Merge pull request #31 from antoinedemathelin/master

fix: Update transfer tree + Fix clone for parameter based methods

Author: antoinedemathelin

…meter_based/decision_trees/tree_utils.py adapt/_tree_utils.pyadapt/parameter_based/decision_trees/tree_utils.py renamed to adapt/_tree_utils.py adapt/parameter_based/decision_trees/tree_utils.py renamed to adapt/_tree_utils.py

@@ -464,7 +464,8 @@ def coherent_new_split(phi,th,rule):
             return 0,1
         else:
             return 1,0
-      
+
+
 def all_coherent_splits(rule,all_splits):
 
     inds = np.zeros(all_splits.shape[0],dtype=bool)

adapt/metrics.py

@@ -413,7 +413,7 @@ def normalized_frechet_distance(Xs, Xt):
     return x_max / Xs.shape[1]
 
 
-def j_score(Xs, Xt, max_centers=100, sigma=None):
+def neg_j_score(Xs, Xt, max_centers=100, sigma=None):
     """
     Compute the negative J-score between Xs and Xt.
     

adapt/parameter_based/__init__.py

@@ -4,5 +4,10 @@
 
 from ._regular import RegularTransferLR, RegularTransferLC, RegularTransferNN
 from ._finetuning import FineTuning
+from ._transfer_tree import TransferTreeClassifier
 
-__all__ = ["RegularTransferLR", "RegularTransferLC", "RegularTransferNN", "FineTuning"]
+__all__ = ["RegularTransferLR",
+           "RegularTransferLC",
+           "RegularTransferNN",
+           "FineTuning",
+           "TransferTreeClassifier"]

adapt/parameter_based/_finetuning.py

@@ -1,5 +1,6 @@
 import tensorflow as tf
 from adapt.base import BaseAdaptDeep, make_insert_doc
+from adapt.utils import check_fitted_network
 
 
 @make_insert_doc(["encoder", "task"], supervised=True)
@@ -67,6 +68,9 @@ def __init__(self,
                  random_state=None,
                  **params):
 
+        encoder = check_fitted_network(encoder)
+        task = check_fitted_network(task)
+        
         names = self._get_param_names()
         kwargs = {k: v for k, v in locals().items() if k in names}
         kwargs.update(params)

adapt/parameter_based/_regular.py

@@ -3,7 +3,6 @@
 """
 
 import numpy as np
-from sklearn.exceptions import NotFittedError
 from sklearn.preprocessing import LabelBinarizer
 from scipy.sparse.linalg import lsqr
 import tensorflow as tf
@@ -14,7 +13,8 @@
 from adapt.utils import (check_arrays,
                          set_random_seed,
                          check_estimator,
-                         check_network)
+                         check_network,
+                         check_fitted_estimator)
 
 
 @make_insert_doc(supervised=True)
@@ -101,9 +101,12 @@ def __init__(self,
                  **params):
 
         if not hasattr(estimator, "coef_"):
-            raise NotFittedError("`estimator` argument has no ``coef_`` attribute, "
-                                 "please call `fit` on `estimator` or use "
-                                 "another estimator.")
+            raise ValueError("`estimator` argument has no ``coef_`` attribute, "
+                             "please call `fit` on `estimator` or use "
+                             "another estimator as `LinearRegression` or "
+                             "`RidgeClassifier`.")
+            
+        estimator = check_fitted_estimator(estimator)
 
         names = self._get_param_names()
         kwargs = {k: v for k, v in locals().items() if k in names}
@@ -137,7 +140,7 @@ def fit(self, Xt=None, yt=None, **fit_params):
         self.estimator_ = check_estimator(self.estimator,
                                           copy=self.copy,
                                           force_copy=True)
-                
+
         if self.estimator_.fit_intercept:
             intercept_ = np.reshape(
                 self.estimator_.intercept_,