def make_bagset(in_path, out_path, k=7):
seq_dict = files.get_seqs(in_path)
files.make_dir(out_path)
for i in range(k):
dataset_i = resample(seq_dict)
out_i = "%s/bag%d" % (out_path, i)
files.save_seqs(dataset_i, out_i)
files.save_seqs(seq_dict, "%s/full" % out_path)
def simple_exp(in_path, out_path=None, n_epochs=1000):
seq_dict = files.get_seqs(in_path)
train, test = files.split(seq_dict)
X, y, names, params = prepare_data(train)
model = deep.make_conv(params)
model.fit(X, y, epochs=n_epochs)
if (out_path):
extrac_feats(model, seq_dict, out_path)
else:
test_model(model, test)
def norm_seqs(in_path, out_path):
seqs = files.get_seqs(in_path)
fun = sigma_norm(seqs)
new_seqs = {}
for name_i, seq_i in seqs.items():
seq_i = [fun(ts_j, j) for j, ts_j in enumerate(seq_i.T)]
seq_i = np.array(seq_i).T
new_seqs[name_i] = seq_i
files.make_dir(out_path)
for name_i, seq_i in new_seqs.items():
out_i = "%s/%s" % (out_path, name_i)
np.save(out_i, seq_i)
def agum_template(in_path, out_path, agum):
seq_dict = files.get_seqs(in_path)
train, test = files.split(seq_dict)
agum_train = []
for name_i, seq_i in train.items():
agum_train.append((name_i, seq_i))
new_seqs = [agum_k(seq_i) for agum_k in agum]
new_seqs = list(itertools.chain(*new_seqs))
for j, seq_j in enumerate(new_seqs):
name_j = "%s_%d" % (name_i, j)
agum_train.append((name_j, seq_j))
agum_data = agum_train + list(test.items())
agum_data = dict(agum_data)
files.save_seqs(agum_data, out_path)
print(len(agum_data))
def show_ts(in_path, out_path):
seqs = files.get_seqs(in_path)
dim = list(seqs.values())[0].shape[-1]
cats = get_cats(seqs.keys())[0]
files.make_dir(out_path)
for cat_i, names_i in cats.items():
in_i = "%s/%d" % (out_path, cat_i)
files.make_dir(in_i)
for name_ij in names_i:
out_ij = "%s/%s" % (in_i, name_ij)
seq_ij = seqs[name_ij]
for ts_k in seq_ij.T:
plt.plot(ts_k)
plt.savefig(out_ij)
plt.clf()
plt.close()
def jackknife(in_path, out_path):
seq_dict = files.get_seqs(in_path)
dim = list(seq_dict.values())[0].shape[1]
X = np.array(list(seq_dict.values()))
names = list(seq_dict.keys())
files.make_dir(out_path)
for i in range(dim):
sub_i = np.delete(X, [i], axis=2)
out_i = "%s/bag%d" % (out_path, i)
sub_dict_i = {name_j: sub_i[j] for j, name_j in enumerate(names)}
files.save_seqs(sub_dict_i, out_i)
files.save_seqs(seq_dict, "%s/full" % out_path)
#make_bagset("../MSR/agum","../MSR/bagging",k=7)
#train_bag("../MSR/subspace","../MSR/sub_feats")
#jackknife("../MSR/agum","../MSR/subspace")
def ae_exp(in_path, out_path):
files.make_dir(out_path)
seq_dict = files.get_seqs(in_path)
pairs = dtw.make_pairwise_distance(seq_dict)
save(pairs, "%s/%s" % (out_path, "pairs"))
def compute_stats(in_path,out_path):
seqs=files.get_seqs(in_path)
feat_dict=feats.Feats()
for name_i,seq_i in seqs.items():
feat_dict[name_i]=EBTF(seq_i)
feat_dict.save(out_path)
def upsample(in_path, out_path, size=128):
seq_dict = files.get_seqs(in_path)
print(len(seq_dict))
spline = SplineUpsampling(size)
seq_dict = {name_i: spline(seq_i) for name_i, seq_i in seq_dict.items()}
files.save_seqs(seq_dict, out_path)