def get_md5s(PS_prefix):
'''return list of md5 hash keys from parameterspace prefix'''
md5s = []
PS = ParameterSpace(os.path.join('parameters', PS_prefix + '.pspace'))
for i, PSet in enumerate(PS.iter_inner()):
id = get_md5(PSet)
md5s += [id]
return md5s
def make_param_dict_list():
"""
create a list of parameter dictionaries for the model network.
"""
# there is certainly a way to do this with NeuroTools.
import numpy
rates = numpy.linspace(start = 10., stop = 100., num = 5)
weights = numpy.linspace(start = 0.1, stop = 1.0, num = 5)
from NeuroTools.parameters import ParameterSet, ParameterSpace, ParameterRange
params = ParameterSpace(ParameterSet({'rate':ParameterRange(rates),
'weight': ParameterRange(weights)}))
dictlist = [p.as_dict() for p in params.iter_inner()]
return dictlist
def make_param_dict_list(N):
"""
create a list of parameter dictionaries for the model network.
"""
N_snr, N_seeds = 5, 10
from NeuroTools.parameters import ParameterSpace, ParameterRange
import numpy
params = ParameterSpace({
'N' : N,
'snr' : ParameterRange(list(numpy.linspace(0.1,2.0,N_snr))),
'kernelseed' : ParameterRange(list([12345+ k for k in range(N_seeds)]))})
dictlist = [p.as_dict() for p in params.iter_inner()]
return dictlist
def make_param_dict_list():
"""
create a list of parameter dictionaries for the model network.
"""
# there is certainly a way to do this with NeuroTools.
import numpy
rates = numpy.linspace(start=10., stop=100., num=5)
weights = numpy.linspace(start=0.1, stop=1.0, num=5)
from NeuroTools.parameters import ParameterSet, ParameterSpace, ParameterRange
params = ParameterSpace(
ParameterSet({
'rate': ParameterRange(rates),
'weight': ParameterRange(weights)
}))
dictlist = [p.as_dict() for p in params.iter_inner()]
return dictlist
def make_param_dict_list(N):
"""
create a list of parameter dictionaries for the model network.
"""
N_snr, N_seeds = 5, 10
from NeuroTools.parameters import ParameterSpace, ParameterRange
import numpy
params = ParameterSpace({
'N':
N,
'snr':
ParameterRange(list(numpy.linspace(0.1, 2.0, N_snr))),
'kernelseed':
ParameterRange(list([12345 + k for k in range(N_seeds)]))
})
dictlist = [p.as_dict() for p in params.iter_inner()]
return dictlist
retina.params['amplitude'] = numpy.ones(retina.params['amplitude'].shape)
# calculates the dimension of the parameter space
results_dim, results_label = p.parameter_space_dimension_labels()
# creates results array with size of parameter space dimension
data = retina.run(retina.params,verbose=False)
lower_edges = data['out_ON_DATA'].time_axis(t_smooth)
N_smooth = len(lower_edges)
temporal_ON, temporal_OFF = [],[]
import progressbar # see http://projects.scipy.org/pipermail/scipy-dev/2008-January/008200.html
pbar=progressbar.ProgressBar(widgets=[name, " ", progressbar.Percentage(), ' ',
progressbar.Bar(), ' ', progressbar.ETA()], maxval=N_exp)
for i_exp,experiment in enumerate(p.iter_inner()):
params = retina.params
params.update(experiment) # updates what changed in the dictionary
# simulate the experiment and get its data
data = retina.run(params,verbose=False)
# calculating the index in the parameter space
index = p.parameter_space_index(experiment)
# put the data at the right position in the results array
temporal_ON.append(sum(data['out_ON_DATA'].firing_rate(t_smooth))/N)#
temporal_OFF.append(sum(data['out_OFF_DATA'].firing_rate(t_smooth))/N)#
pbar.update(i_exp)
results['lower_edges'] = lower_edges
results['temporal_ON'] = temporal_ON
results['temporal_OFF'] = temporal_OFF
parser.error("incorrect number of arguments")
test_script, url = args[:2]
script_args = args[2:]
trials = options.trials
if hasattr(options, "host_list"):
host_list = options.host_list
else:
host_list = [socket.gethostname()] # by default, run just on the current host
# iterate over the parameter space, creating a job each time
parameter_space = ParameterSpace(url)
tempfiles = []
job_manager = JobManager(host_list, delay=0, quiet=False)
for sub_parameter_space in parameter_space.iter_inner(copy=True):
for parameter_set in sub_parameter_space.realize_dists(n=trials, copy=True):
##print parameter_set.pretty()
fd, tmp_url = tempfile.mkstemp(dir=os.getcwd())
os.close(fd)
tempfiles.append(tmp_url)
parameter_set.save(tmp_url)
job_manager.run(test_script, parameter_set._url, *script_args)
# wait until all jobs have finished
job_manager.wait()
# retrieve results stored by the jobs. We use the NeuroTools datastore, so
# the test module itself is the storage key.
test_module = __import__(test_script.replace(".py",""))
ds = datastore.ShelveDataStore(root_dir=parameter_space.results_dir,
cc = numpy.correlate(st1.time_histogram(time_bin=1.0),
st2.time_histogram(time_bin=1.),
mode='same')
corrcoef = numpy.corrcoef(st1.time_histogram(time_bin=1.0),
st2.time_histogram(time_bin=1.))
time_axis_cc = numpy.linspace(-cc.shape[0] / 2., cc.shape[0] / 2.,
cc.shape[0])
return cc, time_axis_cc, corrcoef[0][1]
# creating a results array, with the dimensions of the ParameterSpace
corrcoef_results = numpy.empty(dims)
# scanning the ParameterSpace
for experiment in p.iter_inner():
# calculation of the index in the space
index = p.parameter_space_index(experiment)
# perfomring the experiment
cc, time_axis_cc, corrcoef = calc_cc(experiment)
corrcoef_results[index] = corrcoef
# plotting the cc's
subplot_index = (dims[1] * index[0]) + index[1]
pylab.subplot(dims[0], dims[1], subplot_index + 1)
pylab.plot(time_axis_cc, cc)
pylab.title(make_name(experiment, p.range_keys()))
pylab.xlim(-30, 30.)
pylab.ylim(0, 10.)
# plot the results
pylab.matshow(corrcoef_results)
test_script, url = args[:2]
script_args = args[2:]
trials = options.trials
if hasattr(options, "host_list"):
host_list = options.host_list
else:
host_list = [socket.gethostname()
] # by default, run just on the current host
# iterate over the parameter space, creating a job each time
parameter_space = ParameterSpace(url)
tempfiles = []
job_manager = JobManager(host_list, delay=0, quiet=False)
for sub_parameter_space in parameter_space.iter_inner(copy=True):
for parameter_set in sub_parameter_space.realize_dists(n=trials,
copy=True):
##print parameter_set.pretty()
fd, tmp_url = tempfile.mkstemp(dir=os.getcwd())
os.close(fd)
tempfiles.append(tmp_url)
parameter_set.save(tmp_url)
job_manager.run(test_script, parameter_set._url, *script_args)
# wait until all jobs have finished
job_manager.wait()
# retrieve results stored by the jobs. We use the NeuroTools datastore, so
# the test module itself is the storage key.
test_module = __import__(test_script.replace(".py", ""))
st1.merge(st3.jitter(p.jitter))
st2.merge(st3.jitter(p.jitter))
cc = numpy.correlate(st1.time_histogram(time_bin = 1.0),
st2.time_histogram(time_bin = 1.),mode = 'same')
corrcoef = numpy.corrcoef(st1.time_histogram(time_bin = 1.0),
st2.time_histogram(time_bin = 1.))
time_axis_cc = numpy.linspace(-cc.shape[0]/2.,cc.shape[0]/2.,cc.shape[0])
return cc, time_axis_cc, corrcoef[0][1]
# creating a results array, with the dimensions of the ParameterSpace
corrcoef_results = numpy.empty(dims)
# scanning the ParameterSpace
for experiment in p.iter_inner():
# calculation of the index in the space
index = p.parameter_space_index(experiment)
# perfomring the experiment
cc,time_axis_cc, corrcoef = calc_cc(experiment)
corrcoef_results[index] = corrcoef
# plotting the cc's
subplot_index = (dims[1]*index[0])+index[1]
pylab.subplot(dims[0],dims[1],subplot_index+1)
pylab.plot(time_axis_cc,cc)
pylab.title(make_name(experiment,p.range_keys()))
pylab.xlim(-30,30.)
pylab.ylim(0,10.)
# plot the results
# creates results array with size of parameter space dimension
data = retina.run(retina.params, verbose=False)
lower_edges = data['out_ON_DATA'].time_axis(t_smooth)
N_smooth = len(lower_edges)
temporal_ON, temporal_OFF = [], []
import progressbar # see http://projects.scipy.org/pipermail/scipy-dev/2008-January/008200.html
pbar = progressbar.ProgressBar(widgets=[
name, " ",
progressbar.Percentage(), ' ',
progressbar.Bar(), ' ',
progressbar.ETA()
],
maxval=N_exp)
for i_exp, experiment in enumerate(p.iter_inner()):
params = retina.params
params.update(experiment) # updates what changed in the dictionary
# simulate the experiment and get its data
data = retina.run(params, verbose=False)
# calculating the index in the parameter space
index = p.parameter_space_index(experiment)
# put the data at the right position in the results array
temporal_ON.append(sum(data['out_ON_DATA'].firing_rate(t_smooth)) /
N) #
temporal_OFF.append(
sum(data['out_OFF_DATA'].firing_rate(t_smooth)) / N) #
pbar.update(i_exp)
results['lower_edges'] = lower_edges
results['temporal_ON'] = temporal_ON
