def test_seed(self):
np.random.seed(123412845)
a1 = np.random.randint(1000000)
res1 = generate(128, 3, 100, abort_on_restart=True, seed=1423452)
a2 = np.random.randint(1000000)
res2 = generate(128, 3, 100, abort_on_restart=True, seed=1423452)
a3 = np.random.randint(1000000)
res3 = generate(128, 3, 100, abort_on_restart=True)
a4 = np.random.randint(1000000)
self.assertTrue(np.all(res1 == res2))
self.assertFalse(np.all(res1 == res3))
self.assertFalse(a1 == a2 or a2 == a3 or a3 == a4)
def __init__(self,
data_params,
params,
delay,
optimise_params,
simulator,
terminating_neurons=1,
flag=False):
'''
Calculation of the BiNAM and set up of the simulator
:param data_params: BiNAM data parameters
:param params: neuron params
:param delay: delay in the synapses
:param optimise_params: needs wCH_min, wCH_max, wCTInh = -0.001
:param simulator: nest, nmmmc1, ...
:param terminating_neurons: Number of terminating neurons
:param flag: change between SCM (False) and a single neuron in the
CS(igma) population
'''
# Generate BiNAM
self.mat_in = data.generate(data_params["n_bits_in"],
data_params["n_ones_in"],
data_params["n_samples"])
self.mat_out = data.generate(data_params["n_bits_out"],
data_params["n_ones_out"],
data_params["n_samples"])
print "Data generated!"
# set up simulator
self.scm = pyscm.SpikeCounterModel(self.mat_in, self.mat_out)
self.sim = pynl.PyNNLessIsolated(simulator)
# Initial values
self.weights = {
"wCH": 0.00,
"wCA": 0.00,
"wCSigma": -0.00,
"wCTExt": 0, # 0.02,
"wCTInh": optimise_params["wCTInh"],
"wAbort": 0 # -1.0
}
self.wCH_min, self.wCH_max = optimise_params["wCH_min"], \
optimise_params[
"wCH_max"]
self.params = params
self.data_params = data_params
self.n_ones_out = data_params["n_ones_out"]
self.delay = delay
self.terminating_neurons = terminating_neurons
self.flag = flag
self.simulator = simulator
print "Optimisation set up done!"
def test_generate_all(self):
res = generate(6, 3, 20)
res = res[np.lexsort(res.T)]
expected = np.array([
[1, 1, 1, 0, 0, 0],
[1, 1, 0, 1, 0, 0],
[1, 0, 1, 1, 0, 0],
[0, 1, 1, 1, 0, 0],
[1, 1, 0, 0, 1, 0],
[1, 0, 1, 0, 1, 0],
[0, 1, 1, 0, 1, 0],
[1, 0, 0, 1, 1, 0],
[0, 1, 0, 1, 1, 0],
[0, 0, 1, 1, 1, 0],
[1, 1, 0, 0, 0, 1],
[1, 0, 1, 0, 0, 1],
[0, 1, 1, 0, 0, 1],
[1, 0, 0, 1, 0, 1],
[0, 1, 0, 1, 0, 1],
[0, 0, 1, 1, 0, 1],
[1, 0, 0, 0, 1, 1],
[0, 1, 0, 0, 1, 1],
[0, 0, 1, 0, 1, 1],
[0, 0, 0, 1, 1, 1]], dtype=np.uint8)
numpy.testing.assert_equal(res, expected)
def test_generate_all_multiple_abort_on_restart(self):
res = generate(6, 3, 40, abort_on_restart=True)
res = res[np.lexsort(res.T)]
expected = np.array([
[1, 1, 1, 0, 0, 0],
[1, 1, 0, 1, 0, 0],
[1, 0, 1, 1, 0, 0],
[0, 1, 1, 1, 0, 0],
[1, 1, 0, 0, 1, 0],
[1, 0, 1, 0, 1, 0],
[0, 1, 1, 0, 1, 0],
[1, 0, 0, 1, 1, 0],
[0, 1, 0, 1, 1, 0],
[0, 0, 1, 1, 1, 0],
[1, 1, 0, 0, 0, 1],
[1, 0, 1, 0, 0, 1],
[0, 1, 1, 0, 0, 1],
[1, 0, 0, 1, 0, 1],
[0, 1, 0, 1, 0, 1],
[0, 0, 1, 1, 0, 1],
[1, 0, 0, 0, 1, 1],
[0, 1, 0, 0, 1, 1],
[0, 0, 1, 0, 1, 1],
[0, 0, 0, 1, 1, 1]], dtype=np.uint8)
numpy.testing.assert_equal(res, expected)
def __init__(self, data_params, params, delay, optimise_params, simulator,
terminating_neurons=1, flag=False):
'''
Calculation of the BiNAM and set up of the simulator
:param data_params: BiNAM data parameters
:param params: neuron params
:param delay: delay in the synapses
:param optimise_params: needs wCH_min, wCH_max, wCTInh = -0.001
:param simulator: nest, nmmmc1, ...
:param terminating_neurons: Number of terminating neurons
:param flag: change between SCM (False) and a single neuron in the
CS(igma) population
'''
# Generate BiNAM
self.mat_in = data.generate(data_params["n_bits_in"],
data_params["n_ones_in"],
data_params["n_samples"])
self.mat_out = data.generate(data_params["n_bits_out"],
data_params["n_ones_out"],
data_params["n_samples"])
print "Data generated!"
# set up simulator
self.scm = pyscm.SpikeCounterModel(self.mat_in, self.mat_out)
self.sim = pynl.PyNNLessIsolated(simulator)
# Initial values
self.weights = {
"wCH": 0.00,
"wCA": 0.00,
"wCSigma": -0.00,
"wCTExt": 0, # 0.02,
"wCTInh": optimise_params["wCTInh"],
"wAbort": 0 # -1.0
}
self.wCH_min, self.wCH_max = optimise_params["wCH_min"], \
optimise_params[
"wCH_max"]
self.params = params
self.data_params = data_params
self.n_ones_out = data_params["n_ones_out"]
self.delay = delay
self.terminating_neurons = terminating_neurons
self.flag = flag
self.simulator = simulator
print "Optimisation set up done!"
def test_generate_distribution(self):
res = generate(6, 3, 20)
for i in xrange(2, 22, 2):
s = np.sum(res[0:i], 0)
min_val = np.min(s)
max_val = np.max(s)
self.assertEqual(min_val, i // 2)
self.assertEqual(max_val, i // 2)
# Read in neuron data
with open("data/neuron_data.json", 'r') as outfile:
dict = json.load(outfile)
data_params = dict["data_params"]
params = dict["neuron_params"]
delay = dict["delay"]
terminating_neurons = dict["terminating_neurons"]
flag = dict["Simple_Network"]
# Read in neuron weights, some still have to be set manually
with open("data/optimised_weights.json", 'r') as outfile:
weights = json.load(outfile)
# Generate BiNAM
mat_in = data.generate(data_params["n_bits_in"], data_params["n_ones_in"],
data_params["n_samples"])
mat_out = data.generate(data_params["n_bits_out"], data_params["n_ones_out"],
data_params["n_samples"])
print "Data generated!"
# set up simulator
scm = pyscm.SpikeCounterModel(mat_in, mat_out)
sim = pynl.PyNNLess(sys.argv[1])
net, input_indices, _, input_times = scm.build(params=params, weights=weights,
delay=delay,
terminating_neurons=terminating_neurons,
flag=flag)
print "Preparations done"
# Simulation
# Read in neuron data
with open("data/neuron_data.json", 'r') as outfile:
dict = json.load(outfile)
data_params = dict["data_params"]
params = dict["neuron_params"]
delay = dict["delay"]
terminating_neurons = dict["terminating_neurons"]
flag = dict["Simple_Network"]
# Read in neuron weights, some still have to be set manually
with open("data/optimised_weights.json", 'r') as outfile:
weights = json.load(outfile)
# Generate BiNAM
mat_in = data.generate(data_params["n_bits_in"], data_params["n_ones_in"],
data_params["n_samples"])
mat_out = data.generate(data_params["n_bits_out"], data_params["n_ones_out"],
data_params["n_samples"])
print "Data generated!"
# set up simulator
scm = pyscm.SpikeCounterModel(mat_in, mat_out)
sim = pynl.PyNNLessIsolated(sys.argv[1])
[I, I_norm, fp, fn, I_start, I_norm_start, fp_start, fn_start] = np.zeros(
(8, n))
values = np.linspace(0.1, 30, n)
# Workaround for negative weights
if (stri == "wCSigma"):
values = np.linspace(-1, 0.001, n)