Merge pull request #19 from aigamedev/params

Construction from Layer specification objects.

README.rst

@@ -73,17 +73,17 @@ The library supports both regressors (to estimate continuous outputs) and classi
 
 .. code:: python
 
-    import sknn.mlp
+    from sknn import mlp
 
-    nn = sknn.mlp.MultiLayerPerceptronClassifier(
-        layers=[("Rectifier", 100), ("Linear",)],
+    nn = mlp.Classifier(
+        layers=[mlp.Layer("Rectifier", units=100), mlp.Layer("Linear")],
         learning_rate=0.02,
         n_iter=10)
 
     nn.fit(X_train, y_train)
-    nn.predict(X_test)
+    nn.predict(X_valid)
 
-    nn.score(X_valid, y_valid)
+    nn.score(X_test, y_test)
 
 We currently recommend reading ``sknn/mlp.py`` for more information about the parameters.  There's also `generated documentation <http://scikit-neuralnetwork.readthedocs.org/>`_ as a standalone page.
 

docs/index.rst

@@ -39,17 +39,17 @@ The library supports both regressors (to estimate continuous outputs) and classi
 
 .. code:: python
 
-    import sknn.mlp
+    from sknn import mlp
 
-    nn = sknn.mlp.MultiLayerPerceptronClassifier(
-        layers=[("Rectifier", 100), ("Linear",)],
+    nn = mlp.Classifier(
+        layers=[mlp.Layer("Rectifier", units=100), mlp.Layer("Linear")],
         learning_rate=0.02,
         n_iter=10)
 
     nn.fit(X_train, y_train)
-    nn.predict(X_test)
+    nn.predict(X_valid)
 
-    nn.score(X_valid, y_valid)
+    nn.score(X_test, y_test)
 
 You can also use a ``MultiLayerPerceptronRegressor`` in the exact same way.  See the documentation in :mod:`sknn.mlp` for details about the construction parameters.
 

examples/bench_cifar10.py

@@ -2,8 +2,13 @@
 import numpy as np
 
 def load(name):
-    with open(name, 'rb') as f:
-        return cPickle.load(f)
+    try:
+        with open(name, 'rb') as f:
+            return cPickle.load(f)
+    except IOError:
+        import gzip
+        with gzip.open(name+'.gz', 'rb') as f:
+            return cPickle.load(f)
 
 dataset1 = load('data_batch_1')
 dataset2 = load('data_batch_2')
@@ -24,12 +29,15 @@ def load(name):
 import logging
 logging.basicConfig(format="%(message)s", level=logging.DEBUG, stream=sys.stdout)
 
-from sknn.mlp import MultiLayerPerceptronClassifier
-net = MultiLayerPerceptronClassifier(
-    [("Rectifier", n_feat*2/3), ("Rectifier", n_feat*1/3), ("Linear", n_targets)],
+from sknn import mlp
+net = mlp.Classifier(
+    layers=[
+        mlp.Layer("Rectifier", units=n_feat*2/3),
+        mlp.Layer("Rectifier", units=n_feat*1/3),
+        mlp.Layer("Softmax", units=n_targets)],
     n_iter=50,
     n_stable=10,
-    learning_rate=0.005,
+    learning_rate=0.001,
     valid_size=0.1,
     verbose=1)
 net.fit(data_train, labels_train)

examples/bench_mnist.py

@@ -35,10 +35,10 @@
     classifiers.append(('nolearn.dbn', clf))
 
 if 'sknn' in sys.argv:
-    from sknn.mlp import MultiLayerPerceptronClassifier
+    from sknn import mlp
 
-    clf = MultiLayerPerceptronClassifier(
-        layers=[("Rectifier", 300), ("Softmax",)],
+    clf = mlp.Classifier(
+        layers=[mlp.Layer("Rectifier", units=300), mlp.Layer("Softmax")],
         learning_rate=0.02,
         learning_rule='momentum',
         batch_size=25,

examples/plot_mlp.py

@@ -8,6 +8,7 @@
 
 import sys
 import time
+import logging
 import argparse
 import itertools
 import numpy as np
@@ -19,15 +20,19 @@
 from sklearn.preprocessing import StandardScaler
 from sklearn.datasets import make_moons, make_circles, make_classification
 
-from sknn.mlp import MultiLayerPerceptronClassifier
+# The neural network uses the `sknn` logger to output its information.
+import logging
+logging.basicConfig(format="%(message)s", level=logging.WARNING, stream=sys.stdout)
+
+from sknn import mlp
 
 # All possible parameter options that can be plotted, separately or combined.
 PARAMETERS = {
     'activation': ['Rectifier', 'Tanh', 'Sigmoid', 'Maxout'],
     'alpha': [0.001, 0.005, 0.01, 0.05, 0.1, 0.2],
     'dropout': [False, True],
     'iterations': [100, 200, 500, 1000],
-    'output': ['Linear', 'Softmax', 'LinearGaussian'],
+    'output': ['Softmax', 'Linear', 'Gaussian'],
     'rules': ['sgd', 'momentum', 'nesterov', 'adadelta', 'rmsprop'],
     'units': [16, 64, 128, 256],
 }
@@ -53,8 +58,9 @@
 names = []
 classifiers = []
 for (activation, alpha, dropout, iterations, output, rule, units) in itertools.product(*params):
-    classifiers.append(MultiLayerPerceptronClassifier(
-        layers=[(activation, units, 2), (output,)], random_state=1,
+    params = {'pieces': 2} if activation == "Maxout" else {}
+    classifiers.append(mlp.Classifier(
+        layers=[mlp.Layer(activation, units=units, **params), mlp.Layer(output)], random_state=1,
         n_iter=iterations, n_stable=iterations,
         dropout=dropout, learning_rule=rule, learning_rate=alpha),)