;
struct PopStruct%(id)s{
// Number of neurons
int size;
bool _active;
// Global parameter window
%(float_prec)s window ;
// Global parameter scaling
%(float_prec)s scaling ;
// Global parameter smooth
%(float_prec)s smooth ;
// Local variable r
std::vector< %(float_prec)s > r ;
// Store the last spikes
std::vector< std::vector<long int> > last_spikes;
int get_size() { return size; }
void set_size(int value) { size = value; }
bool is_active() { return _active; }
void set_active(bool value) { _active = value; }
%(float_prec)s get_window() { return window; }
void set_window(%(float_prec)s value) { window = value; }
%(float_prec)s get_scaling() { return scaling; }
void set_scaling(%(float_prec)s value) { scaling = value; }
%(float_prec)s get_smooth() { return smooth; }
void set_smooth(%(float_prec)s value) { smooth = value; }
std::vector< %(float_prec)s > get_r() { return r; }
void set_r(std::vector< %(float_prec)s > value) { r = value; }
%(float_prec)s get_single_r(int rk) { return r[rk]; }
void set_single_r(int rk, %(float_prec)s value) { r[rk] = value; }
void init_population() {
size = %(size)s;
_active = true;
last_spikes = std::vector< std::vector<long int> >(size, std::vector<long int>());
}
void update_rng() {
}
void update() {
// Updating the local variables of Spike2Rate population %(id)s
if(_active){
int pop%(id)s_nb, pop%(id)s_out;
%(omp_code)s
for(int i = 0; i < size; i++){
// Increase when spiking
if (pop%(id_pre)s.last_spike[i] == t-1){
last_spikes[i].push_back(t-1);
}
pop%(id)s_nb = 0;
pop%(id)s_out = -1;
for(int j=0; j < last_spikes[i].size(); j++){
if(last_spikes[i][j] >= t -1 - (long int)(window/dt) ){
pop%(id)s_nb++;
}
else{
pop%(id)s_out = j;
}
}
if (pop%(id)s_out > -1){
last_spikes[i].erase(last_spikes[i].begin());
}
r[i] += dt*(1000.0/scaling / window * %(float_prec)s(pop%(id)s_nb) - r[i] ) / smooth;
}
}
}
};
""" % {'id' : self.id, 'id_pre': self.population.id, 'omp_code': omp_code, 'size': self.size, 'float_prec': Global.config['precision']}
return code
