Machine learning Tensorflow与';正阈值0.500000平均值';?
Tensorflow发布了一系列统计数据:Machine learning Tensorflow与';正阈值0.500000平均值';?,machine-learning,tensorflow,Machine Learning,Tensorflow,Tensorflow发布了一系列统计数据: accuracy: 0.915224 accuracy/baseline_target_mean: 0.220896 accuracy/threshold_0.500000_mean: 0.915224 auc: 0.937926 global_step: 200 labels/actual_target_mean: 0.220896 labels/prediction_mean: 0.203677 loss: 0.247065 precision/p
accuracy: 0.915224
accuracy/baseline_target_mean: 0.220896
accuracy/threshold_0.500000_mean: 0.915224
auc: 0.937926
global_step: 200
labels/actual_target_mean: 0.220896
labels/prediction_mean: 0.203677
loss: 0.247065
precision/positive_threshold_0.500000_mean: 0.991379
recall/positive_threshold_0.500000_mean: 0.621622
baseline\u target\u是什么意思(精确后)和positive\u threshold\u 0.500000\u是什么意思(回忆和精确后)?既然你得到了这些统计数据,我假设你在做二元分类
baseline\u target\u mean
是数据中类标签的平均值,即在本例中,假设类标签为0和1,约22%的测试示例属于类1,其余示例属于类0。类标签可以是任意数字,因此解释取决于您的数据,我所描述的只是一种可能性,因为我不知道您使用的数据
positive_threshold_0.500000_mean
仅表示预测值高于阈值0.5的示例被视为正示例,而低于阈值0.5的示例被视为负示例