def prg_small():
s = snn('default')
s.nneus = [1, 2, 2]
#s.delays = [1.,2.,3.]
s.defWt = 100.
# s.makeSyns = [(0, 0, 0)] #, (1, 0, 0)] #, (0, 0, 1), (0, 1, 1), (1, 1, 0)]
# s.connections <- [(srcLayer, srcN, destLayer, destN, weight)]
# s.connections = [(0,0,1,0),(1,0,1,1)] #,(1,1,1,0),(2,0,2,1),(2,1,2,0)]
s.connections = [(0, 0, 1, 0), (0, 0, 2, 0), (1, 0, 1, 1),
(1, 1, 1, 0, -25.)]
#(1,0,1,1),(1,1,1,0),
#(2,0,2,1),(2,1,2,0)]
s.plasticity = False
n0, n1, n10, n11, n20, n21 = (0, 0), (0, 1), (1, 0), (1, 1), (2, 0), (2, 1)
s.currentAt = {n0: 30.}
s.stoptime = 10.
s.dt = 0.01
s.nTrace = [n0, n10, n11, n20] #,n20,n21] # [(layer <0..> ,indx <0..>)]
s.enable_Log = ("S", False, True, False)
s.enable_RunStats = False
s.init()
s.run()
def prg_2mesh():
# np.random.seed(10)
# initialize snn object with a name
s = snn('default')
# delays to layers
s.delays = [] #[.1,.1,.1]
# neurons in each layer
s.nneus = [100, 100]
# default weight of connections
s.defWt = 40.
# random connections: [(srcLayer, destLayer, fraction of maximum connections, weightMean, weightSD)]
s.randLayerConnections = [(0, 1, .6, 5., 0.), (1, 1, .5, 10., 2.),
(1, 0, .5, -1, 1.)
] #,(1,1,.9,10.,10.),(1,0,.8,-2,1.)]
# random current: [(mean,SD)] for each layer, starts at layer 0
s.randCurrent = [(10., 2.)] #,(30.,10.),(50.,10.),(70.,10.)]
# time at which to stop external current
#s.relaxAt = 50.
#s.showSyns = True
# enable triplet STDP
s.plasticity = False
s.plasticity_params['WFactor'] = 10.
# log (Potential, --, Spike, Weight)
s.enable_Log = (False, False, True, False)
# log potential for neurons: (layer,neuron)
# s.nTrace = [(0,0),(0,1),(1,4)]
s.stoptime, s.dt = 1000., .1
s.extraplots = 1
s.init()
s.run()
rwnd = 100
#--- plot spike rates per layer
axr = s.fig.add_subplot(s.subplotIndx + 1, sharex=s.axes[0])
spall = np.zeros((0, 2))
try:
for l in range(len(s.nneusUser)):
start = sum(s.nneusUser[:l])
spikeRates = np.asarray([
(i,
np.sum(s.spikeRaster[start:(start + s.nneus[l]),
i:(i + rwnd)]) / s.nneusUser[l])
for i in range(rwnd, s.iterations - rwnd)
])
print(spikeRates.shape)
axr.plot(spikeRates[:, 0], spikeRates[:, 1], label='Layer %d' % l)
spall = np.vstack((spall, spikeRates))
axr.legend()
axr.set_xlim(0, s.iterations)
axr.set_title('Average f.rates')
np.save('sprates', spall)
s.axes.append(axr)
except IndexError:
print("Could not plot spike rates")
def prg_Big():
np.random.seed(10)
s = snn('default')
s.nneus = [100] * 150
#s.makeSyns = [(0,0,0)]
s.randLayerConnections = [(sL, sL + 1, 100, 1., .5)
for sL in range(len(s.nneus) - 2)]
s.randCurrent = (5., 1)
s.enable_Log = (False, False, False, False)
s.stoptime, s.dt = 200., .1
s.extraplots = 1
s.init()
s.run()
def prg_small_delays():
s = snn('default')
s.nneus = [3, 1, 1]
s.delays = [1., 1.5, 3.] #, [1.,2.] #[1.,2.] #,2.,3.]
s.defWt = 40.
s.makeSyns = [(0, 0, 0), (0, 1, 0)]
s.plasticity = False
n0, n1 = (0, 0), (1, 0)
#s.spikeTm[1.] = [n0]
s.currentAt = {n0: 20.}
s.stoptime = 20.
s.dt = 0.01
s.nTrace = [n0, n1, (2, 0)] # [(layer <0..> ,indx <0..>)]
s.enable_Log = (True, False, True, False)
s.init()
s.run()
def prg_plasticTest():
s = snn('default')
s.nneus = [2, 1]
s.makeSyns = [(0, 0, 0)]
s.connections = [(0, 0, 0, 1, 10.)]
s.plasticity = True
s.plasticity_params['WFactor'] = 40 # 200000. #200000. #40
s.norm_weights = False
n1, n2, n11 = (0, 0), (0, 1), (1, 0)
# test spike before train:
s.spikeTm[4.0] = [n1]
s.resetAt.append(8.0)
# train:
for st in np.linspace(10., 50., 5):
s.spikeTm[st] = [n1]
s.spikeTm[st + 1.] = [n11]
s.spikeTm[st + 1.2] = [n2]
s.resetAt.append(st + 2.)
# test:
s.spikeTm[60.] = [n1]
print(s.spikeTm)
s.stoptime = 80.0
s.nTrace = [n1, n2, n11] # [(layer <0..> ,indx <0..>)]
s.enable_Log = (True, False, True, True)
s.init()
s.run()
if np.any(s.spikeRaster):
n, itr = np.where(s.spikeRaster)
# for nn,ii in np.where(s.spikeRaster.T[
print("Last spikes on N%d at %f time." % (n[-1], itr[-1] * s.dt))
np.save(
'pltimes',
sorted(zip(n.tolist(), (itr * s.dt).tolist()),
key=lambda x: x[1])) or print("saved to pltimes.")
else:
print("Nothing in raster")
plt.show()
def prg_large():
s = snn('default')
s.nneus = [200, 200, 10]
s.randSyns = 10 # no. of outward connections
s.randCurrent = (100, 20)
# [(layer <0..> ,indx <0..>)]
s.nTrace = zip([1] * 10, [1])
s.plasticity = False
s.enable_Log = (False, False, True, False)
s.stoptime = 20.0
#s.K_DEFN['DISPSPK'] = str(0)
s.norm_weights = True
s.defWt = 10.
s.dt = .01
s.init()
with open('last_syns', 'w') as f:
pickle.dump(s.syns, f)
s.run()
from __future__ import print_function
from cl_osnn import snn
import pickle
import numpy as np
import matplotlib.pyplot as plt
import os
os.environ['PYOPENCL_COMPILER_OUTPUT'] = '0'
os.environ['PYOPENCL_CTX'] = '0'
#np.random.seed(1)
s = snn('default')
s.nneus = [1024, 200] #, 200 ]#, 128]# , 100] #200
#s.delays = [0.,5.,7.]
#s.makeSynsWeight = [(0,100,1,100.)]
# s.randLayerConnections = [(0,0,1.,5.,1.)]#, (1,1,1.,1.,1.)]
# s.randLayerConnections = [(1,1,1.,1.,1.)]
# s.randLayerConnections = [(0,1,.7,5.,1.), (1,1,.7,5.,1.)]
s.randLayerConnections = [
(0, 1, .7, 1., .1), (1, 0, .9, -3., 1.)
] #, (1,0,.7,-2.,.3), (1,2,.1,.1,1.)] #,(1,1,.7,5.,1.)]
s.randCurrent = [(30, 10)]
#s.currentAt[(0,100)] = 100.
#s.relaxAt = 50
s.make_logs['spike'] = True
s.recordSpikes = True
s.nTrace = [(0, 100), (1, 1), (1, 0)]
s.stoptime, s.dt = 100., .01