def create_images(files,repeat,frames = float('inf')):
'''Cycles through image stimuli continously. If the experiment is paused
or the maximum amount of frames has been presented (end of exposure period)
then it finishes.
'''
def render(stim):
if stim.done:
return 0
else:
item = stim.items[stim.i%len(stim.items)]
item['content'].draw()
return item['value']
def move(stim,loop=repeat,max_frames = frames):
if loop and not stim.paused:
if stim.i+1 >= max_frames and (stim.i+1)%len(stim.items) == 0:
stim.done = True
stim.i += 0 if stim.done else 1
elif not stim.paused:
if stim.i+1 == len(stim.items):
stim.done = True
stim.i += 0 if stim.done else 1
# Create items
items = [{ 'content': visual.ImageStim(screen,image = file),
'value': 100 if file == files[0] else 0
} for file in files]
return stimulus(items,render,move,do_nothing)
def create_instructions(instructions):
'''Creates and cycles through the instructions and images in the
experiment.yml file.
'''
def render(stim):
if stim.done:
return 0
else:
item = stim.items[stim.i]
for i in item['content']:
i.draw()
return item['value']
def next(stim):
if stim.i+1 == len(stim.items):
stim.done = True
stim.i += 0 if stim.done else 1
# Create items
items = list()
for instruct in instructions:
content = list()
if instruct['text'] and instruct['img']: # second instruction screen
content.append(visual.TextStim(screen, text=instruct['text'], alignHoriz='center', alignVert='center', pos = (0.0, -0.65), color = (1.0,1.0,1.0), height = .08, wrapWidth = 0.97))
content.append(visual.ImageStim(screen,image = instruct['img'], pos = (0.0, 0.1)))
else: # first instruction screen
content.append(visual.TextStim(screen, text=instruct['text'], alignHoriz='center', alignVert='center', pos = (0.0, 0.0), color = (1.0,1.0,1.0), height = .08, wrapWidth = 0.97))
items.append({ 'content': content, 'value': 0})
return stimulus(items,render,do_nothing,next)
def create_sounds(audio_dict,order,freq):
'''Creates audio stimuli stream based on the order text file. Ensures that
each syllable plays for 300ms. If the experiment is paused or the image
cycle has been completed, it finishes.
'''
def render(stim):
item = stim.items[stim.i]
if not stim.paused and item is not None:
item['content'].play()
return item['value']
else:
return 0
def move(stim):
if not stim.paused:
if stim.i+1 == len(stim.items):
stim.done = True
else:
stim.i += 1
# Create items
items = []
for k in order:
items += [{
'content': Sound(audio_dict['files'][int(k)],secs=0.3),
'value': int(k)
}] + [None for i in range(freq-1)]
return stimulus(items,render,move,do_nothing)
def run(ax, lesion, name, title):
S = np.zeros((len(targets),4))
model = Model()
model.make_lesion(lesion)
if os.path.exists('lesion-%s.npy' % name):
S = np.load('lesion-%s.npy' % name)
else:
for i,target in enumerate(targets):
rho,theta = target
x_,y_ = polar_to_logpolar(rho/90.0, np.pi*theta/180.0)
x_,y_ = 2*x_, 2*y_-1
model.reset()
model.R = stimulus((rho, theta))
model.run(duration=5*second, dt=5*millisecond, epsilon=0.001)
x,y = decode(model.SC_V)
S[i] = x_, y_, x, y
print 'Target at (%d,%d): %f' % (rho,theta, np.sqrt((x-x_)*(x-x_) + (y-y_)*(y-y_)))
np.save('lesion-%s.npy' % name, S)
logpolar_frame(ax)
for i in range(len(S)):
ax.plot([S[i,0],S[i,2]],[S[i,1],S[i,3]], color='k', lw=.5)
ax.scatter(S[:,0], S[:,1], s=35, color='k', marker='o', lw=.5)
ax.scatter(S[:,2], S[:,3], s=35, color='k', marker='o', facecolors='w', lw=.5)
ax.text(0,-1, title, ha="left", va="top", fontsize=14)
if lesion is not None:
position,size = lesion
rho,theta = position
rho,theta = rho/90.0, np.pi*theta/180.0
x,y = polar_to_logpolar(rho,theta)
radius = size/90.0
ax.add_patch(plt.Circle((2*x,2*y-1), radius=radius, ec='k', fc='w', zorder=+10, alpha=.5))
ax.add_patch(plt.Circle((2*x,2*y-1), radius=radius, ec='k', lw=.5, fc='none', zorder=+15))
def run(ax, lesion, name, title):
S = np.zeros((len(targets),4))
model = Model()
model.make_lesion(lesion)
if os.path.exists('lesion-%s.npy' % name):
S = np.load('lesion-%s.npy' % name)
else:
for i,target in enumerate(targets):
rho,theta = target
x_,y_ = polar_to_logpolar(rho/90.0, np.pi*theta/180.0)
x_,y_ = 2*x_, 2*y_-1
model.reset()
model.R = stimulus((rho, theta))
model.run(duration=5*second, dt=5*millisecond, epsilon=0.001)
x,y = decode(model.SC_V)
S[i] = x_, y_, x, y
print 'Target at (%d,%d): %f' % (rho,theta, np.sqrt((x-x_)*(x-x_) + (y-y_)*(y-y_)))
np.save('lesion-%s.npy' % name, S)
logpolar_frame(ax)
for i in range(len(S)):
ax.plot([S[i,0],S[i,2]],[S[i,1],S[i,3]], color='k', lw=.5)
ax.scatter(S[:,0], S[:,1], s=35, color='k', marker='o', lw=.5)
ax.scatter(S[:,2], S[:,3], s=35, color='k', marker='o', facecolors='w', lw=.5)
ax.text(0,-1, title, ha="left", va="top", fontsize=14)
if lesion is not None:
position,size = lesion
rho,theta = position
rho,theta = rho/90.0, np.pi*theta/180.0
x,y = polar_to_logpolar(rho,theta)
radius = size/90.0
ax.add_patch(plt.Circle((2*x,2*y-1), radius=radius, ec='k', fc='w', zorder=+10, alpha=.5))
ax.add_patch(plt.Circle((2*x,2*y-1), radius=radius, ec='k', lw=.5, fc='none', zorder=+15))
def setStim(self, amp, delta, T, times):
self.amp = amp #amplitude
self.delta = delta # period bt stim
self.T = T # period stim
self.times = times
self.mystim = stimulus(self.patterns, self.lagStim, self.delta, self.T,
self.times)
self.mystim.inten = self.amp
def __init__(self): #Global parameters needed
# single neuron
self.n = 10 #n pop
self.tau = 10. #timescale of populations
self.nu = 1. # slope PWL TF
self.theta = 0. # lower threshold
self.uc = 1. #upper threshold
# Learning Rule
self.wmax = 1.8 # maximum learned
self.thres = 0.6 # threshold
self.delay = 15.3 # delay
self.w0 = 0.01 # initial condition dynamics
bf = 10. # slope f and g
xf = 0.7 # threshold f and g
self.a_post = bf
self.b_post = xf
self.a_pre = bf
self.b_pre = xf
self.tau_learning = 400. # time scale LR
# stimulus
self.amp = 5.5 #amplitude
self.patterns = np.identity(self.n) # stimulation matrix
self.patterns = [self.patterns[:, i]
for i in range(self.n)] # puting in a list
self.dt = 0.5 # dt dinamics
self.npts = int(np.floor(self.delay / self.dt) + 1) # points delay
self.lagStim = 100. # lag before stimulation
self.delta = 10. # period bt stim
self.T = 10. # period stim
self.times = 150
self.mystim = stimulus(self.patterns, self.lagStim, self.delta, self.T,
self.times)
self.mystim.inten = self.amp
#Homeostatic LR
self.tau_H = 200000. # tau
self.y0 = .12 * np.ones(self.n) # target rate
#inhibition
self.w_i = 2.
#if there is noise in the simulations
self.amp_noise = 0
a_pre = 0.5 b_pre = -3.0 tau_learning = 10.0 # comment this lines if you don t want to load or save the parameters of the simulation # name='failure_learning2.param' # save_param(name) # load_param(name) w_inh = w_i / n r1_matrix = np.ones((n, n)) patterns = np.identity(n) patterns = [patterns[:, i] for i in range(n)] mystim = stimulus(patterns, lagStim, delta, period, times) mystim.inten = amp # integrator npts = int(np.floor(delay / dt) + 1) # points delay tmax = times * (lagStim + n * (period + delta)) + 40 # initial conditions x0 = 0.01 * np.ones((npts, n)) W0 = [(wmax / n) * np.ones((n, n)) for i in range(npts)] theintegrator = myintegrator(delay, dt, n, tmax) theintegrator.fast = False rowsum, u, Wdiag, Woffdiag, connectivity, W01, t = theintegrator.DDE_Norm_additive(field, x0, W0) # new instance of the model # with no stimulation
a_pre=bf b_pre=xf tau_learning=400.#30000. a1=6. b1=-0.25 #------------------------------------------------------------------- #-----------------Stimulation of Populations------------------------ #------------------------------------------------------------------- # setting up the simulation r1_matrix=np.ones((n,n)) patterns=np.identity(n) patterns=[patterns[:,i] for i in range(n)] mystim=stimulus(patterns,lagStim,delta,period,times) mystim.inten=amp #integrator npts=int(np.floor(delay/dt)+1) # points delay tmax=times*(lagStim+n*(period+delta))+100.+mystim.delay_begin thetmax=tmax+15.5*tau_H #initial conditions a0=np.zeros((npts,n)) x0=0.1*np.ones((npts,n)) W0=[0.1*np.ones((n,n)) for i in range(npts)] H0=[np.array([0.1 for i in range(n)]) for i in range(npts)] #H0=[np.array([19.52158144,13.31267976,13.35448593,13.35612847,13.35535822,13.35451532,13.35366458,13.35281449,13.35258073,13.35252602]) for i in range(npts)] #H0=[0.5*np.ones(n) for i in range(npts)]
x = (Z*X).sum() / Z_sum
y = (Z*Y).sum() / Z_sum
return x,y
p = 100
np.random.seed(123)
if os.path.exists('data/double-target-intensity-5.npy'):
T5 = np.load('data/double-target-intensity-5.npy')
else:
rho = 5
model= Model()
T5 = np.zeros((p,2))
for i,theta in enumerate(np.linspace(10,45,p)):
model.reset()
model.R = np.maximum( stimulus((rho, -theta), size=1, intensity=1) ,
stimulus((rho, +theta), size=1, intensity=1) )
model.R += np.random.uniform(0,0.05,model.R.shape)
model.run(duration=5*second, dt=5*millisecond, epsilon=0.0)
x,y = decode(model.SC_V)
print u"Δθ = %.2f: (%f,%f)" % (2*theta, x, y)
T5[i] = x,y
np.save("data/double-target-intensity-5.npy",T5)
if os.path.exists('data/double-target-intensity-10.npy'):
T10 = np.load('data/double-target-intensity-10.npy')
else:
rho = 10
model= Model()
T10 = np.zeros((p,2))
import os
import numpy as np
from helper import *
from stimulus import *
from graphics import *
from projections import *
if __name__ == "__main__":
import matplotlib.pyplot as plt
fig = plt.figure(figsize=(16, 6), facecolor="w")
# Stimuli luminances are not additive
R = stimulus(position=(5, 0))
P = retina_projection(retina_shape, [1024, 1024])
# ---------------------------------
ax = plt.subplot(2, 1, 1)
ax.tick_params(direction="outward")
ax.spines["right"].set_color("none")
ax.spines["top"].set_color("none")
ax.spines["left"].set_color("none")
ax.spines["bottom"].set_color("none")
ax.xaxis.set_ticks_position("bottom")
ax.spines["bottom"].set_position(("data", -0.05))
X = np.linspace(0, 90, R.shape[1])
for rho in [2, 5, 10, 20, 40, 60, 80]:
R = stimulus(position=(rho, 0))
ax.set_xlim(-90,90)
ax.set_ylim(0,250)
ax.set_zlim(0,300)
ax.set_xlabel(u'Stimuli position [°]', fontsize=fontsize,labelpad=4)
ax.set_ylabel('Time [ms]', fontsize=fontsize,labelpad=5)
ax.set_zlabel('Discharge rate (spike/s)', fontsize=fontsize,labelpad=5)
ax.tick_params(axis='both', which='major', labelsize=fontsize-1)
ax.set_zticks([0,100,200,300])
plt.figure(figsize=(2.5,6.5), facecolor='w',dpi=100)
model = Model()
model.reset()
model.R = stimulus((5.0, -30), size=1, intensity=1)
ax = plt.subplot(311, projection='3d')
run(model,ax)
ax.text2D(0.0, 1.0, "a", va='top', ha='right',
transform=ax.transAxes, fontsize=10, fontweight='bold')
# Get location of -20/+20 on SC
model.reset()
model.R = stimulus((5.0, -20), size=1, intensity=1)
model.run(1.*second, 4*millisecond)
x,_ = polar_to_logpolar(5/90.,0.)
x = int(x*128)
y1 = model.SC_V[:,x].argmax()
max_trials = 1
RP = 4
ITI = 2
# Set images
GO_images = glob('images/obj1*')
NOGO_images = glob('images/obj2*')
# Initialize image matrices
GO_resp = np.ones(np.size(GO_images))
NOGO_resp = np.zeros(np.size(NOGO_images))
ALL_images = GO_images + NOGO_images
ALL_resp = np.concatenate([GO_resp, NOGO_resp])
# Load images
stim = stimulus(ALL_images, w, h)
# Datajoin stuff
#session = dict(mouse_id=1, session_tmst=timer.start_time)
#Sessions().insert1(session)
# Initialize other settings
timer.start()
stim.init_block()
trial = 0
state_time = 0
pygame.mouse.set_visible(0)
# RUN
while trial < max_trials: # Each trial is one block
T = np.zeros((len(targets), 2))
for i, target in enumerate(targets):
rho, theta = target
rho, theta = rho / 90.0, np.pi * theta / 180.0
x, y = polar_to_logpolar(rho, theta)
T[i] = 2 * x, 2 * y - 1
if not os.path.exists('accuracy.npy'):
model = Model()
D = np.zeros((len(targets), 2))
for i, target in enumerate(targets):
rho, theta = target
x_, y_ = polar_to_logpolar(rho / 90.0, np.pi * theta / 180.0)
x_, y_ = 2 * x_, 2 * y_ - 1
model.reset()
model.R = stimulus((rho, theta))
model.run(duration=5 * second, dt=5 * millisecond, epsilon=0.001)
x, y = decode(model.SC_V)
D[i] = x, y
print 'Target at (%d,%d): %f' % (rho, theta,
np.sqrt((x - x_) * (x - x_) +
(y - y_) * (y - y_)))
np.save('accuracy.npy', D)
else:
D = np.load('accuracy.npy')
fig = plt.figure(figsize=(8, 8), facecolor='w')
ax1 = plt.subplot(111, aspect=1)
logpolar_frame(ax1)
for i in range(len(D)):
plt.plot([T[i, 0], D[i, 0]], [T[i, 1], D[i, 1]], color='k')
import time
from matplotlib import pyplot
GO_images = ('image1.jpg', 'image2.jpg')
NOGO_images = ('image3.jpg', 'image4.jpg')
# Initialize image matrices
GO_resp = np.ones(np.size(GO_images))
NOGO_resp = np.zeros(np.size(NOGO_images))
ALL_images = GO_images + NOGO_images
ALL_resp = np.concatenate([GO_resp, NOGO_resp])
pygame.init()
clock = pygame.time.Clock()
FPS = 60
s = stimulus(ALL_images)
s.init_block()
stimes = np.zeros(200)
etimes = np.zeros(200)
for itrial in range(200):
stimes[itrial] = time.time()
s.init_trial()
s.show()
clock.tick(FPS)
s.unshow()
etimes[itrial] = time.time()
times = etimes - stimes
s.close()
from mpl_toolkits.axes_grid1 import ImageGrid
from mpl_toolkits.axes_grid1.inset_locator import zoomed_inset_axes
from mpl_toolkits.axes_grid1.inset_locator import mark_inset
fig = plt.figure(figsize=(10, 8), facecolor='w')
ax1, ax2 = ImageGrid(fig, 111, nrows_ncols=(1, 2), axes_pad=0.5)
polar_frame(ax1, legend=True)
zax = zoomed_inset_axes(ax1, 6, loc=1)
polar_frame(zax, zoom=True)
zax.set_xlim(0.0, 0.1)
zax.set_xticks([])
zax.set_ylim(-.05, .05)
zax.set_yticks([])
zax.set_frame_on(True)
mark_inset(ax1, zax, loc1=2, loc2=4, fc="none", ec="0.5")
# Stimuli luminances are not additive
R = np.maximum(stimulus(position=(1, 0)), stimulus(position=(5, 0)))
polar_imshow(ax1, R)
polar_imshow(zax, R)
logpolar_frame(ax2, legend=True)
P = retina_projection()
SC = R[P[..., 0], P[..., 1]]
logpolar_imshow(ax2, SC)
# plt.savefig("fig-input.pdf")
plt.show()
################################
ax1, ax2 = ImageGrid(fig, 212, nrows_ncols=(1,2), axes_pad=0.5)
polar_frame(ax1, legend=True,reduced=True)
'''
zax = zoomed_inset_axes(ax1, 6, loc=1)
polar_frame(zax, zoom=True)
zax.set_xlim(0.0, 0.15)
zax.set_xticks([])
zax.set_ylim(-.05, .05)
zax.set_yticks([])
zax.set_frame_on(True)
mark_inset(ax1, zax, loc1=2, loc2=4, fc="none", ec="0.5")
'''
# Stimuli luminances are not additive
R = np.maximum(stimulus(position=(1,0)), stimulus(position=(5,0)))
R = np.maximum(R , stimulus(position=(10,0)))
polar_imshow(ax1, R,reduced=True)
#polar_imshow(zax, R)
logpolar_frame(ax2, legend=True)
P = retina_projection()
SC = R[P[...,0], P[...,1]]
logpolar_imshow(ax2, SC)
ax1.text(1.1, 1.1, u"b",
ha="left", va="bottom", fontsize=20, fontweight='bold')
fig.subplots_adjust(left=0.05, bottom=0.05, right=0.95, top=0.95,
wspace=0.15, hspace=0.2)
plt.savefig("figures/Fig-1.pdf")
x = (Z*X).sum() / Z_sum
y = (Z*Y).sum() / Z_sum
return x,y
p = 100
np.random.seed(123)
if os.path.exists('data/double-target-intensity-5.npy'):
T5 = np.load('data/double-target-intensity-5.npy')
else:
rho = 5
model= Model()
T5 = np.zeros((p,2))
for i,theta in enumerate(np.linspace(10,45,p)):
model.reset()
model.R = np.maximum( stimulus((rho, -theta), size=1, intensity=1) ,
stimulus((rho, +theta), size=1, intensity=1) )
model.R += np.random.uniform(0,0.05,model.R.shape)
model.run(duration=5*second, dt=5*millisecond, epsilon=0.0)
x,y = decode(model.SC_V)
print u"Δθ = %.2f: (%f,%f)" % (2*theta, x, y)
T5[i] = x,y
np.save("data/double-target-intensity-5.npy",T5)
if os.path.exists('data/double-target-intensity-10.npy'):
T10 = np.load('data/double-target-intensity-10.npy')
else:
rho = 10
model= Model()
T10 = np.zeros((p,2))
# -----------------------------------------------------------------------------
import os
import numpy as np
from helper import *
from stimulus import *
from graphics import *
from projections import *
if __name__ == '__main__':
import matplotlib.pyplot as plt
fig = plt.figure(figsize=(16, 6), facecolor='w')
# Stimuli luminances are not additive
R = stimulus(position=(5, 0))
P = retina_projection(retina_shape, [1024, 1024])
# ---------------------------------
ax = plt.subplot(2, 1, 1)
ax.tick_params(direction="outward")
ax.spines['right'].set_color('none')
ax.spines['top'].set_color('none')
ax.spines['left'].set_color('none')
ax.spines['bottom'].set_color('none')
ax.xaxis.set_ticks_position('bottom')
ax.spines['bottom'].set_position(('data', -0.05))
X = np.linspace(0, 90, R.shape[1])
for rho in [2, 5, 10, 20, 40, 60, 80]:
R = stimulus(position=(rho, 0))
w0=0.1*random.randn(N,N)#ones((N,N))#random.randn(N, N) w0=list(w0.reshape(N*N)) y0=list(u0)+w0 u0=array(u0) y0=array(y0) ################################ ##Building patterns times=1 delay_inter=1000 delay_intra=40 n=2.0 v1=array([1,0,0])#*(1/sqrt(2)) v2=array([0,1,0])#*(1/sqrt(2)) v3=array([0,0,1])#*(1/sqrt(2)) patterns=[v1,-0.3*v1]#random.binomial(1,0.5,N)i mystim=stimulus(patterns,delay_inter,delay_intra,times) mystim.tstim=90 mystim.inten=1.9 mystim.delay_begin=700 dt=0.1 tmax=mystim.delay_begin+times*n*(mystim.tstim+delay_inter)-times*delay_inter+times*delay_intra+700 ###########Building network####### ubar=array([0.3,0.3,0.3])#0.1*random.randn(N)#array([2,3]) net=network(u0,struct) net.covmean(array(ubar)) net.stim(mystim.stim) net.homeostatic(0) net.alpha=20 net.v12=0
for i,target in enumerate(targets):
rho,theta = target
rho,theta = rho/90.0, np.pi*theta/180.0
x,y = polar_to_logpolar(rho,theta)
T[i] = 2*x,2*y-1
if not os.path.exists('accuracy.npy'):
model = Model()
D = np.zeros((len(targets),2))
for i,target in enumerate(targets):
rho,theta = target
x_,y_ = polar_to_logpolar(rho/90.0, np.pi*theta/180.0)
x_,y_ = 2*x_, 2*y_-1
model.reset()
model.R = stimulus((rho, theta))
model.run(duration=5*second, dt=5*millisecond, epsilon=0.001)
x,y = decode(model.SC_V)
D[i] = x,y
print 'Target at (%d,%d): %f' % (rho,theta, np.sqrt((x-x_)*(x-x_) + (y-y_)*(y-y_)))
np.save('accuracy.npy', D)
else:
D = np.load('accuracy.npy')
fig = plt.figure(figsize=(8,8), facecolor='w')
ax1 = plt.subplot(111, aspect=1)
logpolar_frame(ax1)
for i in range(len(D)):
plt.plot([T[i,0],D[i,0]],[T[i,1],D[i,1]], color='k')
ax1.scatter(T[:,0], T[:,1], s=50, color='k', marker='o')
