def TestErroneous(self):
"""
Test weird shapes, like flat lines other things
"""
t = np.arange(self.t_max) / float(self.t_max)
x_1 = np.exp(-t) + np.random.randn(self.t_max)
x_flat = np.ones(self.t_max)
self.assertTrue(fit(t, x_flat)[0] > fit(t, x_1)[0],
msg="Constant better than exponential")
def TestNan(self):
"""
Test that nan values are higher than any non nan value
"""
t = np.arange(1E5)
x_1 = np.exp(-t)
nan_idx = np.random.randint(1E5, size=5)
x_nan = np.exp(-2 * t)
x_nan[nan_idx] = np.nan
self.assertTrue(fit(t, x_nan)[0] > fit(t, x_1)[0],
msg="Nan fit failure")
def TestErroneous(self):
"""
Test weird shapes, like flat lines other things
"""
t = np.arange(self.t_max) / float(self.t_max)
x_1 = np.exp(-t) + np.random.randn(self.t_max)
x_flat = np.ones(self.t_max)
self.assertTrue(
fit(t, x_flat)[0] > fit(t, x_1)[0],
msg="Constant better than exponential"
)
def TestLongTrial(self):
"""
Test extra long trial to check for floating point craziness
"""
t = np.arange(1E5)
x_1 = np.exp(-t)
x_2 = np.exp(-2 * t)
x_1_noise = np.exp(-t) + 0.1 * np.random.randn(1E5)
x_2_noise = np.exp(-2 * t) + 0.1 * np.random.randn(1E5)
self.assertTrue(fit(t, x_1)[0] > fit(t, x_2)[0],
msg="Long trial monotonicity failure")
self.assertTrue(fit(t, x_1_noise)[0] > fit(t, x_2_noise)[0],
msg="Long trial noisy monotonicity failure")
def TestNan(self):
"""
Test that nan values are higher than any non nan value
"""
t = np.arange(1E5)
x_1 = np.exp(-t)
nan_idx = np.random.randint(1E5, size=5)
x_nan = np.exp(-2 * t)
x_nan[nan_idx] = np.nan
self.assertTrue(
fit(t, x_nan)[0] > fit(t, x_1)[0],
msg="Nan fit failure"
)
def TestLongTrial(self):
"""
Test extra long trial to check for floating point craziness
"""
t = np.arange(1E5)
x_1 = np.exp(-t)
x_2 = np.exp(-2 * t)
x_1_noise = np.exp(-t) + 0.1 * np.random.randn(1E5)
x_2_noise = np.exp(-2 * t) + 0.1 * np.random.randn(1E5)
self.assertTrue(
fit(t, x_1)[0] > fit(t, x_2)[0],
msg="Long trial monotonicity failure"
)
self.assertTrue(
fit(t, x_1_noise)[0] > fit(t, x_2_noise)[0],
msg="Long trial noisy monotonicity failure"
)
def TestMonotonicity(self):
"""
Tests to see that objective function is monotonic
"""
t = np.arange(self.t_max) / float(self.t_max)
x_1 = np.exp(-t)
x_2 = np.exp(-2 * t)
x_3 = np.exp(-3 * t)
self.assertTrue(fit(t, x_1)[0] > fit(t, x_2)[0],
msg="Monotonicity failure")
self.assertTrue(fit(t, x_2)[0] > fit(t, x_3)[0],
msg="Monotonicity failure")
self.assertTrue(fit(t, x_1)[0] > fit(t, x_3)[0],
msg="Monotonicity failure")
def TestNoise(self):
"""
Monotonicity plus varying amount of noise
"""
perc_noise = 0.01
t = np.arange(self.t_max) / float(self.t_max)
x_1 = np.exp(-t) + np.random.randn(self.t_max) * perc_noise
x_2 = np.exp(-2 * t) + np.random.randn(self.t_max) * perc_noise
x_3 = np.exp(-3 * t) + np.random.randn(self.t_max) * perc_noise
self.assertTrue(fit(t, x_1)[0] > fit(t, x_2)[0],
msg="Monotonicity failure")
self.assertTrue(fit(t, x_2)[0] > fit(t, x_3)[0],
msg="Monotonicity failure")
self.assertTrue(fit(t, x_1)[0] > fit(t, x_3)[0],
msg="Monotonicity failure")
def TestEndNoise(self):
"""
Test that a bunch of noise at the end doesn't bork everything
"""
t = np.arange(self.t_max) / float(self.t_max)
x_1 = np.exp(-2 * t)
x_rand_end = np.ones(self.t_max)
x_rand_end[-3:] = 10 * np.random.randn(3)
x_pos_end = np.ones(self.t_max)
x_pos_end[-3:] *= 100
x_neg_end = np.ones(self.t_max)
x_neg_end[-3:] *= -100
self.assertTrue(fit(t, x_rand_end)[0] > fit(t, x_1)[0],
msg="Random end noise does better than exponential")
self.assertTrue(fit(t, x_pos_end)[0] > fit(t, x_1)[0],
msg="Positive end noise wins")
self.assertTrue(fit(t, x_neg_end)[0] > fit(t, x_1)[0],
msg="Negative end noise wins")
def TestMonotonicity(self):
"""
Tests to see that objective function is monotonic
"""
t = np.arange(self.t_max) / float(self.t_max)
x_1 = np.exp(-t)
x_2 = np.exp(-2 * t)
x_3 = np.exp(-3 * t)
self.assertTrue(
fit(t, x_1)[0] > fit(t, x_2)[0],
msg="Monotonicity failure"
)
self.assertTrue(
fit(t, x_2)[0] > fit(t, x_3)[0],
msg="Monotonicity failure"
)
self.assertTrue(
fit(t, x_1)[0] > fit(t, x_3)[0],
msg="Monotonicity failure"
)
def TestNoise(self):
"""
Monotonicity plus varying amount of noise
"""
perc_noise = 0.01
t = np.arange(self.t_max) / float(self.t_max)
x_1 = np.exp(-t) + np.random.randn(self.t_max) * perc_noise
x_2 = np.exp(-2 * t) + np.random.randn(self.t_max) * perc_noise
x_3 = np.exp(-3 * t) + np.random.randn(self.t_max) * perc_noise
self.assertTrue(
fit(t, x_1)[0] > fit(t, x_2)[0],
msg="Monotonicity failure"
)
self.assertTrue(
fit(t, x_2)[0] > fit(t, x_3)[0],
msg="Monotonicity failure"
)
self.assertTrue(
fit(t, x_1)[0] > fit(t, x_3)[0],
msg="Monotonicity failure"
)
def TestEndNoise(self):
"""
Test that a bunch of noise at the end doesn't bork everything
"""
t = np.arange(self.t_max) / float(self.t_max)
x_1 = np.exp(-2 * t)
x_rand_end = np.ones(self.t_max)
x_rand_end[-3:] = 10 * np.random.randn(3)
x_pos_end = np.ones(self.t_max)
x_pos_end[-3:] *= 100
x_neg_end = np.ones(self.t_max)
x_neg_end[-3:] *= -100
self.assertTrue(
fit(t, x_rand_end)[0] > fit(t, x_1)[0],
msg="Random end noise does better than exponential"
)
self.assertTrue(
fit(t, x_pos_end)[0] > fit(t, x_1)[0],
msg="Positive end noise wins"
)
self.assertTrue(
fit(t, x_neg_end)[0] > fit(t, x_1)[0],
msg="Negative end noise wins"
)
