def generate_data(N, Ngrid):
data = None
if (direction in ["uv", "vu"]):
if not os.path.exists(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid)):
data = bh.generate_uv_data(N, 50000, 5, Ngrid=Ngrid)
np.save(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid),
data)
else:
data = np.load("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(
N, Ngrid))
else:
if not os.path.exists(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N,
Ngrid)):
data = mh.generate_vh_data(N, 20000, 50, Ngrid=Ngrid)
np.save(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid),
data)
else:
data = np.load(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid))
#at the moment we are doing a u -> v / v -> h cross prediction.
if (direction in ["vu", "hv"]):
#switch the entries for the v -> u / h -> v prediction
tmp = data[0].copy()
data[0] = data[1].copy()
data[1] = tmp.copy()
return data
def generate_data(N, Ngrid):
data = None
if direction == "u":
if not os.path.exists(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid)):
data = bh.generate_uv_data(N, 20000, 5, Ngrid=Ngrid)
np.save(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid),
data)
else:
data = np.load("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(
N, Ngrid))
else:
if not os.path.exists(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N,
Ngrid)):
data = mh.generate_vh_data(N, 20000, 50, Ngrid=Ngrid)
np.save(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid),
data)
else:
data = np.load(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid))
shared_output_data[:] = data[0, :]
#blur the data
for n in range(ndata):
shared_input_data[n, :, :] = gaussian_filter(shared_output_data[n],
sigma=9.0)
def generate_data(N, Ngrid):
data = None
if direction == "u":
if not os.path.exists(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid)):
data = bh.generate_uv_data(N, 50000, 5, Ngrid=Ngrid)
np.save(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid),
data)
else:
data = np.load("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(
N, Ngrid))
else:
if not os.path.exists(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N,
Ngrid)):
data = mh.generate_vh_data(N, 20000, 50, Ngrid=Ngrid)
np.save(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid),
data)
else:
data = np.load(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid))
#fill the first ndata-predictionLength items with the real data and leave the last predictionLength items free
#these items will never be used, and are only created to reduce the needed code
shared_input_data[:-predictionLength] = data[0, :-predictionLength]
shared_output_data[:-predictionLength] = data[0, predictionLength:]
def generate_data(N, trans, sample_rate, Ngrid):
data = None
if (direction == "u"):
if (os.path.exists("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(
N, Ngrid)) == False):
data = bh.generate_uv_data(N, 50000, 5, Ngrid=Ngrid)
np.save(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid),
data)
else:
data = np.load("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(
N, Ngrid))
else:
if (os.path.exists(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(
N, Ngrid)) == False):
data = mh.generate_vh_data(N, 20000, 50, Ngrid=Ngrid)
np.save(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid),
data)
else:
data = np.load(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid))
#print("blurring...")
for t in range(ndata):
data[1, t, :, :] = gaussian_filter(data[0, t], sigma=9.0)
return data
def generate_data(N, trans, sample_rate, Ngrid):
data = None
if direction in ["uv", "vu"]:
if not os.path.exists(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid)):
data = bh.generate_uv_data(N, 50000, 5, Ngrid=Ngrid)
np.save(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid),
data)
else:
data = np.load("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(
N, Ngrid))
elif direction in ["bocf_uv", "bocf_uw", "bocf_us"]:
if not os.path.exists(
"../../cache/bocf/raw/{0}_{1}.uvws.dat.npy".format(N, Ngrid)):
print(
"NO BOCF data set found. Please generate a chaotic data set manually."
)
else:
data = np.load("../../cache/bocf/raw/{0}_{1}.uvws.dat.npy".format(
N, Ngrid))
else:
if not os.path.exists(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N,
Ngrid)):
data = mh.generate_vh_data(N, 20000, 50, Ngrid=Ngrid)
np.save(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid),
data)
else:
data = np.load(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid))
#at the moment we are doing a u -> v / v -> h cross prediction (index 0 -> index 1)
if (direction in ["vu", "hv"]):
#switch the entries for the v -> u / h -> v prediction
tmp = data[0].copy()
data[0] = data[1].copy()
data[1] = tmp.copy()
if direction in ["bocf_uv", "bocf_uw", "bocf_ws"]:
real_data = np.empty((2, N, Ngrid, Ngrid))
real_data[0] = data[0].copy()
if direction == "bocf_uv":
real_data[1] = data[1].copy()
elif direction == "bocf_uw":
real_data[1] = data[2].copy()
elif direction == "bocf_us":
real_data[1] = data[3].copy()
data = real_data
return data
def generate_data(N, trans, sample_rate, Ngrid):
data = None
if direction == "u":
if not os.path.exists(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid)):
data = bh.generate_uv_data(N, 50000, 5, Ngrid=Ngrid)
np.save(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid),
data)
else:
data = np.load("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(
N, Ngrid))
else:
if not os.path.exists(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N,
Ngrid)):
data = mh.generate_vh_data(N, 20000, 50, Ngrid=Ngrid)
np.save(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid),
data)
else:
data = np.load(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid))
data = data[0]
input_y, input_x, _, _ = create_patch_indices(
(center - (halfInnerSize + borderSize), center +
(halfInnerSize + borderSize) + rightBorderAdd),
(center - (halfInnerSize + borderSize), center +
(halfInnerSize + borderSize) + rightBorderAdd),
(center - (halfInnerSize), center + (halfInnerSize) + rightBorderAdd),
(center - (halfInnerSize), center + (halfInnerSize) + rightBorderAdd))
_, _, output_y, output_x = create_patch_indices(
(center - (halfInnerSize + borderSize), center +
(halfInnerSize + borderSize) + rightBorderAdd),
(center - (halfInnerSize + borderSize), center +
(halfInnerSize + borderSize) + rightBorderAdd),
(center - (1), center + (1) + 0), (center - (1), center + (1) + 0))
inputData = data[:, input_y, input_x]
outputData = data[:, output_y, output_x]
return inputData, outputData
def generate_data(N, trans, sample_rate, Ngrid):
data = None
if direction == "u":
if not os.path.exists("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid)):
data = bh.generate_uv_data(N, 50000, 5, Ngrid=Ngrid)
np.save("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid), data)
else:
data = np.load("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid))
else:
if not os.path.exists("../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid)):
data = mh.generate_vh_data(N, 20000, 50, Ngrid=Ngrid)
np.save("../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid), data)
else:
data = np.load("../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid))
return data[0]
else:
data = np.load("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(
ndata, N))
elif direction in ["bocf_uv", "bocf_uw", "bocf_us"]:
if not os.path.exists("../../cache/bocf/raw/{0}_{1}.uvws.dat.npy".format(
ndata, N)):
print(
"NO BOCF data set found. Please generate a chaotic data set manually."
)
else:
data = np.load("../../cache/bocf/raw/{0}_{1}.uvws.dat.npy".format(
ndata, N))
else:
if not os.path.exists("../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(
N, N)):
data = mh.generate_vh_data(N, 20000, 50, Ngrid=N)
np.save("../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(ndata, N),
data)
else:
data = np.load("../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(
ndata, N))
print("Data loaded")
#at the moment we are doing a u -> v / v -> h cross prediction (index 0 -> index 1)
if (direction in ["vu", "hv"]):
#switch the entries for the v -> u / h -> v prediction
tmp = data[0].copy()
data[0] = data[1].copy()
data[1] = tmp.copy()
def generate_data(N, trans, sample_rate=1, Ngrid=100):
#return the u variable
return generate_vh_data(N, trans, sample_rate=sample_rate, Ngrid=Ngrid)[0]
def generate_data(N, Ngrid):
data = None
if direction in ["uv", "vu"]:
if not os.path.exists(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid)):
data = bh.generate_uv_data(N, 50000, 5, Ngrid=Ngrid)
np.save(
"../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(N, Ngrid),
data)
else:
data = np.load("../../cache/barkley/raw/{0}_{1}.uv.dat.npy".format(
N, Ngrid))
elif direction.startswith("bocf_"):
if not os.path.exists(
"../../cache/bocf/raw/{0}_{1}.uvws.dat.npy".format(N, Ngrid)):
print(
"NO BOCF data set found. Please generate a chaotic data set manually."
)
else:
data = np.load("../../cache/bocf/raw/{0}_{1}.uvws.dat.npy".format(
N, Ngrid))
else:
if not os.path.exists(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N,
Ngrid)):
data = mh.generate_vh_data(N, 20000, 50, Ngrid=Ngrid)
np.save(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid),
data)
else:
data = np.load(
"../../cache/mitchell/raw/{0}_{1}.vh.dat.npy".format(N, Ngrid))
#at the moment we are doing a u -> v / v -> h cross prediction (index 0 -> index 1)
if (direction in ["vu", "hv"]):
#switch the entries for the v -> u / h -> v prediction
tmp = data[0].copy()
data[0] = data[1].copy()
data[1] = tmp.copy()
if direction.startswith("bocf_"):
real_data = np.empty((2, N, Ngrid, Ngrid))
if direction[5] == "u":
real_data[0] = data[0].copy()
elif direction[5] == "v":
real_data[0] = data[1].copy()
elif direction[5] == "w":
real_data[0] = data[2].copy()
elif direction[5] == "s":
real_data[0] = data[3].copy()
if direction[6] == "u":
real_data[1] = data[0].copy()
elif direction[6] == "v":
real_data[1] = data[1].copy()
elif direction[6] == "w":
real_data[1] = data[2].copy()
elif direction[6] == "s":
real_data[1] = data[3].copy()
data = real_data
global means_train
means_train = [0, 0] #np.mean(data[:trainLength], axis=(1, 2))
data[0] -= means_train[0]
data[1] -= means_train[1]
if cpmtps.use_noise:
data[0] += np.random.normal(loc=0.0, scale=noise, size=data[0].shape)
if not direction.startswith("bocf"):
data[0][data[0] < 0.0] = 0.0
data[0][data[0] > 1.0] = 1.0
shared_input_data[:ndata] = data[0]
shared_output_data[:ndata] = data[1]