module Ai4cr::NeuralNetwork::Rnn::RnnSimpleConcerns::CalcGuess
Direct including types
Defined in:
ai4cr/neural_network/rnn/rnn_simple_concerns/calc_guess.crInstance Method Summary
- #bias_default : Float64
- #bias_default=(bias_default : Float64)
- #bias_disabled
- #bias_disabled=(bias_disabled)
- #deriv_scale : Float64
- #deriv_scale=(deriv_scale : Float64)
-
#errors
TODO For 'errors', research using a key of an Enum instead of String.
-
#eval(input_set_given)
steps for 'eval' aka 'guess':
- #guesses_ceiled
-
#guesses_sorted
guesses
- #guesses_top_n(n)
- #hidden_layer_qty
- #hidden_size
- #hidden_size_given
- #input_set_given
- #input_size
- #inputs_for(li, ti)
-
#io_offset
The 'io_offset' param is for setting, for a given time column, how much the inputs and outputs should be offset.
- #learning_rate : Float64
- #learning_rate=(learning_rate : Float64)
- #learning_styles
- #learning_styles=(learning_styles)
- #mini_net_set
- #mini_net_set=(mini_net_set)
- #momentum : Float64
- #momentum=(momentum : Float64)
- #node_input_sizes
- #node_input_sizes=(node_input_sizes)
- #node_output_sizes
- #node_output_sizes=(node_output_sizes)
- #output_size
- #outputs_guessed
- #synaptic_layer_index_last
- #synaptic_layer_indexes
- #synaptic_layer_indexes_reversed
- #synaptic_layer_qty
- #time_col_index_last
- #time_col_indexes
- #time_col_indexes_reversed
- #time_col_qty
- #valid
- #weight_init_scale : Float64
- #weight_init_scale=(weight_init_scale : Float64)
Instance Method Detail
TODO For 'errors', research using a key of an Enum instead of String. (Using Symbol's seems incompatible with 'from_json'.)
The 'io_offset' param is for setting, for a given time column, how much the inputs and outputs should be offset. For example, let's say the inputs and outputs are weather data and you want to guess tomorrow's weather based on today's and the past weather. * Setting 'io_offset' value to '-1' would mean that (we're just init'ing it or...) the outputs in tc number 0 would also represent weather data for day number -1 (which would be guessing yesterday's weather, which would overlap with the input data and probably not be of much help) * Setting 'io_offset' value to '0' would mean that the outputs in tc # 0 would also represent weather data for day # 0 (straight pass-thru; not good for guessing the future, but good for translating one set of data to another, like English to Spanish or speech to text) * Setting 'io_offset' value to '1' would mean that the outputs in tc # 0 would also represent weather data for day # 1 (and would let you guess tomorrow's weather based on today's and the past weather)