Question

我想在输出层添加新节点以便以后训练它，我正在做：

def add_outputs(self, n_new_outputs):
    out = self.model.get_layer('fc8').output
    last_layer = self.model.get_layer('fc7').output
    out2 = Dense(n_new_outputs, activation='softmax', name='fc9')(last_layer)
    output = merge([out, out2], mode='concat')
    self.model = Model(input=self.model.input, output=output)

其中'fc7'是输出图层'fc8'之前的完全连接图层。我考虑只使用out = self.model.get_layer('fc8').output的最后一层，但输出是所有模型。有没有办法从网络中获取一个图层？也许还有其他更容易的方法......

感谢!!!!

Answer 1

最后我找到了解决方案：

1）从最后一层获得权重

2）在权重中添加零并随机初始化它的连接

3）弹出输出图层并创建一个新的

4）为新图层设置新权重

这里是代码：

 def add_outputs(self, n_new_outputs):
        #Increment the number of outputs
        self.n_outputs += n_new_outputs
        weights = self.model.get_layer('fc8').get_weights()
        #Adding new weights, weights will be 0 and the connections random
        shape = weights[0].shape[0]
        weights[1] = np.concatenate((weights[1], np.zeros(n_new_outputs)), axis=0)
        weights[0] = np.concatenate((weights[0], -0.0001 * np.random.random_sample((shape, n_new_outputs)) + 0.0001), axis=1)
        #Deleting the old output layer
        self.model.layers.pop()
        last_layer = self.model.get_layer('batchnormalization_1').output
        #New output layer
        out = Dense(self.n_outputs, activation='softmax', name='fc8')(last_layer)
        self.model = Model(input=self.model.input, output=out)
        #set weights to the layer
        self.model.get_layer('fc8').set_weights(weights)
        print(weights[0])

在Keras中将新节点添加到输出层

1 个答案: