def run_log_regression():
nb_in_sample = 100
nb_out_of_sample = 100000
nb_runs = 100
nb_epochs = 0
nb_Eout = 0
lr = 0.01
eps = 0.01
for i in range(nb_runs):
# generate random function
l = randomline()
f = target_random_function(l)
# generate in sample data and out of sample data
data_in_sample = data(nb_in_sample)
data_out_of_sample = data(nb_out_of_sample)
# create training set structure [[w0,w1,...],target_value]
t_set_in = build_training_set_fmultipleparams(data_in_sample,f)
t_set_out = build_training_set_fmultipleparams(data_out_of_sample,f)
# run logistic regression in sample
epochs,w = log_regression_sgd(t_set_in,eps,lr)
# compute the out of sample error given the previously compute weights.
e_out = log_regression_compute_Eout(t_set_out,w)
print "Run: %s - epochs: %s"%(i, epochs)
print "Eout: %s"%(e_out)
nb_Eout += e_out
nb_epochs += epochs
print 'Number of runs:%s'%(nb_runs)
print "Avg epochs: %s"%(nb_epochs / nb_runs*1.0)
print "Avg Eout: %s"%(nb_Eout / nb_runs*1.0)
def run_log_regression():
nb_in_sample = 100
nb_out_of_sample = 100000
nb_runs = 100
nb_epochs = 0
nb_Eout = 0
lr = 0.01
eps = 0.01
for i in range(nb_runs):
# generate random function
l = randomline()
f = target_random_function(l)
# generate in sample data and out of sample data
data_in_sample = data(nb_in_sample)
data_out_of_sample = data(nb_out_of_sample)
# create training set structure [[w0,w1,...],target_value]
t_set_in = build_training_set_fmultipleparams(data_in_sample, f)
t_set_out = build_training_set_fmultipleparams(data_out_of_sample, f)
# run logistic regression in sample
epochs, w = log_regression_sgd(t_set_in, eps, lr)
# compute the out of sample error given the previously compute weights.
e_out = log_regression_compute_Eout(t_set_out, w)
print "Run: %s - epochs: %s" % (i, epochs)
print "Eout: %s" % (e_out)
nb_Eout += e_out
nb_epochs += epochs
print 'Number of runs:%s' % (nb_runs)
print "Avg epochs: %s" % (nb_epochs / nb_runs * 1.0)
print "Avg Eout: %s" % (nb_Eout / nb_runs * 1.0)
def generate_t_set(N, f=None):
'''
Generate a training set of N = 1000 points on X = [1; 1] * [1; 1] with uniform
probability of picking each x that belongs X . Generate simulated noise by
fipping the sign of a random 10% subset of the generated training set
'''
d = data(N)
if f is None:
f = lambda x: sign(x[0]**2 + x[1]**2 - 0.6)
t_set = build_training_set_fmultipleparams(d, f)
t_set = t_set_errorNoise(t_set, N / 10)
return t_set, f
def generate_t_set(N,f=None):
'''
Generate a training set of N = 1000 points on X = [1; 1] * [1; 1] with uniform
probability of picking each x that belongs X . Generate simulated noise by
fipping the sign of a random 10% subset of the generated training set
'''
d = data(N)
if f is None:
f = lambda x: sign(x[0]**2 + x[1]**2 -0.6)
t_set = build_training_set_fmultipleparams(d,f)
t_set = t_set_errorNoise(t_set,N/10)
return t_set,f
