Package: nnlib2Rcpp

Type: Package

Title: A Tool for Creating Custom Neural Networks in C++ and using Them in R

Version: 0.2.5.5

Version: 0.2.5.6

Author: Vasilis Nikolaidis [aut, cph, cre] (<https://orcid.org/0000-0003-1471-8788>)

Maintainer: Vasilis Nikolaidis <v.nikolaidis@uop.gr>

Description: Contains a module to define neural networks from custom components and versions of Autoencoder, BP, LVQ, MAM NN.

INPUT_LAYER.input_data_from_vector(input,input_dim);

recall();

int current_winner_pe=0;

int current_winner_pe_rewards=0;

DATA current_win_output = OUTPUT_LAYER.PE(0).output;// this should be the distance, see lvq_output_layer::recall.

if(min_rewards>0)

{

bool found_rewarded =false;

for(int i=0;i<output_dimension() AND NOT found_rewarded;i++)

{

if(OUTPUT_LAYER.PE(i).misc >= min_rewards)// misc in output PEs is just a counter of rewards given to the PE

{

current_winner_pe = i;

found_rewarded =true;

}

}

if(NOT found_rewarded)

{

error(NN_METHOD_ERR,"No output node has requested number of rewards");

return(-1);

}

}

DATA current_win_output = OUTPUT_LAYER.PE(current_winner_pe).output;// this should be the distance, see lvq_output_layer::recall.

for(int i=0;i<output_dimension();i++)

{

OUTPUT_LAYER.PE(i).bias = LVQ_DEACTI_PE;

OUTPUT_LAYER.PE(i).bias = LVQ_DEACTI_PE;// (is this really necessary?) deactivate all

if(OUTPUT_LAYER.PE(i).misc >= min_rewards)// misc in output PEs is just a counter of rewards given to the PE

{

{

current_win_output = d;

current_winner_pe = i;

current_winner_pe_rewards = OUTPUT_LAYER.PE(i).misc;

}

}

}

returned_class =

(int)(current_winner_pe / m_number_of_output_nodes_per_class);

if(current_winner_pe_rewards<min_rewards)

{

error(NN_METHOD_ERR,"No winning output node had required number of rewards");

returned_class = -1;

}

}

return returned_class;

break;

case NN_METHOD_ERR:

m1 ="methodfailure";

m1 ="methoderror";

severity=1;

break;