optmu2 = mu2init
optcor = 0
for j1 in range(7):
for j2 in range(7):
mu = muinit * (ps_step**j1)
mu2 = mu2init * (ps_step**j2)
cor = 0
for k in range(5): # 5 for full 5 fold CV
train1_idx = SP.concatenate(
(train_idx[:int(n_train * k * 0.2)],
train_idx[int(n_train * (k + 1) * 0.2):n_train]))
valid_idx = train_idx[int(n_train * k *
0.2):int(n_train * (k + 1) *
0.2)]
res1 = lmm_lasso.train(X[train1_idx],
K[train1_idx][:, train1_idx],
y[train1_idx], mu, mu2, group)
w1 = res1['weights']
yhat = lmm_lasso.predict(y[train1_idx], X[train1_idx, :],
X[valid_idx, :],
K[train1_idx][:, train1_idx],
K[valid_idx][:, train1_idx],
res1['ldelta0'], w1)
cor += SP.dot(yhat.T - yhat.mean(), y[valid_idx] -
y[valid_idx].mean()) / (yhat.std() *
y[valid_idx].std())
print mu, mu2, cor[0, 0]
if cor > optcor:
optcor = cor
optmu = mu
for k in range(n_reps):
opterr = 10000
optw = 0
stop = 4
w0 = 0
nm = {}
mu = muinit
train_idx = SP.random.permutation(SP.arange(n_s))
for j in range(20):
mu *= ps_step
train1_idx = train_idx[:int(n_s * 0.8)]
valid_idx = train_idx[int(n_s * 0.8):]
res = lmm_lasso.train(X[train1_idx],
K[train1_idx][:, train1_idx],
y[train1_idx],
mu,
0, [],
w0=w0,
null_model=nm)
w0 = res['weights']
nm = res['null_model']
yhat = lmm_lasso.predict(y[train1_idx], X[train1_idx, :],
X[valid_idx, :],
K[train1_idx][:, train1_idx],
K[valid_idx][:, train1_idx],
res['ldelta0'], res['lstsq'])
err = LA.norm(yhat - y[valid_idx])
print mu, err
if err < opterr:
optw = w0
opterr = err