Appendix C: Definition of the Genotype

Position Name Usage
1 #Layers Determines number of hidden layers in the network.
2-31 layer_worth determines the "relative worth" of the different layers. The top "#Layers' " layers will actually exist. Position 2 determines layer #1's worth, position 3 determines layer #2's worth, ... position 31 determines layer #30's worth. This value also controls how many nodes the corresponding layer will have.
32-61 input_from determines where layer get their input from. Position 32 determines layer #1's input, position 733 determines layer #2's input, ... position 61 determines layer #30's input.
62-91 output_to Position 62 determines layer #1's output, position 63 determines layer #2's output, ... position 91 determines layer #30's output.
92-121 transfer_function determines what transfer function will be used for each node. Position 92 determines layer #1's transfer function, position 93 determines layer #2's transfer function, ... position 121 determines layer #30's transfer function.
122 learning_function Will determine which learning function will be used to train the network.
123 LF_parameter_1 Learning function's parameter #1.
124 LF_parameter_2 Learning function's parameter #2.
125 LF_parameter_3 Learning function's parameter #3.
126 sentence_group_1_worth determines how many sentences of the training set will be chosen from group 1.
127 sentence_group_2_worth determines how many sentences of the training set will be chosen from group 2.
128 sentence_group_3_worth determines how many sentences of the training set will be chosen from group 3.
129 increase_group_1 determines if the more sentences from group 1 should be added to the training set, and how many.
130 increase_group_2 determines if the more sentences from group 2 should be added to the training set, and how many.
131 increase_group_3 determines if the more sentences from group 3 should be added to the training set, and how many.
132 divide_sentences_into: determines into how many sessions the training should be divided into.
133 use_loose_words determines if independent words should be part of the training set, and how many.

Back to the Table of Content

Back to the previous subject

To the next subject