class VADEstimator(BaseEstimator):
def fit( self, x , y , size=1 ):
self.model = Sequential()
self.model.add(Dense( int( embeddings_dim / 2.0 ) , input_dim=embeddings_dim , init='uniform' , activation='tanh'))
self.model.add(Dense( int( embeddings_dim / 4.0 ) , init='uniform' , activation='tanh'))
self.model.add(Dense(size , init='uniform' ) )
self.model.compile(loss='mse', optimizer='rmsprop')
self.model = KernelRidge( kernel='rbf' )
self.model.fit( x , y )
def predict( self, x ):
if isinstance( self.model , Sequential ): return self.model.predict( x , verbose=0 )[ 0 ]
return self.model.predict( x )
class VADEstimator(BaseEstimator):
def fit(self, x, y, size=1):
self.model = Sequential()
self.model.add(
Dense(int(embeddings_dim / 2.0),
input_dim=embeddings_dim,
init='uniform',
activation='tanh'))
self.model.add(
Dense(int(embeddings_dim / 4.0), init='uniform',
activation='tanh'))
self.model.add(Dense(size, init='uniform'))
self.model.compile(loss='mse', optimizer='rmsprop')
self.model = KernelRidge(kernel='poly', degree=4)
self.model.fit(x, y)
def predict(self, x):
if isinstance(self.model, Sequential):
return self.model.predict(x, verbose=0)[0]
return self.model.predict(x)
print ("Method = Linear ridge regression with bag-of-words features")
model = KernelRidge( kernel='linear' )
model.fit( train_matrix , train_labels )
results = model.predict( test_matrix )
if not(is_geocoding):
print ("RMSE = " + repr( np.sqrt(mean_squared_error( test_labels , results )) ) )
print ("MAE = " + repr( mean_absolute_error( test_labels , results ) ) )
else:
print ("Mean error = " + repr( np.mean( [ geodistance( results[i] , test_labels[i] ) for i in range(results.shape[0]) ] ) ) )
print ("Median error = " + repr( np.median( [ geodistance( results[i] , test_labels[i] ) for i in range(results.shape[0]) ] ) ) )
print ("")
print ("Method = MLP with bag-of-words features")
np.random.seed(0)
model = Sequential()
model.add(Dense(embeddings_dim, input_dim=train_matrix.shape[1], init='uniform', activation='relu'))
model.add(Dropout(0.25))
model.add(Dense(embeddings_dim, activation='relu'))
model.add(Dropout(0.25))
model.add(Dense(reg_dimensions, activation='sigmoid'))
if not(is_geocoding): model.compile(loss='mean_absolute_error', optimizer='adam')
else: model.compile(loss=geoloss, optimizer='adam')
model.fit( train_matrix , train_labels , nb_epoch=30, batch_size=16)
results = model.predict( test_matrix )
if not(is_geocoding):
print ("RMSE = " + repr( np.sqrt(mean_squared_error( test_labels , results )) ) )
print ("MAE = " + repr( mean_absolute_error( test_labels , results ) ) )
else:
print ("Mean error = " + repr( np.mean( [ geodistance( results[i] , test_labels[i] ) for i in range(results.shape[0]) ] ) ) )
print ("Median error = " + repr( np.median( [ geodistance( results[i] , test_labels[i] ) for i in range(results.shape[0]) ] ) ) )
model.fit( train_matrix2 , train_labels_aux )
results = encoder.inverse_transform( model.predict( test_matrix2 ) )
results = np.array( [ healpix2latlon( code , resolution ) for code in results ] )
print ("Mean error = " + repr( np.mean( [ geodistance( results[i] , test_labels2[i] ) for i in range(results.shape[0]) ] ) ) )
print ("Median error = " + repr( np.median( [ geodistance( results[i] , test_labels2[i] ) for i in range(results.shape[0]) ] ) ) )
print ("")
np.random.seed(0)
print ("Method = MLP classification - Default features")
encoder = preprocessing.LabelEncoder( )
train_labels_aux = np.array( [ latlon2healpix( lat , lon , resolution ) for (lat,lon) in train_labels1 ] )
train_labels_aux = np.array( encoder.fit_transform( train_labels_aux ) )
num_classes = len( set( train_labels_aux ) )
train_labels_aux = np_utils.to_categorical( train_labels_aux , num_classes )
model = Sequential( )
model.add(Dense(hidden_dim, input_dim=train_matrix1.shape[1], init='uniform', activation='sigmoid'))
model.add(Dropout(0.25))
model.add(Dense(hidden_dim / 2, activation='sigmoid', init='uniform'))
model.add(Dropout(0.25))
model.add(Dense( num_classes , activation='softmax' , init='uniform' ))
model.compile(loss='categorical_crossentropy', optimizer='adam')
model.fit( train_matrix1 , train_labels_aux , nb_epoch=1500, batch_size=16, verbose=1)
results = encoder.inverse_transform( np_utils.categorical_probas_to_classes( model.predict( test_matrix1 ) ) )
results = np.array( [ healpix2latlon( code , resolution ) for code in results ] )
print ("Mean error = " + repr( np.mean( [ geodistance( results[i] , test_labels1[i] ) for i in range(results.shape[0]) ] ) ) )
print ("Median error = " + repr( np.median( [ geodistance( results[i] , test_labels1[i] ) for i in range(results.shape[0]) ] ) ) )
print ("Method = MLP classification - Default features + chromatic features")
np.random.seed(0)
encoder = preprocessing.LabelEncoder( )
train_labels_aux = np.array( [ latlon2healpix( lat , lon , resolution ) for (lat,lon) in train_labels1 ] )
train_labels_aux = np.array( encoder.fit_transform( train_labels_aux ) )
geodistance(results[i], test_labels[i])
for i in range(results.shape[0])
])))
print("Median error = " + repr(
np.median([
geodistance(results[i], test_labels[i])
for i in range(results.shape[0])
])))
print("")
print("Method = MLP with bag-of-words features")
np.random.seed(0)
model = Sequential()
model.add(
Dense(embeddings_dim,
input_dim=train_matrix.shape[1],
init='uniform',
activation='relu'))
model.add(Dropout(0.25))
model.add(Dense(embeddings_dim, activation='relu'))
model.add(Dropout(0.25))
model.add(Dense(reg_dimensions, activation='sigmoid'))
if not (is_geocoding):
model.compile(loss='mean_absolute_error', optimizer='adam')
else:
model.compile(loss=geoloss, optimizer='adam')
model.fit(train_matrix, train_labels, nb_epoch=30, batch_size=16)
results = model.predict(test_matrix)
if not (is_geocoding):
print("RMSE = " + repr(np.sqrt(mean_squared_error(test_labels, results))))
print("MAE = " + repr(mean_absolute_error(test_labels, results)))
