data = pd.read_csv('errors/li_durbin_errors_{0}.csv'.format(s))
traces = []
for n in islands:
traces.append(Box(y=list(data[str(n)]), name=n))
traces = Data(traces)
layout = Layout(
title='{0} switches errors'.format(s),
xaxis=XAxis(
title='Islands'
),
yaxis=YAxis(
title='Squared error'
)
)
fig = Figure(data=traces, layout=layout)
plot_url = py.plot(fig, filename='{0} switches'.format(s))
#########################
### Time distribution ###
#########################
file = 'individuals/{0}_switches/{1}_islands.json'
islands = 12
for s in switches:
individual = json.loads(open(file.format(s, islands)).read())
m = model.StSICMR(individual['n'], individual['T'], individual['M'])
plotting.plotModel(m, liDurbin_tk, liDurbin_lk, logScale=True,
save=str(s), show=False)
l0 = 1 / l_list[0]
t_list *= l0
l_list *= l0
#########################
### Genetic Algorithm ###
#########################
# Build a group of models
pop = genalg.Population(model.StSICMR, t_list, l_list,
maxIslands=100, switches=3,
size=1000, repetitions=5)
# Enhance them all
pop.enhance(200)
# Plot the best one
plotting.plotModel(pop.best.model, t_list, l_list,
logScale=True)
################
### Boxplots ###
################
#switches = range(2, 7)
#islands = range(2, 31)
#for s in switches:
# data = pd.DataFrame()
# for n in islands:
# errors = []
# minError = np.inf
# bestIndi = None
# for _ in range(20):
# pop = genalg.Population(model.StSICMR, liDurbin_tk, liDurbin_lk,
import sys
sys.path.append('../../lib/')
from inference import genalg
import model
import plotting
import psmcfit
data = psmcfit.get_psmc_history('Norduz.psmc')
times, lambdas = psmcfit.search_increase(data['times'], data['lambdas'])
l0 = 1 / lambdas[0]
times *= l0
lambdas *= l0
#~ # Build a group of models
pop = genalg.Population(model.StSICMR, times, lambdas,
maxIslands=100, switches=4,
size=1000, repetitions=4)
# Enhance them all
pop.enhance(1000)
# Plot the best one
plotting.plotModel(pop.best.model, times, lambdas, logScale=True)