def main(argv=None):
if argv is None:
argv = sys.argv
(t0, tf, dt, paramFn, graphFn, outFn, abs_error, rel_error, save_full,
save_spikes, quiet, spike_thresh, refractory) = parse_args(argv)
# compute the number of steps required
Nstep = np.ceil(tf / dt)
if not quiet:
print "Loading parameters, graph, and setting up IC's"
my_params = prebotc.params(paramFn)
num_vertices, num_edges, graph_params = prebotc.graph(graphFn)
y, N = prebotc.ics(num_vertices, num_edges)
#y, N = prebotc.ics(num_vertices, num_edges, random=False)
# rhs of ODE with parameters evaluated
# f is the rhs with parameters evaluated
f = lambda t, y: prebotc.rhs(t, y, graph_params, my_params)
# data structure to output, timeseries or sparse raster
if save_full:
# all state variables
save_state = np.zeros((N, Nstep + 1))
elif save_spikes:
# just spike times, as sparse matrix
# not boolean because type conversion bug in matlab output
save_state = scipy.sparse.dok_matrix((num_vertices, Nstep + 1))
last_spike = np.ones(num_vertices) * (-np.inf)
else:
# timeseries of spikes
save_state = np.zeros((num_vertices, Nstep + 1))
r = scipy.integrate.ode(f)
r.set_initial_value(y, t0)
# # other integration methods
# # method 1: BDF
# r.set_integrator(
# 'vode',
# method='bdf',
# with_jacobian = False,
# order=3,
# rtol= rel_error,
# atol= abs_error
# )
# # method 2: Dormand-Price
# r.set_integrator(
# 'dopri5',
# rtol = rel_error,
# atol = abs_error
# )
# # method 3: VODE
# r.set_integrator('vode',
# rtol = rel_error,
# atol = abs_error)
if not quiet:
print "Running integration loop...."
t = time.time()
bar_updates = 100
widgets = [
progressbar.Bar('=', '[', ']'), ' ',
progressbar.Percentage()
]
bar = progressbar.ProgressBar(maxval=bar_updates, widgets=widgets)
bar.start()
j = 0
i = 0
while r.successful() and r.t < tf:
r.integrate(r.t + dt)
y = r.y.copy()
if save_full:
save_state[:, i] = y
elif save_spikes:
spikers = prebotc.spiking(y, num_vertices, spike_thresh)
for neur in spikers:
# only count if the new trigger occurs after reasonable delay
if dt * (float(i) - last_spike[neur]) > refractory:
save_state[neur, i] = 1
last_spike[neur] = i
else:
save_state[:, i] = prebotc.voltages(y, num_vertices)
i += 1
if not quiet:
if (i % np.floor(Nstep / bar_updates)) == 0:
bar.update(j)
j += 1
if not save_spikes:
save_state = save_state[:, 0:(i - 1)]
else:
save_state.resize((num_vertices, i - 1))
if not quiet:
bar.finish()
elapsed = time.time() - t
print "Done!\nElapsed: %1.2fs" % elapsed
# Time saving
t = time.time()
print "Saving output...."
if save_full:
save_str = 'full'
elif save_spikes:
save_str = 'spikes'
else:
save_str = 'V'
# save output
scipy.io.savemat(outFn,
mdict={
'Y': save_state,
'dt': dt,
't0': t0,
'tf': tf,
'paramFn': os.path.abspath(paramFn),
'graphFn': os.path.abspath(graphFn),
'absErr': abs_error,
'relErr': rel_error,
'saveStr': save_str,
'finalState': y
},
oned_as='column')
if not quiet:
elapsed = time.time() - t
print "Done!\nSave time: %1.2fs" % elapsed
def main(argv=None):
should_save = True
if argv is None:
argv = sys.argv
else:
should_save = False
(t0, tf, dt, param_fn, graph_fn, outFn, abs_error, rel_error, save_full,
save_spikes, quiet, spike_thresh, refractory, ic_str) = parse_args(argv)
# compute the number of steps required
Nstep = np.ceil(tf/dt)
if not quiet:
print("Loading parameters, graph, and setting up ICs")
my_params = prebotc.params(param_fn)
num_vertices, num_edges, graph_params = prebotc.graph(graph_fn)
#Checks to see if it neets to load a set of initial conditions or generate random ones
if ic_str == 'random':
if not quiet:
print('Setting random ICs')
y = prebotc.ics(num_vertices, num_edges)
elif ic_str == 'testing_ic':
if not quiet:
print('Setting random ICs')
y = prebotc.ics(num_vertices, num_edges, random = False)
else:
if not quiet:
print('Loading ICs from ' + ic_str)
y, graph_fn_loaded = prebotc.load_ics(ic_str)
if os.path.abspath(graph_fn_loaded) != os.path.abspath(graph_fn):
warnings.warn(('simulation is running on a graph which differs'
'from the ics'))
print(graph_fn_loaded)
print(graph_fn)
N = y.size
#y, N = prebotc.ics(num_vertices, num_edges, random=False)
# rhs of ODE with parameters evaluated
# f is the rhs with parameters evaluated
f = lambda t, y: prebotc.rhs(t, y,
graph_params,
my_params)
# data structure to output, timeseries or sparse raster
if save_full:
# all state variables
save_state = np.zeros((N, Nstep+1))
elif save_spikes:
# just spike times, as sparse matrix
# not boolean because type conversion bug in matlab output
save_state = scipy.sparse.dok_matrix((num_vertices, Nstep+1))
last_spike = np.ones(num_vertices) * (-np.inf)
else:
# timeseries of spikes
save_state = np.zeros((num_vertices, Nstep+1))
#Creates and sets integrator. Using VODE and backward differentiation formulas (BDF) as the method
#This is done since we are using a stiff DFQ, the default method is Adams (implicit, non-stiff)
r = scipy.integrate.ode(f)
r.set_initial_value(y, t0)
r.set_integrator(
'vode',
method='bdf',
atol = abs_error,
rtol = rel_error
)
#Visual representation of running progress
if not quiet:
print("Running integration loop....")
t = time.time()
bar_updates = 100
widgets=[progressbar.Bar('=', '[', ']'), ' ', progressbar.Percentage()]
bar = progressbar.ProgressBar(maxval=bar_updates, widgets=widgets)
bar.start()
j = 0
i = 0
while r.successful() and r.t < tf:
r.integrate(r.t + dt)
y = r.y.copy()
if save_full:
save_state[:, i] = y.flatten()
elif save_spikes:
spikers = prebotc.spiking(y, num_vertices, spike_thresh)
for neur in spikers:
# only count if the new trigger occurs after reasonable delay
if dt*( float(i) - last_spike[neur] ) > refractory:
save_state[neur, i] = 1
last_spike[neur] = i
else:
save_state[:, i] = prebotc.voltages(y, num_vertices)
i += 1
if not quiet:
if ( i % np.floor(Nstep/bar_updates) ) == 0:
bar.update(j)
j += 1
if not save_spikes:
save_state = save_state[:, 0:(i-1)]
else:
save_state.resize( (num_vertices, i-1) )
if not quiet:
bar.finish()
elapsed = time.time() - t
print("Done!\nElapsed: %1.2fs" % elapsed)
# Time saving
t = time.time()
print("Saving output....")
if save_full:
save_str = 'full'
elif save_spikes:
save_str = 'spikes'
else:
save_str = 'V'
mdict={ 'Y': save_state,
'dt': dt,
't0': t0,
'tf': tf,
'paramFn': os.path.abspath(param_fn),
'graphFn': os.path.abspath(graph_fn),
'absErr': abs_error,
'relErr': rel_error,
'saveStr': save_str,
'finalState': y,
'icStr': ic_str
}
# save output in .mat files with variable names given below
#Checks to see if it is being run for testing so, if so it will not save to avoid loading the file again
if should_save:
savemat(outFn, mdict,oned_as = 'column', do_compression=True)
else:
return mdict
if not quiet:
elapsed = time.time() - t
print("Done!\nSave time: %1.2fs" % elapsed)
def main(argv=None):
if argv is None:
argv = sys.argv
(t0, tf, dt, paramFn, graphFn, outFn, abs_error, rel_error, save_full,
save_spikes, quiet, spike_thresh, refractory) = parse_args(argv)
# compute the number of steps required
Nstep = np.ceil(tf/dt)
if not quiet:
print "Loading parameters, graph, and setting up IC's"
my_params = prebotc.params(paramFn)
num_vertices, num_edges, graph_params = prebotc.graph(graphFn)
y, N = prebotc.ics(num_vertices, num_edges)
#y, N = prebotc.ics(num_vertices, num_edges, random=False)
# rhs of ODE with parameters evaluated
# f is the rhs with parameters evaluated
f = lambda t, y: prebotc.rhs(t, y,
graph_params,
my_params)
# data structure to output, timeseries or sparse raster
if save_full:
# all state variables
save_state = np.zeros( (N, Nstep+1) )
elif save_spikes:
# just spike times, as sparse matrix
# not boolean because type conversion bug in matlab output
save_state = scipy.sparse.dok_matrix( (num_vertices, Nstep+1) )
last_spike = np.ones( num_vertices ) * (-np.inf)
else:
# timeseries of spikes
save_state = np.zeros( (num_vertices, Nstep+1) )
r = scipy.integrate.ode(f)
r.set_initial_value(y, t0)
# # other integration methods
# # method 1: BDF
# r.set_integrator(
# 'vode',
# method='bdf',
# with_jacobian = False,
# order=3,
# rtol= rel_error,
# atol= abs_error
# )
# # method 2: Dormand-Price
# r.set_integrator(
# 'dopri5',
# rtol = rel_error,
# atol = abs_error
# )
# # method 3: VODE
# r.set_integrator('vode',
# rtol = rel_error,
# atol = abs_error)
if not quiet:
print "Running integration loop...."
t = time.time()
bar_updates = 100
widgets=[progressbar.Bar('=', '[', ']'), ' ', progressbar.Percentage()]
bar = progressbar.ProgressBar(maxval=bar_updates, widgets=widgets)
bar.start()
j = 0
i = 0
while r.successful() and r.t < tf:
r.integrate(r.t + dt)
y = r.y.copy()
if save_full:
save_state[:, i] = y
elif save_spikes:
spikers = prebotc.spiking(y, num_vertices, spike_thresh)
for neur in spikers:
# only count if the new trigger occurs after reasonable delay
if dt*( float(i) - last_spike[neur] ) > refractory:
save_state[neur, i] = 1
last_spike[neur] = i
else:
save_state[:, i] = prebotc.voltages(y, num_vertices)
i += 1
if not quiet:
if ( i % np.floor(Nstep/bar_updates) ) == 0:
bar.update(j)
j += 1
if not save_spikes:
save_state = save_state[:, 0:(i-1)]
else:
save_state.resize( (num_vertices, i-1) )
if not quiet:
bar.finish()
elapsed = time.time() - t
print "Done!\nElapsed: %1.2fs" % elapsed
# Time saving
t = time.time()
print "Saving output...."
if save_full:
save_str = 'full'
elif save_spikes:
save_str = 'spikes'
else:
save_str = 'V'
# save output
scipy.io.savemat(outFn,
mdict={'Y': save_state,
'dt': dt,
't0': t0,
'tf': tf,
'paramFn': os.path.abspath(paramFn),
'graphFn': os.path.abspath(graphFn),
'absErr': abs_error,
'relErr': rel_error,
'saveStr': save_str,
'finalState': y
},
oned_as = 'column')
if not quiet:
elapsed = time.time() - t
print "Done!\nSave time: %1.2fs" % elapsed