Clueanup 1

w4k2 · Oct 24, 2023 · 267a983 · 267a983
1 parent 11736ca
commit 267a983
Show file tree

Hide file tree

Showing 58 changed files with 32 additions and 81 deletions.
diff --git a/vis_real.py → E1P_vis_real.py b/vis_real.py → E1P_vis_real.py
diff --git a/vis_semi.py → E1P_vis_semi.py b/vis_semi.py → E1P_vis_semi.py
diff --git a/vis_synthetic.py → E1P_vis_synthetic.py b/vis_synthetic.py → E1P_vis_synthetic.py
diff --git a/extract_real.py → E1_extract_real.py b/extract_real.py → E1_extract_real.py
@@ -1,5 +1,5 @@
 """
-Przygotowanie strumieni i meta - rzeczywiste
+Experiment 1 -- collect streams and metafetaures -- real-world
 """
 
 import numpy as np
@@ -38,7 +38,7 @@
         fname=(f.split('/')[1]).split('.')[0]
 
         drfs = np.load('real_streams_gt/%s.npy' % fname)
-        
+
         concept=0
         out = []
 
@@ -48,7 +48,6 @@
             # GET CONCEPT
             if chunk in drfs:
                 concept+=1
-                # print(concept)
 
             # CALCULATE
             try:
@@ -58,17 +57,13 @@
                 break
 
             if len(np.unique(y))<2:
-                # print('skip', chunk)
                 continue
-                # exit()
 
             mfe = MFE(groups=[measure_key])
             mfe.fit(X,y)
             ft_labels, ft = mfe.extract()
             ft.append(concept)
 
             out.append(ft)
-        # print(np.array(out).shape)
-        # exit()
 
         np.save('res/real_%i_%s.npy' % (f_id, measure_key), np.array(out))
diff --git a/extract_semi.py → E1_extract_semi.py b/extract_semi.py → E1_extract_semi.py
@@ -1,5 +1,5 @@
 """
-Przygotowanie strumieni i meta - semi
+Experiment 1 -- collect streams and metafetaures -- semi-synthetic
 """
 
 import numpy as np

diff --git a/extract_synthetic.py → E1_extract_synthetic.py b/extract_synthetic.py → E1_extract_synthetic.py
@@ -1,5 +1,5 @@
 """
-Przygotowanie strumieni i meta - synthetic
+Experiment 1 -- collect streams and metafetaures -- synthetic
 """
 
 import numpy as np

diff --git a/analyze_real.py → E2P_analyze_real.py b/analyze_real.py → E2P_analyze_real.py
diff --git a/analyze_semi.py → E2P_analyze_semi.py b/analyze_semi.py → E2P_analyze_semi.py
diff --git a/analyze_synthetic.py → E2P_analyze_synthetic.py b/analyze_synthetic.py → E2P_analyze_synthetic.py
diff --git a/clf_real.py → E2_clf_real.py b/clf_real.py → E2_clf_real.py
diff --git a/clf_semi.py → E2_clf_semi.py b/clf_semi.py → E2_clf_semi.py
diff --git a/clf_synthetic.py → E2_clf_synthetic.py b/clf_synthetic.py → E2_clf_synthetic.py
diff --git a/vis_selection_real.py → E3P_vis_selection_real.py b/vis_selection_real.py → E3P_vis_selection_real.py
diff --git a/vis_selection_semi.py → E3P_vis_selection_semi.py b/vis_selection_semi.py → E3P_vis_selection_semi.py
diff --git a/vis_selection_synthetic.py → E3P_vis_selection_synthetic.py b/vis_selection_synthetic.py → E3P_vis_selection_synthetic.py
@@ -34,10 +34,7 @@ def sqspace(start, end, num):
 for d_id, drift_type in enumerate(['Sudden', 'Gradual', 'Incremental']):    
     clf_temp = clf[d_id]
     clf_temp_mean = np.mean(clf[d_id], axis=(0,2))
-
-    #print('A', clf.shape)
-    #exit()
-
+
     for cm_id, cm in enumerate(clf_temp_mean.T):
         ax[d_id].plot(n_features, cm, label=base_clfs[cm_id], c=c[cm_id])
     ax[d_id].set_title('%s drifts' % drift_type)    

diff --git a/clf_real_selection.py → E3_clf_real_selection.py b/clf_real_selection.py → E3_clf_real_selection.py
diff --git a/clf_semi_selection.py → E3_clf_semi_selection.py b/clf_semi_selection.py → E3_clf_semi_selection.py
diff --git a/clf_synthetic_selection.py → E3_clf_synthetic_selection.py b/clf_synthetic_selection.py → E3_clf_synthetic_selection.py
diff --git a/clf_real_v2.py → E4_clf_real_v2.py b/clf_real_v2.py → E4_clf_real_v2.py
diff --git a/clf_semi_v2.py → E4_clf_semi_v2.py b/clf_semi_v2.py → E4_clf_semi_v2.py
diff --git a/clf_synthetic_v2.py → E4_clf_synthetic_v2.py b/clf_synthetic_v2.py → E4_clf_synthetic_v2.py
diff --git a/cov_real.py → E5_cov_real.py b/cov_real.py → E5_cov_real.py
diff --git a/cov_semi.py → E5_cov_semi.py b/cov_semi.py → E5_cov_semi.py
diff --git a/cov_syn.py → E5_cov_syn.py b/cov_syn.py → E5_cov_syn.py
diff --git a/__pycache__/utils.cpython-310.pyc b/__pycache__/utils.cpython-310.pyc
diff --git a/bar.png b/bar.png
diff --git a/baz.png b/baz.png
diff --git a/logo-dataset.png → fig/logo-dataset.png b/logo-dataset.png → fig/logo-dataset.png
diff --git a/figure-one.py b/figure-one.py
@@ -1,3 +1,7 @@
+'''
+Script generating sample figure for manuscript
+
+'''
 import matplotlib.pyplot as plt
 import numpy as np
 from strlearn.streams import StreamGenerator

diff --git a/metric_combine.py b/metric_combine.py
@@ -1,5 +1,5 @@
 """
-Sklejenie meta z obiecujących podgrup
+Combina metagratures from promising froups
 """
 import numpy as np
 

diff --git a/olaboga.png b/olaboga.png
diff --git a/olaboga2.png b/olaboga2.png
diff --git a/real_gt.py b/real_gt.py
@@ -1,15 +1,14 @@
 """
-Oznaczanie momentów dryfu w rzeczywistych
+Script for marking moments of drift for real-world datastreams.
 """
 import numpy as np
 from sklearn.naive_bayes import GaussianNB
 from sklearn.neural_network import MLPClassifier
 from tqdm import tqdm
 import strlearn as sl
 import matplotlib.pyplot as plt
-from sklearn.tree import DecisionTreeClassifier
+from sklearn.svm import SVC
 
-from utils import ELMI
 real_streams = [
     'real_streams/covtypeNorm-1-2vsAll-pruned.arff',
     'real_streams/electricity.npy',
@@ -21,10 +20,8 @@
 
 stream_static = { 'chunk_size': 300 }
 
-
 pbar = tqdm(total=len(real_streams))
 
-
 for f_id, f in enumerate(real_streams):
     out = []
 
@@ -86,21 +83,22 @@
     if f_id==5:
         drfs=[9,35,60,180,220]
 
-    # clf = [GaussianNB(), MLPClassifier()]
-    # clf = [GaussianNB(), MLPClassifier(), ELMI(probing_rate=1., update_rate=1.)]
-    clf = [GaussianNB(), MLPClassifier(), DecisionTreeClassifier()]
+    clf = [GaussianNB(), MLPClassifier()]
 
     evaluator = sl.evaluators.TestThenTrain()
     evaluator.process(stream, clf)
 
     fig, ax = plt.subplots(1,1,figsize=(10,5))
 
     for i in range(len(clf)):
-        plt.plot(evaluator.scores[i,:,1], alpha=0.3, label=['gnb', 'mlp', 'elm9', 'elm1'][i])
-    plt.vlines(drfs,0.5,1, color='r')
-    # plt.xticks(np.linspace(0,250,30))
-    plt.legend()
-    plt.grid()
+        ax.plot(evaluator.scores[i,:,1], alpha=0.9, label=['GNB', 'MLP'][i], c=['blue','tomato'][i], lw=1)
+    plt.xticks(drfs, rotation=90)
+    plt.legend(frameon=False)
+    plt.grid(ls=':')
+    ax.spines['top'].set_visible(False)
+    ax.spines['right'].set_visible(False)
+    ax.set_xlabel('chunk')
+    ax.set_ylabel('BAC')
 
     plt.tight_layout()
     plt.savefig('real_streams_gt/%s.png' % fname)

diff --git a/real_streams_gt/INSECTS-abrupt_imbalanced_norm.png b/real_streams_gt/INSECTS-abrupt_imbalanced_norm.png
diff --git a/real_streams_gt/INSECTS-gradual_imbalanced_norm.png b/real_streams_gt/INSECTS-gradual_imbalanced_norm.png
diff --git a/real_streams_gt/INSECTS-incremental_imbalanced_norm.png b/real_streams_gt/INSECTS-incremental_imbalanced_norm.png
diff --git a/real_streams_gt/clf_INSECTS-abrupt_imbalanced_norm.npy b/real_streams_gt/clf_INSECTS-abrupt_imbalanced_norm.npy
diff --git a/real_streams_gt/clf_INSECTS-gradual_imbalanced_norm.npy b/real_streams_gt/clf_INSECTS-gradual_imbalanced_norm.npy
diff --git a/real_streams_gt/clf_INSECTS-incremental_imbalanced_norm.npy b/real_streams_gt/clf_INSECTS-incremental_imbalanced_norm.npy
diff --git a/real_streams_gt/clf_covtypeNorm-1-2vsAll-pruned.npy b/real_streams_gt/clf_covtypeNorm-1-2vsAll-pruned.npy
diff --git a/real_streams_gt/clf_electricity.npy b/real_streams_gt/clf_electricity.npy
diff --git a/real_streams_gt/clf_poker-lsn-1-2vsAll-pruned.npy b/real_streams_gt/clf_poker-lsn-1-2vsAll-pruned.npy
diff --git a/real_streams_gt/covtypeNorm-1-2vsAll-pruned.png b/real_streams_gt/covtypeNorm-1-2vsAll-pruned.png
diff --git a/real_streams_gt/electricity.png b/real_streams_gt/electricity.png
diff --git a/real_streams_gt/poker-lsn-1-2vsAll-pruned.png b/real_streams_gt/poker-lsn-1-2vsAll-pruned.png
diff --git a/res/real_0_clustering.npy b/res/real_0_clustering.npy
diff --git a/res/real_0_complexity.npy b/res/real_0_complexity.npy
diff --git a/res/real_1_clustering.npy b/res/real_1_clustering.npy
diff --git a/res/real_1_complexity.npy b/res/real_1_complexity.npy
diff --git a/res/real_2_clustering.npy b/res/real_2_clustering.npy
diff --git a/res/real_2_complexity.npy b/res/real_2_complexity.npy
diff --git a/res/real_3_clustering.npy b/res/real_3_clustering.npy
diff --git a/res/real_3_complexity.npy b/res/real_3_complexity.npy
diff --git a/res/real_4_clustering.npy b/res/real_4_clustering.npy
diff --git a/res/real_4_complexity.npy b/res/real_4_complexity.npy
diff --git a/res/real_5_clustering.npy b/res/real_5_clustering.npy
diff --git a/res/real_5_complexity.npy b/res/real_5_complexity.npy
diff --git a/utils.py b/utils.py
@@ -51,57 +51,14 @@ def find_real_drift(chunks, drifts):
         selected2_indexes.append(m_id)
 
 selected2_measure_names = measure_labels_selected_flat[selected2_indexes]
-
-from sklearn.base import BaseEstimator, ClassifierMixin
-
-class ELMI(BaseEstimator, ClassifierMixin):
-    def __init__(self, hidden_layer_size=1024, 
-                 probing_rate=.1,
-                 update_rate=.1):
-        self.hidden_layer_size = hidden_layer_size
-        self.probing_rate = probing_rate
-        self.update_rate = update_rate
-
-    def partial_fit(self, X, y, classes=None):
-
-        if classes is None:
-            classes = np.unique(y)
-        if not hasattr(self, 'enc'):
-            self.enc = np.arange(len(classes))
-        _y = (np.array([yi==self.enc for yi in y]).astype(int))
-
-        # Check if first
-        if not hasattr(self, 'beta_'):
-            # Get problem info
-            self.n_classes = _y.shape[1]
-            self.n_features = X.shape[1]
-
-            # Initialize W
-            self.coefs_ = np.random.uniform(-1, 1, size=(self.n_features,
-                                                         self.hidden_layer_size))
-            # Initialize bias
-            self.intercepts_ = np.random.normal(size=(self.hidden_layer_size,))
-
-            # Initialize empty beta
-            self.beta_ = np.zeros((self.hidden_layer_size, self.n_classes))
-
-        pmask = np.random.uniform(size=_y.shape[0]) < self.probing_rate
-
-        H = self.activation(X[pmask].dot(self.coefs_) + self.intercepts_)  # Propagate
-        H_pinv = np.linalg.pinv(H)                                  # Inverse by Moore–Penrose
-
-        # Calculate partial beta and update beta
-        partial_beta = H_pinv.dot(_y[pmask]) 
-        self.beta_ = self.beta_ * (1-self.update_rate) + partial_beta * self.update_rate
-
-        return self
-
-    def predict_proba(self, X):
-        H = self.activation(X.dot(self.coefs_) + self.intercepts_)
-        return H.dot(self.beta_)
-
-    def predict(self, X):
-        return np.argmax(self.predict_proba(X), axis=1)
-
-    def activation(self, x):
-        return 1. / (1. + np.exp(-x))
+
+drift_gt = {
+    'covtypeNorm-1-2vsAll-pruned': [ 57, 121, 131, 155, 205, 260, 295, 350],
+    'electricity': [ 20,  38,  55, 115, 145],
+    'poker-lsn-1-2vsAll-pruned': [  45,   90,  110,  120,  160,  182,  245,  275,  292,  320,  358,  400,  450,  468,
+    480,  516,  540,  550,  590,  600,  640,  710,  790,  831,  850,  880,  900,  920,
+    965, 1000, 1010],
+    'INSECTS-abrupt_imbalanced_norm': [125],
+    'INSECTS-gradual_imbalanced_norm': [  9,  60,  90, 125, 190],
+    'INSECTS-incremental_imbalanced_norm': [  9,  35,  60, 180, 220]
+}