def experiment_loop_comparative(
base_dat: str,
cmip_dat: str,
variable: str,
spatial_window: int,
subsample: Optional[int] = None,
trial: int = 1,
regrid_filename: Optional[str] = None,
) -> Dict:
"""Performs one experimental loop for calculating the IT measures
for the models"""
# 1.1) get base model
base_dat = get_base_model(base_dat, variable)
# 1.2) get cmip5 model
cmip_dat = get_cmip5_model(cmip_dat, variable)
# 2) regrid data w. reference grid and
reference_ds = get_reference_dataset("noresm1_m")
base_dat = regrid_data(reference_ds,
base_dat,
filename=regrid_filename + ".nc")
cmip_dat = regrid_data(reference_ds,
cmip_dat,
filename=regrid_filename + ".nc")
# 3) find overlapping times
base_dat, cmip_dat = get_time_overlap(base_dat, cmip_dat)
# 4) get density cubes
base_df = get_spatial_cubes(base_dat, spatial_window)
cmip_df = get_spatial_cubes(cmip_dat, spatial_window)
# 5) normalize data
base_df = normalize_data(base_df)
cmip_df = normalize_data(cmip_df)
# 6) Resample/ Subsample data
base_df = resample(base_df, n_samples=subsample, random_state=trial)
cmip_df = resample(cmip_df, n_samples=subsample, random_state=trial)
# 7.1) Mutual Information
mutual_info, time_mi = run_rbig_models(base_df,
cmip_df,
measure="mi",
verbose=None)
# 7.2 - Pearson, Spearman, KendelTau
pearson = stats.pearsonr(base_df.ravel(), cmip_df.ravel())[0]
spearman = stats.spearmanr(base_df.ravel(), cmip_df.ravel())[0]
kendelltau = stats.kendalltau(base_df.ravel(), cmip_df.ravel())[0]
results = {
"mi": mutual_info,
"time_mi": time_mi,
"pearson": pearson,
"spearman": spearman,
"kendelltau": kendelltau,
}
return results
def experiment_loop_individual(
base: str,
cmip: str,
variable: str,
spatial_window: int,
subsample: Optional[int] = None,
trial: int = 1,
batch_size: Optional[int] = None,
regrid_filename: Optional[str] = None,
) -> Dict:
"""Performs one experimental loop for calculating the IT measures
for the models"""
# 1.1) get base model
base_dat = get_base_model(base, variable)
# 1.2) get cmip5 model
cmip_dat = get_cmip5_model(cmip, variable)
# 2) regrid data w. reference grid and
reference_ds = get_reference_dataset("ipsl_cm5a_lr")
base_dat = regrid_data(reference_ds,
base_dat,
filename=regrid_filename + ".nc")
cmip_dat = regrid_data(reference_ds,
cmip_dat,
filename=regrid_filename + ".nc")
# 3) find overlapping times
base_dat, cmip_dat = get_time_overlap(base_dat, cmip_dat)
# 4) get density cubes
base_df = get_spatial_cubes(base_dat, spatial_window)
cmip_df = get_spatial_cubes(cmip_dat, spatial_window)
# 5) normalize data
base_df = normalize_data(base_df)
cmip_df = normalize_data(cmip_df)
# 6) Resample/ Subsample data
base_df = resample(base_df, n_samples=subsample, random_state=trial)
cmip_df = resample(cmip_df, n_samples=subsample, random_state=trial)
# 7.1 - Entropy
tc_base, h_base, h_time_base = run_rbig_models(base_df,
measure="h",
verbose=None)
# 7.1 - Total Correlation
tc_cmip, h_cmip, h_time_cmip = run_rbig_models(cmip_df,
measure="h",
verbose=None)
results = {
"h_base": h_base,
"tc_base": tc_base,
"h_cmip": h_cmip,
"tc_cmip": tc_cmip,
"t_base": h_time_base,
"t_cmip": h_time_cmip,
}
return results
def experiment_loop_comparative(
base: str,
cmip: str,
variable: str,
spatial_window: int,
subsample: Optional[int] = None,
batch_size: Optional[int] = None,
) -> Dict:
"""Performs one experimental loop for calculating the IT measures
for the models"""
# 1.1) get base model
base_dat = get_base_model(base, variable)
# 1.2) get cmip5 model
cmip_dat = get_cmip5_model(cmip, variable)
# 2) regrid data
base_dat, cmip_dat = regrid_2_lower_res(base_dat, cmip_dat)
# 3) find overlapping times
base_dat, cmip_dat = get_time_overlap(base_dat, cmip_dat)
# 4) get density cubes
base_df = get_spatial_cubes(base_dat, spatial_window)
cmip_df = get_spatial_cubes(cmip_dat, spatial_window)
# 5) normalize data
base_df = normalize_data(base_df)
cmip_df = normalize_data(cmip_df)
# 7.1) Mutual Information
print("Mutual Information")
mutual_info, time_mi = run_rbig_models(
base_df[:subsample],
cmip_df[:subsample],
measure="mi",
verbose=1,
batch_size=batch_size,
)
# 7.2 - Pearson, Spearman, KendelTau
print("Pearson")
pearson = stats.pearsonr(base_df[:subsample].ravel(),
cmip_df[:subsample].ravel())[0]
print("Spearman")
spearman = stats.spearmanr(base_df[:subsample].ravel(),
cmip_df[:subsample].ravel())[0]
print("KendelTau")
kendelltau = stats.kendalltau(base_df[:subsample].ravel(),
cmip_df[:subsample].ravel())[0]
results = {
"mi": mutual_info,
"time_mi": time_mi,
"pearson": pearson,
"spearman": spearman,
"kendelltau": kendelltau,
}
return results
def process_loop_compare(
base_dat: str,
cmip_dat: str,
spatial_window: int,
subsample: Optional[int] = None,
trial: int = 1,
) -> Dict:
"""Performs one experimental loop for calculating the IT measures
for the models"""
# 1.1) get base model
# get time stamp
base_time_stamp = base_dat.time.values[0]
cmip_time_stamp = cmip_dat.time.values[0]
# 4) get density cubes
base_df = get_spatial_cubes(base_dat, spatial_window)
cmip_df = get_spatial_cubes(cmip_dat, spatial_window)
# 5) normalize data
base_df = normalize_data(base_df)
cmip_df = normalize_data(cmip_df)
# 6) Resample/ Subsample data
base_df = resample(base_df, n_samples=subsample, random_state=trial)
cmip_df = resample(cmip_df, n_samples=subsample, random_state=trial)
# 7.1) Mutual Information
mutual_info, time_mi = run_rbig_models(base_df,
cmip_df,
measure="mi",
verbose=None)
# 7.2 - Pearson, Spearman, KendelTau
pearson = stats.pearsonr(base_df.ravel(), cmip_df.ravel())[0]
spearman = stats.spearmanr(base_df.ravel(), cmip_df.ravel())[0]
kendelltau = stats.kendalltau(base_df.ravel(), cmip_df.ravel())[0]
results = {
"base_time_stamp": base_time_stamp,
"cmip_time_stamp": cmip_time_stamp,
"mi": mutual_info,
"time_mi": time_mi,
"pearson": pearson,
"spearman": spearman,
"kendelltau": kendelltau,
}
return results
def process_loop_individual(
base_dat: str,
cmip_dat: str,
spatial_window: int,
subsample: Optional[int] = None,
trial: int = 1,
batch_size: Optional[int] = None,
) -> Dict:
# get time stamp
base_time_stamp = base_dat.time.values[0]
cmip_time_stamp = cmip_dat.time.values[0]
# 4) get density cubes
base_df = get_spatial_cubes(base_dat, spatial_window)
cmip_df = get_spatial_cubes(cmip_dat, spatial_window)
# 5) normalize data
base_df = normalize_data(base_df)
cmip_df = normalize_data(cmip_df)
# 6) Resample/ Subsample data
base_df = resample(base_df, n_samples=subsample, random_state=trial)
cmip_df = resample(cmip_df, n_samples=subsample, random_state=trial)
# 7.1 - Entropy
tc_base, h_base, h_time_base = run_rbig_models(base_df,
measure="h",
verbose=None)
tc_cmip, h_cmip, h_time_cmip = run_rbig_models(cmip_df,
measure="h",
verbose=None)
results = {
"base_time_stamp": base_time_stamp,
"cmip_time_stamp": cmip_time_stamp,
"h_base": h_base,
"h_cmip": h_cmip,
"tc_base": tc_base,
"tc_cmip": tc_cmip,
"t_base": h_time_base,
"t_cmip": h_time_cmip,
}
return results
def experiment_loop_individual(
base: str,
cmip: str,
variable: str,
spatial_window: int,
subsample: Optional[int] = None,
batch_size: Optional[int] = None,
) -> Dict:
"""Performs one experimental loop for calculating the IT measures
for the models"""
# 1.1) get base model
base_dat = get_base_model(base, variable)
# 1.2) get cmip5 model
cmip_dat = get_cmip5_model(cmip, variable)
# 2) regrid data
base_dat, cmip_dat = regrid_2_lower_res(base_dat, cmip_dat)
# 3) find overlapping times
base_dat, cmip_dat = get_time_overlap(base_dat, cmip_dat)
# 4) get density cubes
base_df = get_spatial_cubes(base_dat, spatial_window)
cmip_df = get_spatial_cubes(cmip_dat, spatial_window)
# 5) normalize data
base_df = normalize_data(base_df)
cmip_df = normalize_data(cmip_df)
# 7.1) Total Correlation
print("TOTAL CORRELATION")
print("Base Model")
tc_base, time_base = run_rbig_models(base_df[:subsample],
measure="t",
verbose=1,
batch_size=batch_size)
print("CMIP Model")
tc_cmip, time_cmip = run_rbig_models(cmip_df[:subsample],
measure="t",
verbose=1,
batch_size=batch_size)
# 7.2 - ENTROPY
print("ENTROPY")
print("Base Model")
h_base, time_base = run_rbig_models(base_df[:subsample],
measure="h",
verbose=1,
batch_size=batch_size)
print("CMIP Model")
h_cmip, time_cmip = run_rbig_models(cmip_df[:subsample],
measure="h",
verbose=1,
batch_size=batch_size)
results = {
"h_base": h_base,
"tc_base": tc_base,
"h_cmip": h_cmip,
"tc_cmip": tc_cmip,
"t_base": time_base,
"t_cmip": time_cmip,
}
return results