updated losses for linear svms

notebooks/ml/MachineLearningReport.ipynb

@@ -72,7 +72,7 @@
    "source": [
     "import numpy as np\n",
     "\n",
-    "from yase.ml.svm import SVC\n",
+    "from optiml.ml.svm import SVC\n",
     "\n",
     "from sklearn.datasets import fetch_openml\n",
     "from sklearn.preprocessing import OneHotEncoder\n",
@@ -241,8 +241,8 @@
    "source": [
     "import numpy as np\n",
     "\n",
-    "from yase.ml.neural_network import NeuralNetworkClassifier\n",
-    "from yase.ml.neural_network.layers import FullyConnected\n",
+    "from optiml.ml.neural_network import NeuralNetworkClassifier\n",
+    "from optiml.ml.neural_network.layers import FullyConnected\n",
     "\n",
     "from sklearn.datasets import fetch_openml\n",
     "from sklearn.preprocessing import OneHotEncoder\n",
@@ -309,10 +309,10 @@
     }
    ],
    "source": [
-    "from yase.ml.neural_network.activations import sigmoid\n",
-    "from yase.ml.neural_network.losses import mean_squared_error\n",
+    "from optiml.ml.neural_network.activations import sigmoid\n",
+    "from optiml.ml.neural_network.losses import mean_squared_error\n",
     "\n",
-    "from yase.optimization.unconstrained.line_search import SteepestGradientDescent\n",
+    "from optiml.optimization.unconstrained.line_search import SteepestGradientDescent\n",
     "\n",
     "X, y = fetch_openml('monks-problems-1', return_X_y=True)\n",
     "y = y.astype(np.float32)\n",
@@ -395,10 +395,10 @@
     }
    ],
    "source": [
-    "from yase.ml.neural_network.activations import sigmoid, softmax\n",
-    "from yase.ml.neural_network.losses import sparse_categorical_cross_entropy\n",
+    "from optiml.ml.neural_network.activations import sigmoid, softmax\n",
+    "from optiml.ml.neural_network.losses import sparse_categorical_cross_entropy\n",
     "\n",
-    "from yase.optimization.unconstrained.line_search import SteepestGradientDescent\n",
+    "from optiml.optimization.unconstrained.line_search import SteepestGradientDescent\n",
     "\n",
     "X, y = fetch_openml('monks-problems-2', return_X_y=True)\n",
     "y = y.astype(np.float32)\n",
@@ -474,13 +474,13 @@
     }
    ],
    "source": [
-    "from yase.ml.neural_network.activations import sigmoid, softmax\n",
-    "from yase.ml.neural_network.losses import categorical_cross_entropy\n",
-    "from yase.ml.neural_network.regularizers import L2\n",
+    "from optiml.ml.neural_network.activations import sigmoid, softmax\n",
+    "from optiml.ml.neural_network.losses import categorical_cross_entropy\n",
+    "from optiml.ml.neural_network.regularizers import L2\n",
     "\n",
-    "from yase.optimization.unconstrained.line_search import BFGS\n",
+    "from optiml.optimization.unconstrained.line_search import BFGS\n",
     "\n",
-    "from yase.optimization.unconstrained.stochastic import Adam\n",
+    "from optiml.optimization.unconstrained.stochastic import Adam\n",
     "\n",
     "X, y = fetch_openml('monks-problems-3', return_X_y=True)\n",
     "y = y.astype(np.float32)\n",
@@ -567,8 +567,8 @@
     }
    ],
    "source": [
-    "from yase.ml.utils import plot_validation_curve, plot_learning_curve\n",
-    "from yase.ml.svm import SVR\n",
+    "from optiml.ml.utils import plot_validation_curve, plot_learning_curve\n",
+    "from optiml.ml.svm import SVR\n",
     "\n",
     "from sklearn.metrics import make_scorer\n",
     "from sklearn.model_selection import GridSearchCV\n",
@@ -689,4 +689,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 4
-}
+}

notebooks/ml/SupportVectorMachines.ipynb

@@ -248,8 +248,8 @@
     }
    ],
    "source": [
-    "from yase.ml.svm import SVC\n",
-    "from yase.ml.utils import generate_linearly_separable_data, plot_svm_hyperplane\n",
+    "from optiml.ml.svm import SVC\n",
+    "from optiml.ml.utils import generate_linearly_separable_data, plot_svm_hyperplane\n",
     "\n",
     "from sklearn.svm import SVC as SKLSVC\n",
     "from sklearn.model_selection import train_test_split\n",
@@ -403,8 +403,8 @@
     }
    ],
    "source": [
-    "from yase.ml.svm import SVC\n",
-    "from yase.ml.utils import generate_linearly_separable_overlap_data, plot_svm_hyperplane\n",
+    "from optiml.ml.svm import SVC\n",
+    "from optiml.ml.utils import generate_linearly_separable_overlap_data, plot_svm_hyperplane\n",
     "\n",
     "from sklearn.svm import SVC as SKLSVC\n",
     "from sklearn.model_selection import train_test_split\n",
@@ -509,8 +509,8 @@
     }
    ],
    "source": [
-    "from yase.ml.svm import SVC\n",
-    "from yase.ml.utils import generate_non_linearly_separable_data, plot_svm_hyperplane\n",
+    "from optiml.ml.svm import SVC\n",
+    "from optiml.ml.utils import generate_non_linearly_separable_data, plot_svm_hyperplane\n",
     "\n",
     "from sklearn.svm import SVC as SKLSVC\n",
     "from sklearn.model_selection import train_test_split\n",
@@ -593,8 +593,8 @@
     }
    ],
    "source": [
-    "from yase.ml.svm import SVC\n",
-    "from yase.ml.utils import generate_non_linearly_separable_data, plot_svm_hyperplane\n",
+    "from optiml.ml.svm import SVC\n",
+    "from optiml.ml.utils import generate_non_linearly_separable_data, plot_svm_hyperplane\n",
     "\n",
     "from sklearn.svm import SVC as SKLSVC\n",
     "from sklearn.model_selection import train_test_split\n",
@@ -677,8 +677,8 @@
     }
    ],
    "source": [
-    "from yase.ml.svm import SVC\n",
-    "from yase.ml.utils import generate_non_linearly_separable_data, plot_svm_hyperplane\n",
+    "from optiml.ml.svm import SVC\n",
+    "from optiml.ml.utils import generate_non_linearly_separable_data, plot_svm_hyperplane\n",
     "\n",
     "from sklearn.svm import SVC as SKLSVC\n",
     "from sklearn.model_selection import train_test_split\n",
@@ -767,8 +767,8 @@
     }
    ],
    "source": [
-    "from yase.ml.svm import SVC\n",
-    "from yase.ml.utils import generate_non_linearly_separable_data, plot_svm_hyperplane\n",
+    "from optiml.ml.svm import SVC\n",
+    "from optiml.ml.utils import generate_non_linearly_separable_data, plot_svm_hyperplane\n",
     "\n",
     "from sklearn.svm import SVC as SKLSVC\n",
     "from sklearn.model_selection import train_test_split\n",
@@ -958,8 +958,8 @@
     }
    ],
    "source": [
-    "from yase.ml.svm import SVR\n",
-    "from yase.ml.utils import generate_non_linearly_regression_data, plot_svm_hyperplane\n",
+    "from optiml.ml.svm import SVR\n",
+    "from optiml.ml.utils import generate_non_linearly_regression_data, plot_svm_hyperplane\n",
     "\n",
     "from sklearn.svm import SVR as SKLSVR\n",
     "from sklearn.model_selection import train_test_split\n",
@@ -1002,4 +1002,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 4
-}
+}

notebooks/optimization/LineSearchMethods.ipynb

@@ -38,7 +38,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from yase.optimization.unconstrained import Quadratic, Rosenbrock, SixHumpCamel\n",
+    "from optiml.optimization.unconstrained import Quadratic, Rosenbrock, SixHumpCamel\n",
     "\n",
     "quad = Quadratic(Q=[[5, -3], \n",
     "                    [-3, 5]], q=[10, 5])\n",
@@ -52,8 +52,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from yase.optimization.unconstrained.line_search import *\n",
-    "from yase.optimization.utils import plot_trajectory_optimization"
+    "from optiml.optimization.unconstrained.line_search import *\n",
+    "from optiml.optimization.utils import plot_trajectory_optimization"
    ]
   },
   {
@@ -1407,4 +1407,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 4
-}
+}

notebooks/optimization/StochasticMethods.ipynb

@@ -43,15 +43,15 @@
    "source": [
     "import numpy as np\n",
     "\n",
-    "from yase.ml.utils import generate_centred_and_normalized_regression_data\n",
-    "from yase.ml.neural_network import NeuralNetworkRegressor\n",
-    "from yase.ml.neural_network.layers import FullyConnected\n",
-    "from yase.ml.neural_network.activations import linear\n",
-    "from yase.ml.neural_network.losses import mean_squared_error\n",
-    "from yase.ml.neural_network.regularizers import L2\n",
+    "from optiml.ml.utils import generate_centred_and_normalized_regression_data\n",
+    "from optiml.ml.neural_network import NeuralNetworkRegressor\n",
+    "from optiml.ml.neural_network.layers import FullyConnected\n",
+    "from optiml.ml.neural_network.activations import linear\n",
+    "from optiml.ml.neural_network.losses import mean_squared_error\n",
+    "from optiml.ml.neural_network.regularizers import L2\n",
     "\n",
-    "from yase.optimization.utils import plot_trajectory_optimization\n",
-    "from yase.optimization.unconstrained.stochastic import *\n",
+    "from optiml.optimization.utils import plot_trajectory_optimization\n",
+    "from optiml.optimization.unconstrained.stochastic import *\n",
     "\n",
     "X, y = generate_centred_and_normalized_regression_data(random_state=1)"
    ]
@@ -812,4 +812,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 4
-}
+}

optiml/ml/neural_network/_base.py

@@ -124,7 +124,7 @@ def _store_train_val_info(self, opt, X_batch, y_batch, X_val, y_val):
                 print('\tloss: {:1.4e}'.format(self._avg_epoch_loss), end='')
             self._avg_epoch_loss = 0.
             if self.validation_split:
-                val_loss = self.loss(opt.x, X_val, y_val)
+                val_loss = self.loss.function(opt.x, X_val, y_val)
                 self.val_loss_history.append(val_loss)
                 if self.verbose and not opt.epoch % self.verbose:
                     print(' - val_loss: {:1.4e}'.format(val_loss), end='')

optiml/ml/svm/__init__.py

@@ -1,3 +1,3 @@
-__all__ = ['SVM', 'PrimalSVC', 'DualSVC', 'PrimalSVR', 'DualSVR']
+__all__ = ['SVM', 'LinearSVC', 'SVC', 'LinearSVR', 'SVR']
 
-from ._base import SVM, PrimalSVC, DualSVC, PrimalSVR, DualSVR
+from ._base import SVM, LinearSVC, SVC, LinearSVR, SVR