h2o deeplearning：我是否需要重新调整输入数值变量

Question

我试图用h2o deeplearning模型预测出租车时间：

  deep<-h2o.deeplearning(
        training_frame = train,
        validation_frame = valid,
        x=predictors,
        y=target,
        #distribution = "gaussian",
        #loss = "Automatic",
        #hidden=c(30,30),
        epochs = 50,
        #activation="Rectifier",
        stopping_metric="deviance",
        stopping_tolerance=1e-5,      # stops when deviance does 
                                not improve by >=0.0001 for 5 scoring events
        stopping_rounds=5

        )

这是输入变量的样子，TAXI_OUT是目标，它在几分钟内，当然总是> 0：

   DAY_OF_WEEK CARRIER ORIGIN DEST TAXI_OUT congestion sin_deptime cos_deptime dep_blk_sin dep_blk_cos Temp Dew_point
18           1      DL    ATL  PHL       32         53 -0.80644460   0.5913096  -0.3246995   0.9458172   11        12
24           1      DL    ATL  EWR       23         75 -0.40673664   0.9135455   0.8371665   0.5469482   11        12
25           1      DL    ATL  EWR       24         55  0.68199836  -0.7313537   0.4759474  -0.8794738   11        12
30           1      DL    ATL  FLL       35         52 -0.04361939  -0.9990482  -0.7357239  -0.6772816   11        12
32           1      DL    ATL  PBI       30         68 -0.78260816  -0.6225146  -0.9694003   0.2454855   11        12
36           1      DL    ATL  DTW       13         50 -0.68835458   0.7253744   0.6142127   0.7891405   11        12
   Humidity Sea_Level_Press Visibility Wind Event_1 Event_2      Event_3
18       99            1019          2   11     Fog    Rain Thunderstorm
24       99            1019          2   11     Fog    Rain Thunderstorm
25       99            1019          2   11     Fog    Rain Thunderstorm
30       99            1019          2   11     Fog    Rain Thunderstorm
32       99            1019          2   11     Fog    Rain Thunderstorm
36       99            1019          2   11     Fog    Rain Thunderstorm

我是否需要重新调整某个范围内的数字输入变量，例如[0,1]或[-1,1]，还是让h2o处理它们？

Answer 1

H2O自动处理缩放。你不需要做任何事情。