def optimize_logit_for(self, pair):
# extract the right images from the set
[a,b] = [int(pair[0]), int(pair[1])]
data = np.array(pd.read_csv('./data/kaggle/train.csv', header=0)).astype('float64')
data_t = data[0:, :]
data_a = data_t[data_t[:,0]==a]
data_b = data_t[data_t[:,0]==b]
data_ab = np.vstack([data_a, data_b])
xt = data_ab[:, 1:]
yt = data_ab[:, 0].astype('int8')
yt = np.atleast_2d(yt).T
print 'xt shape', xt.shape
# Perform logistic regression
xt2 = np.column_stack([np.ones((xt.shape[0],1)), xt])
yt2 = yt.flatten()
count = 0
for i in range(yt2.size):
if yt2[i]==a:
yt2[i] = 1
count +=1
else:
yt2[i] = 0
ini_thetas = 0.005*np.random.rand(xt2.shape[1],1)
L = 1e+5
opt_thetas = logit.optimizeThetas(ini_thetas, xt2, yt2, L, visual=False)
return opt_thetas
def test_optimize_thetas_reg():
l = Logysterical(tr, None)
ini_thetas = np.zeros((l.tr["X"].shape[1], 1))
pp = l.optimizeThetas(ini_thetas)
print pp
diff = pp - np.array(
[
1.273005,
0.624876,
1.177376,
-2.020142,
-0.912616,
-1.429907,
0.125668,
-0.368551,
-0.360033,
-0.171068,
-1.460894,
-0.052499,
-0.618889,
-0.273745,
-1.192301,
-0.240993,
-0.207934,
-0.047224,
-0.278327,
-0.296602,
-0.453957,
-1.045511,
0.026463,
-0.294330,
0.014381,
-0.328703,
-0.143796,
-0.924883,
]
)
for e in diff:
assert e < 0.01
