Question

我正在使用带RNN的编码器 - 解码器模型来生成文档的摘要。经过测试，我发现输出完全没用，而且它们都是一样的。我打印了隐藏值，发现它几乎只包含1和-1。

解码器的输出：

const StackScreens = StackNavigator({
    Main: {
        screen: LoginScreen,
        navigationOptions:{
            header:null,

        }
    },
    RegisterForm: {screen: RegisterForm},
    CourseListing:{screen: CourseListing,
    navigationOptions:{
        header:null,

    }
    },
    Home:{screen: HomeScreen,
    },
    EconomicNews: {screen: EconomicNews},
    EconomicDetails: {screen: EconomicDetails},
    CategoryDetail: {screen: CategoryDetail},
    DetailedView: {screen: DetailedView},
    IndividualSection: {screen: IndividualSection},
    Mcq:{screen: Mcq},
    QASection: {screen: QASection},
    ForgotPassword: {screen: ForgotPassword,
    navigationOptions:{
        header:null
    }
    }   
})

export const MyDrawer = DrawerNavigator({
   Home: {
     screen: StackScreens,

  },
  Profile: {
      screen: Profile
  },
  FAQ: {
      screen: Faq
  },
  LogOut: {
      screen: LoginScreen,
      navigationOptions: ({navigation}) => ({
           tabBarOnPress: (scene, jumpToIndex) => {

               return Alert.alert(

                   'Do you really want to logout?',
                   [
                     {text: 'Accept', onPress: () => { navigation.dispatch(NavigationActions.navigate({ routeName: 'LoginScreen' }))

                     }},

                     {text: 'Cancel'}
                    ]

               );
           },
       })
  }
});

隐藏价值的一部分：

['<SOS>', 'the', 'new', 'york', 'city', "'s", 'mother', 'was', 'found',      'in', 'the', 'world', 'cup', '.', 'the', 'former', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'the', 'yoselyn', 'is', 'the', 'the', 'much-loved', 'condition', '.', '<EOS>']

我是深度学习的初学者，我无法理解为什么会这样。我检查了我的数据集，我认为它们很好。

编码器和解码器：

( 0 ,.,.) = 

Columns 0 to 8 
   1.0000  0.9846 -1.0000 -1.0000  1.0000 -1.0000  0.0000  1.0000  1.0000

Columns 9 to 17 
  -1.0000  0.0000  1.0000 -1.0000 -0.9019  1.0000 -1.0000  1.0000 -1.0000

Columns 18 to 26 
   1.0000  1.0000  1.0000  1.0000 -1.0000  0.0000  1.0000  1.0000 -1.0000

Columns 27 to 35 
  -1.0000 -1.0000  1.0000  0.0000 -1.0000 -1.0000 -1.0000  0.0000 -1.0000

Columns 36 to 44 
   1.0000 -1.0000 -1.0000  1.0000  0.0000 -1.0000  1.0000 -1.0000 -1.0000

RNN隐藏值几乎只包含1和-1

0 个答案: