Q2. add logistic regression

ipranjal · Dec 2, 2023 · 44148a8 · 44148a8
1 parent 0ec7fa8
commit 44148a8
Show file tree

Hide file tree

Showing 16 changed files with 23 additions and 2 deletions.
diff --git a/digits.py b/digits.py
@@ -71,6 +71,10 @@
             max_depth = [5,10,15,20,50,100]
             classifer_hparam['tree'] = [{'max_depth': depth} for depth in max_depth]
 
+            solvers = ['lbfgs', 'liblinear', 'newton-cg', 'sag', 'saga']
+            classifer_hparam['lr'] = [{'solver': solver} for solver in solvers]
+
+
         # Predict the value of the digit on the test subset
         # 6.Predict and Evaluate 
             for model in models:

diff --git a/models/m23csa018_lr_lbfgs.joblib b/models/m23csa018_lr_lbfgs.joblib
diff --git a/models/m23csa018_lr_liblinear.joblib b/models/m23csa018_lr_liblinear.joblib
diff --git a/models/m23csa018_lr_newton-cg.joblib b/models/m23csa018_lr_newton-cg.joblib
diff --git a/models/m23csa018_lr_sag.joblib b/models/m23csa018_lr_sag.joblib
diff --git a/models/m23csa018_lr_saga.joblib b/models/m23csa018_lr_saga.joblib
diff --git a/models/svmgamma:0.001_C:1.joblib b/models/svmgamma:0.001_C:1.joblib
diff --git a/models/svmgamma:0.001_C:2.joblib b/models/svmgamma:0.001_C:2.joblib
diff --git a/models/svmgamma:0.001_C:5.joblib b/models/svmgamma:0.001_C:5.joblib
diff --git a/models/svmgamma:1_C:1.joblib b/models/svmgamma:1_C:1.joblib
diff --git a/models/treemax_depth:10.joblib b/models/treemax_depth:10.joblib
diff --git a/models/treemax_depth:100.joblib b/models/treemax_depth:100.joblib
diff --git a/models/treemax_depth:15.joblib b/models/treemax_depth:15.joblib
diff --git a/models/treemax_depth:20.joblib b/models/treemax_depth:20.joblib
diff --git a/models/treemax_depth:50.joblib b/models/treemax_depth:50.joblib
diff --git a/utils.py b/utils.py
@@ -1,6 +1,6 @@
 
 # Import datasets, classifiers and performance metrics
-from sklearn import svm,datasets
+from sklearn import svm,datasets,linear_model
 from sklearn.model_selection import train_test_split
 from sklearn import tree
 import matplotlib.pyplot as plt
@@ -37,6 +37,8 @@ def train_model(X, y, model_params,model_type = 'svm'):
         clf = svm.SVC(**model_params)
     if model_type == 'tree':
         clf = tree.DecisionTreeClassifier(**model_params)
+    if model_type == 'lr':
+        clf = linear_model.LogisticRegression(**model_params)
     clf.fit(X, y)
     return clf
 
@@ -62,18 +64,33 @@ def tune_hparams(X_train, Y_train, X_dev, y_dev, list_of_all_param_combination,
             cur_model = train_model(X_train, Y_train, {'gamma': param_combination['gamma'],'C':param_combination['C']}, model_type='svm')
         if model_type == 'tree':
             cur_model = train_model(X_train, Y_train, {'max_depth': param_combination['max_depth']}, model_type='tree')
+        if model_type == 'lr':
+            cur_model = train_model(X_train, Y_train, {'solver': param_combination['solver']}, model_type='lr')
 
         cur_accuracy,_ = predict_and_eval(cur_model, X_dev, y_dev)
+        if model_type == 'lr':
+            curr_model_path = "./models/m23csa018_{}".format(model_type)+"_{}".format(param_combination['solver'])+".joblib"
+            dump(cur_model,curr_model_path)
+            print(f"For model {model_type} and solver {param_combination['solver']} accuracy is {cur_accuracy}")
+
         if cur_accuracy > best_accuracy_so_far:
             best_accuracy_so_far = cur_accuracy
             if model_type == 'svm':
                 optimal_gamma = param_combination['gamma']
                 optimal_C = param_combination['C']
                 best_hparams = {'gamma': optimal_gamma,'C':optimal_C}
+                best_model_path = "./models/{}".format(model_type)+"_".join(["{}:{}".format(k,v) for k,v in best_hparams.items()])+".joblib"
+
             if model_type == 'tree':
                 optimal_max_depth = param_combination['max_depth']
                 best_hparams = {'max_depth': optimal_max_depth}
-            best_model_path = "./models/{}".format(model_type)+"_".join(["{}:{}".format(k,v) for k,v in best_hparams.items()])+".joblib"
+                best_model_path = "./models/{}".format(model_type)+"_".join(["{}:{}".format(k,v) for k,v in best_hparams.items()])+".joblib"
+
+            if model_type == 'lr':
+                optimal_solver = param_combination['solver']
+                best_hparams = {'solver': optimal_solver}
+                best_model_path = "./models/m23csa018_{}".format(model_type)+"_".join(["_{}".format(v) for k,v in best_hparams.items()])+".joblib"
+
 
             best_model = cur_model