|
|
|
The library's functioning parameters are described by the parameters contained in the ini files.
|
|
|
|
|
|
|
|
The settings_common.ini file describes the general parameters of the system functioning:
|
|
|
|
|
|
|
|
[MainParams]
|
|
|
|
processing_type (Novelty filter or Predict) – choose system processing type mode
|
|
|
|
continuous_mode (bool) – if True, then if GUI mode will perform automatically. Otherwise, it’s will need to manually press the "Nxt" button to perform each tact
|
|
|
|
draw_layers (bool) – if True, then in GUI RNN1 and RNN2 layers will be drawn
|
|
|
|
|
|
|
|
[Forecasting] (use if processing_type == Predict)
|
|
|
|
predictstepsnum (int) – forecasting horizon
|
|
|
|
|
|
|
|
[NoveltyFiltering] (use if processing_type == Novelty filter)
|
|
|
|
Inithistoryperiod (int) – the number of the first elements of the processed sample during which the novelty is not determined (initialization period)
|
|
|
|
novfiltweightsgain (float) – the gain factor of the weights responsible for the novelty data in RNN2
|
|
|
|
novfiltdetectborder (float) – novelty detection border
|
|
|
|
|
|
|
|
[RnnGeometry] – layers and logical fields sizes in RNN1 and RNN2
|
|
|
|
l (int)
|
|
|
|
m (int)
|
|
|
|
d (int)
|
|
|
|
q (int)
|
|
|
|
|
|
|
|
[RnnDataStreaming]
|
|
|
|
sspsubmitinterval (int) – tacts interval for next ssp submitting
|
|
|
|
|
|
|
|
[NeuronParams]
|
|
|
|
refractinterval (int) – the number of tacts of neuronal refractoriness after excitation.
|
|
|
|
|
|
|
|
|
|
|
|
The settings_rnn1.ini file file describes RNN1 parameters:
|
|
|
|
|
|
|
|
[OutputFields] (struct [int, int, int]) – the number of the layer, the number horizontally and the number vertically of the logical field from which the output data is taken
|
|
|
|
|
|
|
|
[IOParams]
|
|
|
|
dictionary_filename – the path to the file with the list of words (word links) processed in the neural network
|
|
|
|
input_data_filename – the path to the file with the processed data encoded in SSP format
|
|
|
|
|
|
|
|
[ControlParams]
|
|
|
|
flag_learning (bool) – if True, then the weights of synapses change during the passage of data through the neural network
|
|
|
|
|
|
|
|
|
|
|
|
The settings_rnn2.ini file file describes RNN2 parameters:
|
|
|
|
|
|
|
|
[NeuronParams]
|
|
|
|
border_type (Const or Concurrent) – neurons excitation border type
|
|
|
|
border_const_value (float) – border value for border_type == Const
|
|
|
|
border_concurrent_winners (int) – number of winner neurons in one logical field for border_type == Concurrent
|
|
|
|
|
|
|
|
[OutputFields] – similarly settings_rnn1.ini |
