我正试图在model = Sequential()
model.add(Dense(16, 8, init='uniform', activation='tanh'))
model.add(Dropout(0.5))
model.add(Dense(8, 2, init='uniform', activation='tanh'))
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='mean_squared_error', optimizer=sgd)
model.fit(X_train, y_train, nb_epoch=1000, batch_size=50)
score = model.evaluate(X_test, y_test, batch_size=50)
中构建一个非常简单的多层感知器（MLP）：
X_train.shape我的训练数据形状：(34180, 16)给出了y_train.shape
标签属于具有形状的二进制类：(34180,)给出keras
因此，我的16x8 => 8x2代码应该生成具有以下连接的网络：ValueError: Input dimension mis-match. (input[0].shape[1] = 2, input[1].shape[1] = 1)
Apply node that caused the error: Elemwise{sub,no_inplace}(Elemwise{Composite{tanh((i0 + i1))}}[(0, 0)].0, <TensorType(float64, matrix)>)
Inputs types: [TensorType(float64, matrix), TensorType(float64, matrix)]
Inputs shapes: [(50, 2), (50, 1)]
Inputs strides: [(16, 8), (8, 8)]
产生形状不匹配错误：
Epoch 0 model.fit(X_train, y_train, nb_epoch=1000, batch_size=50)行<div ng-app="app" ng-controller="demoController">
<ion-item ng-repeat="post in data | filter:{activated:true}" ng-click="deativate(post.id)">
{{post.col1}} {{post.activated }} {{"click Me"}} </br>
</ion-item>
</div>
var app= angular.module("app",[]);
app.controller('demoController', ['$scope',
function($scope) {
$scope.data = [
{id:1,col1:"abc",activated:true},
{id:2,col1:"abc1",activated:true},
{id:3,col1:"abc2",activated:true},
{id:4,col1:"abc3",activated:true},
{id:5,col1:"abc4",activated:false},
{id:6,col1:"abc5",activated:true},
];
$scope.deativate = function(id){
angular.forEach($scope.data,function(item){
if(item.id === id){
item.activated = false;
}
})
}
}]);
。我是否在监督Keras中显而易见的事情？
编辑：我已经完成了问题here，但没有解决我的问题
答案 0 :(得分：10)
我遇到了同样的问题，然后找到了这个帖子;
https://github.com/fchollet/keras/issues/68
您似乎要说明最终输出图层为2或任意数量的类别，标签需要是分类类型，其中每个观察点基本上是一个二进制向量，例如3类输出向量[0,2] ，1,0,1,0]变为[[1,0,0]，[0,0,1]，[0,1,0]，[1,0,0]，[0,1,0] ，[1,0,0]。
np_utils.to_categorical函数为我解决了这个问题;
from keras.utils import np_utils, generic_utils
y_train, y_test = [np_utils.to_categorical(x) for x in (y_train, y_test)]