OpenCV 3-MLP神经网络的已保存文件的含义

Question

我在OpenCV中保存了经过训练的MLP神经网络。 YAML文件如下：

%YAML:1.0
---
opencv_ml_ann_mlp:
   format: 3
   layer_sizes: [ 2, 2, 1 ]
   activation_function: SIGMOID_SYM
   f_param1: 6.6666666666666663e-01
   f_param2: 1.7159000000000000e+00
   min_val: -9.4999999999999996e-01
   max_val: 9.4999999999999996e-01
   min_val1: -9.7999999999999998e-01
   max_val1: 9.7999999999999998e-01
   training_params:
      train_method: RPROP
      dw0: 1.0000000000000001e-01
      dw_plus: 1.2000000000000000e+00
      dw_minus: 5.0000000000000000e-01
      dw_min: 1.1920928955078125e-07
      dw_max: 50.
      term_criteria:
         epsilon: 1.0000000000000000e-04
         iterations: 1000
   input_scale:
      - 2.
      - -1.
      - 2.
      - -1.
   output_scale:
      - 5.2631578947368418e-01
      - 4.9999999999999994e-01
   inv_output_scale:
      - 1.8999999999999999e+00
      - -9.4999999999999996e-01
   weights:
      -
         - -4.7194814611664775e+00
         - -6.1634082465651581e+00
         - -4.3155547465702089e+00
         - -5.1915700181726363e+00
         - 3.8901623511805283e+00
         - -5.2437247580543920e+00
      -
         - 1.2414966548773920e+00
         - -1.2394795554405691e+00
         - -1.9857133283213504e+00

我到处都找不到这些标签的解释。 input_scale，output_scale和inv_output_scale的含义是什么？定义它们的算法是什么？