data-manipulation-utilities 0.2.1__py3-none-any.whl → 0.2.3__py3-none-any.whl

{data_manipulation_utilities-0.2.1.dist-info → data_manipulation_utilities-0.2.3.dist-info}/METADATA

@@ -1,11 +1,11 @@
 Metadata-Version: 2.2
 Name: data_manipulation_utilities
-Version: 0.2.1
+Version: 0.2.3
 Description-Content-Type: text/markdown
 Requires-Dist: logzero
 Requires-Dist: PyYAML
 Requires-Dist: scipy
-Requires-Dist: awkward==2.4.6
+Requires-Dist: awkward
 Requires-Dist: tqdm
 Requires-Dist: joblib
 Requires-Dist: scikit-learn
@@ -424,6 +424,10 @@ where the settings for the training go in a config dictionary, which when writte
 
 ```yaml
 dataset:
+    # Before training, new features can be defined as below
+    define :
+        x : v + w
+        y : v - w
     # If the key is found to be NaN, replace its value with the number provided
     # This will be used in the training.
     # Otherwise the entries with NaNs will be dropped
@@ -433,7 +437,7 @@ dataset:
         z : -999 
 training :
     nfold    : 10
-    features : [w, x, y, z]
+    features : [x, y, z]
     hyper    :
       loss              : log_loss
       n_estimators      : 100
@@ -493,7 +497,9 @@ When training on real data, several things might go wrong and the code will try
 will end up in different folds. The tool checks for wether a model is evaluated for an entry that was used for training and raise an exception. Thus, repeated
 entries will be removed before training.
 
-- **NaNs**: Entries with NaNs will break the training with the scikit GradientBoostClassifier base class. Thus, we also remove them from the training.
+- **NaNs**: Entries with NaNs will break the training with the scikit `GradientBoostClassifier` base class. Thus, we: 
+    - Can use the `nan` section shown above to replace `NaN` values with something else
+    - For whatever remains we remove the entries from the training.
 
 ## Application
 
@@ -516,15 +522,24 @@ The picking process happens through the comparison of hashes between the samples
 The hashes of the training samples are stored in the pickled model itself; which therefore is a reimplementation of
 `GradientBoostClassifier`, here called `CVClassifier`.
 
-If a sample exist, that was used in the training of _every_ model, no model can be chosen for the prediction and an
+If a sample exists, that was used in the training of _every_ model, no model can be chosen for the prediction and a
 `CVSameData` exception will be risen.
 
+During training, the configuration will be stored in the model. Therefore, variable definitions can be picked up for evaluation
+from that configuration and the user does not need to define extra columns.
+
 ### Caveats
 
 When evaluating the model with real data, problems might occur, we deal with them as follows:
 
 - **Repeated entries**: When there are repeated features in the dataset to be evaluated we assign the same probabilities, no filtering is used.
-- **NaNs**: Entries with NaNs will break the evaluation. These entries will be _patched_  with zeros and evaluated. However, before returning, the probabilities will be
+- **NaNs**: Entries with NaNs will break the evaluation. These entries will be:
+    - Replaced by other values before evaluation IF a replacement was specified during training. The training configuration will be stored in the model
+    and can be accessed through:
+    ```python
+    model.cfg
+    ```
+    - For whatever entries that are still NaN, they will be _patched_  with zeros and evaluated. However, before returning, the probabilities will be
 saved as -1. I.e. entries with NaNs will have probabilities of -1.
 
 # Pandas dataframes
@@ -563,6 +578,24 @@ These are utility functions meant to be used with ROOT dataframes.
 
 ## Adding a column from a numpy array
 
+### With numba
+
+For this do:
+
+```python
+import dmu.rdataframe.utilities as ut
+
+arr_val = numpy.array([10, 20, 30])
+rdf     = ut.add_column_with_numba(rdf, arr_val, 'values', identifier='some_name')
+```
+
+where the identifier needs to be unique, every time the function is called.
+This is the case, because the addition is done internally by declaring a numba function whose name
+cannot be repeated as mentioned
+[here](https://root-forum.cern.ch/t/ways-to-work-around-the-redefinition-of-compiled-functions-in-one-single-notebook-session/41442/1)
+
+### With awkward
+
 For this do:
 
 ```python

{data_manipulation_utilities-0.2.1.dist-info → data_manipulation_utilities-0.2.3.dist-info}/RECORD

@@ -1,10 +1,10 @@
-data_manipulation_utilities-0.2.1.data/scripts/publish,sha256=-3K_Y2_4CfWCV50rPB8CRuhjxDu7xMGswinRwPovgLs,1976
+data_manipulation_utilities-0.2.3.data/scripts/publish,sha256=-3K_Y2_4CfWCV50rPB8CRuhjxDu7xMGswinRwPovgLs,1976
 dmu/arrays/utilities.py,sha256=PKoYyybPptA2aU-V3KLnJXBudWxTXu4x1uGdIMQ49HY,1722
 dmu/generic/utilities.py,sha256=0Xnq9t35wuebAqKxbyAiMk1ISB7IcXK4cFH25MT1fgw,1741
 dmu/logging/log_store.py,sha256=umdvjNDuV3LdezbG26b0AiyTglbvkxST19CQu9QATbA,4184
 dmu/ml/cv_classifier.py,sha256=8Jwx6xMhJaRLktlRdq0tFl32v6t8i63KmpxrlnXlomU,3759
-dmu/ml/cv_predict.py,sha256=AhCsCnHWPWGIRVTdGS1NxA2m4yH7t2lV_OdALwQAcAE,4927
-dmu/ml/train_mva.py,sha256=xJCJZKaly4Mml7Dy-TWQxpB-VNftL7EjQ79QKxROWx0,16475
+dmu/ml/cv_predict.py,sha256=4G7F_1yOvnLftsDC6zUpdvkxuHXGkPemhj0RsYySYDM,6708
+dmu/ml/train_mva.py,sha256=SZ5cQHl7HBxn0c5Hh4HlN1aqMZaJUAlNmsfjnUSQrTg,16894
 dmu/ml/utilities.py,sha256=l348bufD95CuSYdIrHScQThIy2nKwGKXZn-FQg3CEwg,3930
 dmu/pdataframe/utilities.py,sha256=ypvLiFfJ82ga94qlW3t5dXnvEFwYOXnbtJb2zHwsbqk,987
 dmu/plotting/matrix.py,sha256=pXuUJn-LgOvrI9qGkZQw16BzLjOjeikYQ_ll2VIcIXU,4978
@@ -13,7 +13,7 @@ dmu/plotting/plotter_1d.py,sha256=g6H2xAgsL9a6vRkpbqHICb3qwV_qMiQPZxxw_oOSf9M,51
 dmu/plotting/plotter_2d.py,sha256=J-gKnagoHGfJFU7HBrhDFpGYH5Rxy0_zF5l8eE_7ZHE,2944
 dmu/plotting/utilities.py,sha256=SI9dvtZq2gr-PXVz71KE4o0i09rZOKgqJKD1jzf6KXk,1167
 dmu/rdataframe/atr_mgr.py,sha256=FdhaQWVpsm4OOe1IRbm7rfrq8VenTNdORyI-lZ2Bs1M,2386
-dmu/rdataframe/utilities.py,sha256=x8r379F2-vZPYzAdMFCn_V4Kx2Tx9t9pn_QHcZ1euew,2756
+dmu/rdataframe/utilities.py,sha256=pNcQARMP7txMhy6k27UnDcYf0buNy5U2fshaJDl_h8o,3661
 dmu/rfile/rfprinter.py,sha256=mp5jd-oCJAnuokbdmGyL9i6tK2lY72jEfROuBIZ_ums,3941
 dmu/rfile/utilities.py,sha256=XuYY7HuSBj46iSu3c60UYBHtI6KIPoJU_oofuhb-be0,945
 dmu/stats/fitter.py,sha256=vHNZ16U3apoQyeyM8evq-if49doF48sKB3q9wmA96Fw,18387
@@ -47,8 +47,8 @@ dmu_scripts/rfile/compare_root_files.py,sha256=T8lDnQxsRNMr37x1Y7YvWD8ySHrJOWZki
 dmu_scripts/rfile/print_trees.py,sha256=Ze4Ccl_iUldl4eVEDVnYBoe4amqBT1fSBR1zN5WSztk,941
 dmu_scripts/ssh/coned.py,sha256=lhilYNHWRCGxC-jtyJ3LQ4oUgWW33B2l1tYCcyHHsR0,4858
 dmu_scripts/text/transform_text.py,sha256=9akj1LB0HAyopOvkLjNOJiptZw5XoOQLe17SlcrGMD0,1456
-data_manipulation_utilities-0.2.1.dist-info/METADATA,sha256=ojD6P0bBj9GFohtPd7ULl7sDW80bIVD6JZ-bnNpHYmc,26649
-data_manipulation_utilities-0.2.1.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-data_manipulation_utilities-0.2.1.dist-info/entry_points.txt,sha256=1TIZDed651KuOH-DgaN5AoBdirKmrKE_oM1b6b7zTUU,270
-data_manipulation_utilities-0.2.1.dist-info/top_level.txt,sha256=n_x5J6uWtSqy9mRImKtdA2V2NJNyU8Kn3u8DTOKJix0,25
-data_manipulation_utilities-0.2.1.dist-info/RECORD,,
+data_manipulation_utilities-0.2.3.dist-info/METADATA,sha256=STJ7vYfcSIM9dtMRzywGLwDzH1sUBE5DL9FqvskMcxo,27923
+data_manipulation_utilities-0.2.3.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
+data_manipulation_utilities-0.2.3.dist-info/entry_points.txt,sha256=1TIZDed651KuOH-DgaN5AoBdirKmrKE_oM1b6b7zTUU,270
+data_manipulation_utilities-0.2.3.dist-info/top_level.txt,sha256=n_x5J6uWtSqy9mRImKtdA2V2NJNyU8Kn3u8DTOKJix0,25
+data_manipulation_utilities-0.2.3.dist-info/RECORD,,

dmu/ml/cv_predict.py

@@ -32,11 +32,56 @@ class CVPredict:
         if rdf is None:
             raise ValueError('No ROOT dataframe passed')
 
-        self._l_model = models
-        self._rdf     = rdf
+        self._l_model   = models
+        self._rdf       = rdf
+        self._d_nan_rep : dict[str,str]
 
         self._arr_patch : numpy.ndarray
     # --------------------------------------------
+    def _initialize(self):
+        self._rdf       = self._define_columns(self._rdf)
+        self._d_nan_rep = self._get_nan_replacements()
+    # --------------------------------------------
+    def _define_columns(self, rdf : RDataFrame) -> RDataFrame:
+        cfg = self._l_model[0].cfg
+
+        if 'define' not in cfg['dataset']:
+            log.debug('No define section found in config, will not define extra columns')
+            return self._rdf
+
+        d_def = cfg['dataset']['define']
+        log.debug(60 * '-')
+        log.info('Defining columns in RDF before evaluating classifier')
+        log.debug(60 * '-')
+        for name, expr in d_def.items():
+            log.debug(f'{name:<20}{"<---":20}{expr:<100}')
+            rdf = rdf.Define(name, expr)
+
+        return rdf
+    # --------------------------------------------
+    def _get_nan_replacements(self) -> dict[str,str]:
+        cfg = self._l_model[0].cfg
+
+        if 'nan' not in cfg['dataset']:
+            log.debug('No define section found in config, will not define extra columns')
+            return {}
+
+        return cfg['dataset']['nan']
+    # --------------------------------------------
+    def _replace_nans(self, df : pnd.DataFrame) -> pnd.DataFrame:
+        if len(self._d_nan_rep) == 0:
+            log.debug('Not doing any NaN replacement')
+            return df
+
+        log.debug(60 * '-')
+        log.info('Doing NaN replacements')
+        log.debug(60 * '-')
+        for var, val in self._d_nan_rep.items():
+            log.debug(f'{var:<20}{"--->":20}{val:<20.3f}')
+            df[var] = df[var].fillna(val)
+
+        return df
+    # --------------------------------------------
     def _get_df(self):
         '''
         Will make ROOT rdf into dataframe and return it
@@ -45,6 +90,7 @@ class CVPredict:
         l_ft  = model.features
         d_data= self._rdf.AsNumpy(l_ft)
         df_ft = pnd.DataFrame(d_data)
+        df_ft = self._replace_nans(df_ft)
         df_ft = ut.patch_and_tag(df_ft)
 
         if 'patched_indices' in df_ft.attrs:
@@ -136,6 +182,8 @@ class CVPredict:
         '''
         Will return array of prediction probabilities for the signal category
         '''
+        self._initialize()
+
         df_ft = self._get_df()
         model = self._l_model[0]
 

dmu/ml/train_mva.py

@@ -26,7 +26,7 @@ from dmu.plotting.matrix     import MatrixPlotter
 from dmu.logging.log_store   import LogStore
 
 npa = numpy.ndarray
-log = LogStore.add_logger('data_checks:train_mva')
+log = LogStore.add_logger('dmu:ml:train_mva')
 # ---------------------------------------------
 class TrainMva:
     '''
@@ -334,10 +334,10 @@ class TrainMva:
         if 'max' in self._cfg['plotting']['roc']:
             [max_x, max_y] = self._cfg['plotting']['roc']['max']
 
-        self._plot_probabilities(xval_ts, yval_ts, l_prb_ts)
-
         plt.plot(xval_ts, yval_ts, color='b', label=f'Test: {area_ts:.3f}')
         plt.plot(xval_tr, yval_tr, color='r', label=f'Train: {area_tr:.3f}')
+        self._plot_probabilities(xval_ts, yval_ts, l_prb_ts, l_lab_ts)
+
         plt.xlabel('Signal efficiency')
         plt.ylabel('Background rejection')
         plt.title(f'Fold: {ifold}')
@@ -351,13 +351,17 @@ class TrainMva:
     def _plot_probabilities(self,
                             arr_seff: npa,
                             arr_brej: npa,
-                            arr_sprb: npa) -> None:
+                            arr_sprb: npa,
+                            arr_labl: npa) -> None:
 
         roc_cfg = self._cfg['plotting']['roc']
         if 'annotate' not in roc_cfg:
             log.debug('Annotation section in the ROC curve config not found, skipping annotation')
             return
 
+        l_sprb   = [ sprb for sprb, labl in zip(arr_sprb, arr_labl) if labl == 1 ]
+        arr_sprb = numpy.array(l_sprb)
+
         plt_cfg = roc_cfg['annotate']
         if 'sig_eff' not in plt_cfg:
             l_seff_target = [0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.95]
@@ -366,17 +370,24 @@ class TrainMva:
             del plt_cfg['sig_eff']
 
         arr_seff_target = numpy.array(l_seff_target)
+        arr_quantile    = 1 - arr_seff_target
 
-        l_score = numpy.quantile(arr_sprb, 1 - arr_seff_target)
+        l_score = numpy.quantile(arr_sprb, arr_quantile)
         l_seff  = []
         l_brej  = []
-        for seff_target in l_seff_target:
+
+        log.debug(60 * '-')
+        log.debug(f'{"SigEff":20}{"BkgRej":20}{"Score":20}')
+        log.debug(60 * '-')
+        for seff_target, score in zip(arr_seff_target, l_score):
             arr_diff = numpy.abs(arr_seff - seff_target)
             ind      = numpy.argmin(arr_diff)
 
             seff     = arr_seff[ind]
             brej     = arr_brej[ind]
 
+            log.debug(f'{seff:<20.3f}{brej:<20.3f}{score:<20.2f}')
+
             l_seff.append(seff)
             l_brej.append(brej)
 

dmu/rdataframe/utilities.py

@@ -1,6 +1,7 @@
 '''
 Module containing utility functions to be used with ROOT dataframes
 '''
+# pylint: disable=no-name-in-module
 
 import re
 from dataclasses import dataclass
@@ -10,7 +11,7 @@ import pandas  as pnd
 import awkward as ak
 import numpy
 
-from ROOT import RDataFrame, RDF
+from ROOT import RDataFrame, RDF, Numba
 
 from dmu.logging.log_store import LogStore
 
@@ -34,6 +35,8 @@ def add_column(rdf : RDataFrame, arr_val : Union[numpy.ndarray,None], name : str
          exclude_re : Regex with patter of column names that we won't pick
     '''
 
+    log.warning(f'Adding column {name} with awkward')
+
     d_opt = {} if d_opt is None else d_opt
     if arr_val is None:
         raise ValueError('Array of values not introduced')
@@ -66,12 +69,35 @@ def add_column(rdf : RDataFrame, arr_val : Union[numpy.ndarray,None], name : str
     if arr_val.dtype == 'object':
         arr_val = arr_val.astype(float)
 
-    d_data[name] = arr_val
+    d_data[name] = ak.from_numpy(arr_val)
 
     rdf = ak.to_rdataframe(d_data)
 
     return rdf
 # ---------------------------------------------------------------------
+def add_column_with_numba(
+        rdf        : RDataFrame,
+        arr_val    : Union[numpy.ndarray,None],
+        name       : str,
+        identifier : str) -> RDataFrame:
+    '''
+    Will take a dataframe, an array of numbers and a string
+    Will add the array as a colunm to the dataframe
+
+    The `identifier` argument is a string need in order to avoid collisions
+    when using Numba to define a function to get the value from.
+    '''
+    identifier=f'fun_{identifier}'
+
+    @Numba.Declare(['int'], 'float', name=identifier)
+    def get_value(index):
+        return arr_val[index]
+
+    log.debug(f'Adding column {name} with numba')
+    rdf = rdf.Define(name, f'Numba::{identifier}(rdfentry_)')
+
+    return rdf
+# ---------------------------------------------------------------------
 def rdf_report_to_df(rep : RDF.RCutFlowReport) -> pnd.DataFrame:
     '''
     Takes the output of rdf.Report(), i.e. an RDataFrame cutflow report.