Draft: New dockerfile

notebook/compressed_sensing.ipynb

 ``` python
+import contextlib
+import numpy as np
 import os
 import pandas as pd
-import numpy as np
+import shutil
+
 from itertools import combinations, product
 from time import time
 import matplotlib.pyplot as plt
 from IPython.display import HTML
 from jupyter_jsmol import JsmolView
 import pathlib
+from pathlib import Path
 
 from compressed_sensing.sisso import SissoRegressor
 from compressed_sensing.combine_features import combine_features
 from sissopp.py_interface.import_dataframe import get_unit
 from sissopp.sklearn import SISSORegressor as SISSORegressor_skl
 
+from pysr import PySRRegressor
+from sympy import Abs, root, sign
+
 warnings.filterwarnings('ignore')
 from bokeh.io import export_png
+%% Cell type:code id: tags:
+``` python
+@contextlib.contextmanager
+def cwd(path, mkdir=False, debug=False, clear=True):
+    CWD = os.getcwd()
+
+    if os.path.exists(path) is False and mkdir:
+        os.makedirs(path)
+
+    if debug:
+        os.chdir(path)
+        yield
+        os.chdir(CWD)
+        return
+
+    os.chdir(path)
+    yield
+    os.chdir(CWD)
+    if clear:
+        shutil.rmtree(path)
 %% Cell type:markdown id: tags:
 feat_space = FeatureSpace(inputs)
 
 sisso = SISSORegressor(inputs, feat_space)
+with cwd(Path("SISSO_training"), mkdir=True):
+    sisso.fit()
 
-sisso.fit()
 for i in range(n_nonzero_coefs):
     print(str(i+1)+'D model')
     print("RMSE: {:.4} | Descriptor: {}".format(sisso.models[i][0].rmse, sisso.models[i][0]))
 
 sisso = SISSORegressor(inputs, feat_space)
 
-sisso.fit()
+with cwd(Path("SISSO_training"), mkdir=True):
+    sisso.fit()
+
 for i in range(n_nonzero_coefs):
     print(str(i+1)+'D model')
     print("RMSE: {:.4} | Descriptor: {}".format(sisso.models[i][0].rmse, sisso.models[i][0]))
 
 df_sisso_cv = df_plus_reduced.drop(columns=["energy_diff"])
 
-for indices_train, index_test in loo.split(P):
-    i_cv = index_test[0]
-    print('%2s) Leave out %s: Ediff_ref = %.3f eV/atom'
-          % (index_test[0]+1, chemical_formulas[i_cv], P[i_cv]))
-
-    sisso.fit(df_sisso_cv.iloc[indices_train, :], P[indices_train])
-    data_dct = {col.split("(")[0].strip(): df_sisso_cv.iloc[i_cv, cc] for cc, col in enumerate(df_sisso_cv.columns)}
-    predicted_value = [models[0].eval(data_dct) for models in sisso.solver.models[:]]
-    P_predict[:, i_cv] = predicted_value
-    for dim in dimensions:
-        print('Ediff_predicted(%sD) = %.3f eV/atom' %(dim, predicted_value[dim - 1]))
-    print('-----')
+with cwd(Path("SISSO_CV"), mkdir=True):
+    for indices_train, index_test in loo.split(P):
+        i_cv = index_test[0]
+        print('%2s) Leave out %s: Ediff_ref = %.3f eV/atom'
+              % (index_test[0]+1, chemical_formulas[i_cv], P[i_cv]))
+
+        sisso.fit(df_sisso_cv.iloc[indices_train, :], P[indices_train])
+        data_dct = {col.split("(")[0].strip(): df_sisso_cv.iloc[i_cv, cc] for cc, col in enumerate(df_sisso_cv.columns)}
+        predicted_value = [models[0].eval(data_dct) for models in sisso.solver.models[:]]
+        P_predict[:, i_cv] = predicted_value
+        for dim in dimensions:
+            print('Ediff_predicted(%sD) = %.3f eV/atom' %(dim, predicted_value[dim - 1]))
+        print('-----')
 %% Cell type:code id: tags:
 ``` python
-import contextlib
-import os
-from pathlib import Path
-
-from pysr import PySRRegressor
-from sympy import Abs, root, sign
-
-@contextlib.contextmanager
-def cwd(path, mkdir=False, debug=False):
-    CWD = os.getcwd()
-
-    if os.path.exists(path) is False and mkdir:
-        os.makedirs(path)
-
-    if debug:
-        os.chdir(path)
-        yield
-        os.chdir(CWD)
-        return
-
-    os.chdir(path)
-    yield
-    os.chdir(CWD)
-
 pysr = PySRRegressor(
     niterations=100,
     binary_operators=["+", "-", "/"],
 
 P_predict_pysr = []
 loo = LeaveOneOut()
-for indices_train, index_test in loo.split(P):
-    pysr_path = f"pySR/{index_test[0]}"
-    with cwd(pysr_path, mkdir=True):
+with cwd(Path("pySR"), mkdir=True):
+    for indices_train, index_test in loo.split(P):
         pysr.fit(pysr_df.iloc[indices_train, :], P[indices_train])
         P_predict_pysr.append(pysr.predict(pysr_df.iloc[index_test, :])[0])
         print("%2i Ediff_ref: %.3f, Ediff_pred: %.3f, equation: {%s}"
 ``` python
-pysr_path = f"pySR/all_pts"
+pysr_path = Path(f"pySR/all_pts")
 with cwd(pysr_path, mkdir=True):
     pysr.fit(pysr_df, P)
     print(pysr.get_best().sympy_format)
 plt.ylabel("Count")
 plt.show()
+%% Cell type:code id: tags:
+``` python