resolution=0.1,
stim_duration=1,
ITImodel='exponential',
ITImin=1,
ITImean=2,
ITImax=5,
confoundorder=3, # this cannot be 0
hardprob=True,
)
# optimize the design for detection efficiency only using GA
POP_GA = optimisation(experiment=EXP,
weights=[0, 0.5, 0.5, 0],
preruncycles=2,
cycles=cycles,
seed=1,
outdes=5,
I=10,
folder='/tmp/',
optimisation='GA',
R=[0.5, 0.5, 0.0])
POP_GA.optimise()
# print the best model score
print("Score: %s " % POP_GA.optima[::-1][0])
print("N trials: %d " % len(POP_GA.bestdesign.onsets))
# Let's look at the resulting experimental designs.
# this plots the columns of the X matrix convolved with the HRF
plt.figure(figsize=(10, 7))
order = neurodesign.generate.order(nstim=4,
ntrials=100,
probabilities=[0.25, 0.25, 0.25, 0.25],
ordertype='random',
seed=1234)
print(order[:10])
Counter(order)
iti, lam = neurodesign.generate.iti(ntrials=40,
model='exponential',
min=2,
mean=3,
max=8,
resolution=0.1,
seed=2134)
print(iti[:10])
print("mean ITI: %s \n\
min ITI: %s \n\
max ITI: %s" %
(round(sum(iti) / len(iti), 2), round(min(iti), 2), round(max(iti), 2)))
POP = neurodesign.optimisation(experiment=EXP,
weights=[0, 0.5, 0.25, 0.25],
preruncycles=10,
cycles=100,
folder="./",
seed=100)
POP.optimise()
stim_duration=1,
t_pre = 0,
t_post = 2,
restnum=0,
restdur=0,
ITImodel = "exponential",
ITImin = 1,
ITImean = 2,
ITImax=4
)
POP = optimisation(
experiment=EXP,
weights=[0,0.5,0.25,0.25],
preruncycles = 10,
cycles = 10,
seed=1,
outdes=5,
folder='/Users/Joke/'
)
#########################
# run natural selection #
#########################
POP.optimise()
POP.download()
POP.evaluate()
################
# step by step #
P=[0.25, 0.25, 0.25],
C=[[1, 0, 0], [0, 1, 0], [0, 0, 1], [1, 0, -1]],
n_stimuli=3,
rho=0.3,
resolution=0.1,
stim_duration=1,
ITImodel="exponential",
ITImin=0.3,
ITImean=1,
ITImax=4)
# In[3]:
POP_Max = optimisation(experiment=EXP,
weights=[0, 0.5, 0.25, 0.25],
preruncycles=cycles,
cycles=2,
optimisation='GA')
POP_Max.optimise()
# In[4]:
EXP.FeMax = POP_Max.exp.FeMax
EXP.FdMax = POP_Max.exp.FdMax
# Below we define two populations of designs. We will optimise one using the genetic algorithm, and the other using randomly drawn designs.
#
# We optimise for statistical power (weights = [0,1,0,0]). We run 100 cycles.
# In[5]:
rho=0.3,
resolution=0.1,
stim_duration=1,
t_pre=0,
t_post=2,
restnum=0,
restdur=0,
ITImodel="exponential",
ITImin=1,
ITImean=2,
ITImax=4)
POP = optimisation(experiment=EXP,
weights=[0, 0.5, 0.25, 0.25],
preruncycles=10,
cycles=10,
seed=1,
outdes=5,
folder=os.getcwd())
#########################
# run natural selection #
#########################
POP.optimise()
POP.download()
POP.evaluate()
################
# step by step #
################