Skip to content

Commit

Permalink
update Predictor
Browse files Browse the repository at this point in the history
  • Loading branch information
@EhsanGharibNezhad
EhsanGharibNezhad committed Apr 11, 2024
1 parent d531ea2 commit ba9abef
Showing 1 changed file with 52 additions and 52 deletions.
104 changes: 52 additions & 52 deletions TelescopeML/Predictor.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -189,7 +189,7 @@ def ProcessObservationalDataset(self,
Return
-------
Fnu_obs , Fnu_obs_err, Fnu_obs_absolute, Fnu_obs_absolute_err
Fnu_obs , Fnu_obs_err, Fnu_obs_TOA, Fnu_obs_TOA_err
Example
-------
Expand All @@ -215,13 +215,13 @@ def ProcessObservationalDataset(self,
Flam_errors = self.F_lambda_obs_err,
wavelengths = wl_obs,) # Convert F_lambda_obs to Fnu_obs

Fnu_obs_absolute, Fnu_obs_absolute_err = self.Fnu_to_Fnu_abs(
Fnu_obs_TOA, Fnu_obs_TOA_err = self.Fnu_to_Fnu_abs(
Fnu_values = Fnu_obs,
Fnu_errors = Fnu_obs_err,
bd_literature_dic = bd_literature_dic)


return Fnu_obs , Fnu_obs_err, Fnu_obs_absolute, Fnu_obs_absolute_err
return Fnu_obs , Fnu_obs_err, Fnu_obs_TOA, Fnu_obs_TOA_err


def Fnu_to_Fnu_abs(self,
Expand All @@ -231,7 +231,7 @@ def Fnu_to_Fnu_abs(self,
__plot__ = False
):

# Calculate the absolute Fnu_obs values
# Calculate the TOA Fnu_obs values
bd_literature_dic = self.bd_literature_dic if bd_literature_dic is None else bd_literature_dic

Fnu_values_abs = Fnu_values * (bd_literature_dic['bd_distance_pc'] * ((u.pc).to(u.jupiterRad)) / (
Expand Down Expand Up @@ -289,7 +289,7 @@ def Flam_to_Fnu(self,
return Fnu_values, Fnu_errors

def flux_interpolated(self,
Fnu_obs_absolute,
Fnu_obs_TOA,
interpolated_wl = None,
__print_results__=False,
__plot_spectra_errorbar__=False,
Expand All @@ -308,28 +308,28 @@ def flux_interpolated(self,
Return
-------
Fnu_obs_absolute_intd, Fnu_obs_absolute_intd_df
Fnu_obs_TOA_intd, Fnu_obs_TOA_intd_df
"""
Fnu_obs_absolute = self.Fnu_obs_absolute if Fnu_obs_absolute is None else Fnu_obs_absolute
Fnu_obs_TOA = self.Fnu_obs_TOA if Fnu_obs_TOA is None else Fnu_obs_TOA
interpolated_wl = np.sort(self.wl_synthetic.wl.values) if interpolated_wl is None else interpolated_wl

if __use_spectres__:
Fnu_obs_absolute_intd = spectres.spectres(interpolated_wl,
Fnu_obs_TOA_intd = spectres.spectres(interpolated_wl,
np.float64(self.wl_obs),
np.float64(Fnu_obs_absolute))
np.float64(Fnu_obs_TOA))
else:
flux_intd = pchip(self.wl_obs, self.Fnu_obs_absolute)
Fnu_obs_absolute_intd = flux_intd(interpolated_wl)
flux_intd = pchip(self.wl_obs, self.Fnu_obs_TOA)
Fnu_obs_TOA_intd = flux_intd(interpolated_wl)

Fnu_obs_absolute_intd_df = pd.DataFrame(list(Fnu_obs_absolute_intd),
Fnu_obs_TOA_intd_df = pd.DataFrame(list(Fnu_obs_TOA_intd),
index=[str(x) for x in np.round(interpolated_wl, 3)]).T


if __print_results__:
print('--- Object Flux ----')
print(self.Fnu_obs_absolute_intd_df)
print(self.Fnu_obs_TOA_intd_df)
print('--- Object Flux Interpolate ----')
print(pd.DataFrame(self.Fnu_obs_absolute_intd))
print(pd.DataFrame(self.Fnu_obs_TOA_intd))

if __plot_spectra_errorbar__:
plot_spectra_errorbar(
Expand All @@ -351,13 +351,13 @@ def flux_interpolated(self,
plot_spectra_errorbar(
object_name=self.object_name,
x_obs=self.obs_data_df['wl'],
y_obs=self.Fnu_obs_absolute,
y_obs_err=self.Fnu_obs_absolute_err,
y_obs=self.Fnu_obs_TOA,
y_obs_err=self.Fnu_obs_TOA_err,
y_label='Flux_abs (F_abs)',
)


return Fnu_obs_absolute_intd, Fnu_obs_absolute_intd_df
return Fnu_obs_TOA_intd, Fnu_obs_TOA_intd_df



Expand All @@ -381,37 +381,37 @@ def Process_Observational_Dataset(self,
F_lambda_obs = self.obs_data_df['F_lambda_obs'] if F_lambda_obs == None else F_lambda_obs
F_lambda_obs_err = self.obs_data_df['F_lambda_obs_err'] if F_lambda_obs_err == None else F_lambda_obs_err

Fnu_obs , Fnu_obs_err, Fnu_obs_absolute, Fnu_obs_absolute_err = self.ProcessObservationalDataset(
Fnu_obs , Fnu_obs_err, Fnu_obs_TOA, Fnu_obs_TOA_err = self.ProcessObservationalDataset(
F_lambda_obs=F_lambda_obs.values,
F_lambda_obs_err=F_lambda_obs_err.values,
wl_obs=self.obs_data_df['wl'].values,
# wl_synthetic=self.wl_synthetic['wl'].values,
bd_literature_dic=self.bd_literature_dic,
)

# self.Fnu_obs, Fnu_obs_err, self.Fnu_obs_absolute, self.Fnu_obs_absolute_err = Fnu_obs , Fnu_obs_err, Fnu_obs_absolute, Fnu_obs_absolute_err
# self.Fnu_obs, Fnu_obs_err, self.Fnu_obs_TOA, self.Fnu_obs_TOA_err = Fnu_obs , Fnu_obs_err, Fnu_obs_TOA, Fnu_obs_TOA_err
self.obs_data_df['Fnu_obs'] = Fnu_obs
self.obs_data_df['Fnu_obs_err'] = Fnu_obs_err
self.obs_data_df['Fnu_obs_absolute'] = Fnu_obs_absolute
self.obs_data_df['Fnu_obs_absolute_err'] = Fnu_obs_absolute_err
self.obs_data_df['Fnu_obs_TOA'] = Fnu_obs_TOA
self.obs_data_df['Fnu_obs_TOA_err'] = Fnu_obs_TOA_err

# Extract the original ML features from the observational spectrum
self.Fnu_obs_absolute_intd, self.Fnu_obs_absolute_intd_df = \
self.flux_interpolated(Fnu_obs_absolute = Fnu_obs_absolute,
self.Fnu_obs_TOA_intd, self.Fnu_obs_TOA_intd_df = \
self.flux_interpolated(Fnu_obs_TOA = Fnu_obs_TOA,
interpolated_wl = None,
__print_results__=False,
__plot_spectra_errorbar__=False,
__use_spectres__=True)

if __print_results__:
print('------------ Interpolated Observational Spectra: Absolute F𝜈 ------------')
print(self.Fnu_obs_absolute_intd_df)
print('------------ Interpolated Observational Spectra: TOA F𝜈 ------------')
print(self.Fnu_obs_TOA_intd_df)

# Extract the engineered ML features from the observational spectrum
df_Fnu_obs_absolute_intd_min = self.Fnu_obs_absolute_intd_df.min(axis=1)
df_Fnu_obs_absolute_intd_max = self.Fnu_obs_absolute_intd_df.max(axis=1)
df_Fnu_obs_TOA_intd_min = self.Fnu_obs_TOA_intd_df.min(axis=1)
df_Fnu_obs_TOA_intd_max = self.Fnu_obs_TOA_intd_df.max(axis=1)

self.df_MinMax_obs = pd.DataFrame((df_Fnu_obs_absolute_intd_min, df_Fnu_obs_absolute_intd_max)).T
self.df_MinMax_obs = pd.DataFrame((df_Fnu_obs_TOA_intd_min, df_Fnu_obs_TOA_intd_max)).T

if __print_results__:
print('------------ df_MinMax Single Observational Spectrum ------------')
Expand All @@ -422,10 +422,10 @@ def Process_Observational_Dataset(self,

# XminXmax_Stand = self.trained_data_processor.normalize_X_ColumnWise.transform(self.df_MinMax_obs.values)

bd_mean = self.Fnu_obs_absolute_intd_df.mean(axis=1)[0]
bd_std = self.Fnu_obs_absolute_intd_df.std(axis=1)[0]
bd_mean = self.Fnu_obs_TOA_intd_df.mean(axis=1)[0]
bd_std = self.Fnu_obs_TOA_intd_df.std(axis=1)[0]

X_Scaled = (self.Fnu_obs_absolute_intd_df.values[0] - bd_mean) / bd_std
X_Scaled = (self.Fnu_obs_TOA_intd_df.values[0] - bd_mean) / bd_std


y_pred_train = np.array(
Expand Down Expand Up @@ -505,7 +505,7 @@ def predict_from_random_spectra(
spectra['F_lambda_obs'].interpolate(inplace=True)

# Process the randomly generated Observational spectra
Fnu_obs , Fnu_obs_err, Fnu_obs_absolute, Fnu_obs_absolute_err = self.ProcessObservationalDataset(
Fnu_obs , Fnu_obs_err, Fnu_obs_TOA, Fnu_obs_TOA_err = self.ProcessObservationalDataset(
F_lambda_obs=spectra['F_lambda_obs'].values,
F_lambda_obs_err=self.obs_data_df['F_lambda_obs_err'].values,
wl_obs=self.obs_data_df['wl'].values,
Expand All @@ -515,35 +515,35 @@ def predict_from_random_spectra(

# self.obs_data_df['Fnu_obs'] = self.Fnu_obs
# self.obs_data_df['Fnu_obs_err'] = self.Fnu_obs_err
# self.obs_data_df['Fnu_obs_absolute'] = self.Fnu_obs_absolute
# self.obs_data_df['Fnu_obs_absolute_err'] = self.Fnu_obs_absolute_err
# self.obs_data_df['Fnu_obs_TOA'] = self.Fnu_obs_TOA
# self.obs_data_df['Fnu_obs_TOA_err'] = self.Fnu_obs_TOA_err

# Extract the original ML features from the observational spectrum
Fnu_obs_absolute_intd, Fnu_obs_absolute_intd_df = \
self.flux_interpolated(Fnu_obs_absolute = Fnu_obs_absolute,
Fnu_obs_TOA_intd, Fnu_obs_TOA_intd_df = \
self.flux_interpolated(Fnu_obs_TOA = Fnu_obs_TOA,
interpolated_wl = None,
__print_results__=False,
__plot_spectra_errorbar__=False,
__use_spectres__=True)

# Extract the engineered ML features from the observational spectrum
Fnu_obs_absolute_intd_df_min = Fnu_obs_absolute_intd_df.min(axis=1)
Fnu_obs_absolute_intd_df_max = Fnu_obs_absolute_intd_df.max(axis=1)
Fnu_obs_TOA_intd_df_min = Fnu_obs_TOA_intd_df.min(axis=1)
Fnu_obs_TOA_intd_df_max = Fnu_obs_TOA_intd_df.max(axis=1)

df_MinMax_obs = pd.DataFrame(
(Fnu_obs_absolute_intd_df_min, Fnu_obs_absolute_intd_df_max)
(Fnu_obs_TOA_intd_df_min, Fnu_obs_TOA_intd_df_max)
).T

XminXmax_Stand = self.trained_X_ColWise_MinMax[0].transform(df_MinMax_obs.values)
# XminXmax_Stand = self.trained_data_processor.normalize_X_ColumnWise.transform(df_MinMax_obs.values)


bd_mean = Fnu_obs_absolute_intd_df.mean(axis=1)[0]
bd_std = Fnu_obs_absolute_intd_df.std(axis=1)[0]
bd_mean = Fnu_obs_TOA_intd_df.mean(axis=1)[0]
bd_std = Fnu_obs_TOA_intd_df.std(axis=1)[0]


# X_Scaled = (Fnu_obs_absolute_intd_df.div((self.bd_literature_dic['bd_radius_Rjup'])**2).values[0] - bd_mean) / bd_std
X_Scaled = (Fnu_obs_absolute_intd_df.values[0] - bd_mean) / bd_std
# X_Scaled = (Fnu_obs_TOA_intd_df.div((self.bd_literature_dic['bd_radius_Rjup'])**2).values[0] - bd_mean) / bd_std
X_Scaled = (Fnu_obs_TOA_intd_df.values[0] - bd_mean) / bd_std

y_pred_train = np.array(
self.trained_ML_model.predict(
Expand All @@ -559,7 +559,7 @@ def predict_from_random_spectra(
)


spectra_list_obs.append(Fnu_obs_absolute_intd_df.values)
spectra_list_obs.append(Fnu_obs_TOA_intd_df.values)
param_list.append(y_pred_random[0])

filtered_df4 = interpolate_df(dataset=self.training_dataset_df,
Expand Down Expand Up @@ -600,7 +600,7 @@ def predict_from_random_spectra(
p = figure(
title=self.object_name + ": Randomly generated spectra within 1σ",
x_axis_label='Wavelength [𝜇m]',
y_axis_label='Flux (F𝜈) [erg/s/cm2/Hz]',
y_axis_label='TOA Flux (F𝜈) [erg/s/cm2/Hz]',
width=800,
height=300,
y_axis_type="log",
Expand Down Expand Up @@ -647,8 +647,8 @@ def predict_from_random_spectra(
plot_pred_vs_obs_errorbar(
object_name=self.object_name,
x_obs=self.obs_data_df['wl'],
y_obs=Fnu_obs_absolute,
y_obs_error=Fnu_obs_absolute_err,
y_obs=Fnu_obs_TOA,
y_obs_error=Fnu_obs_TOA_err,
training_dataset=self.training_dataset_df,
x_pred=self.wl_synthetic,
predicted_targets_dic=self.dic_random_pred_mean,
Expand All @@ -668,11 +668,11 @@ def predict_from_random_spectra(
confidence_level=0.95,
object_name=self.object_name,
x_obs=self.obs_data_df['wl'],
y_obs=self.obs_data_df['Fnu_obs_absolute'], #self.obs_data_df['Fnu_obs_absolute'],# self.Fnu_obs_absolute,
y_obs_err=self.obs_data_df['Fnu_obs_absolute_err'], #self.obs_data_df['Fnu_obs_absolute_err'],#self.Fnu_obs_absolute_err,
y_obs=self.obs_data_df['Fnu_obs_TOA'],
y_obs_err=self.obs_data_df['Fnu_obs_TOA_err'],
training_datasets=self.training_dataset_df,
x_pred=self.wl_synthetic,
predicted_targets_dic=self.dic_random_pred_mean, # self.dic_random_pred_mean,
predicted_targets_dic=self.dic_random_pred_mean,
radius = self.bd_literature_dic['bd_radius_Rjup'],
__print_results__=False,
)
Expand All @@ -682,8 +682,8 @@ def predict_from_random_spectra(
confidence_level=0.95,
object_name=self.object_name,
x_obs=self.obs_data_df['wl'],
y_obs=self.obs_data_df['Fnu_obs_absolute'],
y_obs_err=self.obs_data_df['Fnu_obs_absolute_err'],
y_obs=self.obs_data_df['Fnu_obs_TOA'],
y_obs_err=self.obs_data_df['Fnu_obs_TOA_err'],
training_datasets=self.training_dataset_df,
x_pred=self.wl_synthetic,
predicted_targets_dic=self.dic_random_pred_mean,
Expand Down

0 comments on commit ba9abef

Please sign in to comment.