Trainer::Trainer(fann* ann, float bit_fail_limit,
const TrainingAlgorithm& algorithm,
float learning_rate, float learning_momentum)
: ann_(ann) {
set_bit_fail_limit(bit_fail_limit);
set_training_algorithm(algorithm);
set_learning_rate(learning_rate);
set_learning_momentum(learning_momentum);
}
int main() {
ndata_object data,tdata;
nnet_object net;
int inp, oup, patterns;
int N,tsize;
int isheader = 1;
char* file = "datasets/gene.train";
char* tfile = "datasets/gene.test";
char *delimiter = " ";
float tmse;
inp = 120;
oup = 3;
patterns = 1588;
tsize = 1587;
data = ndata_init(inp, oup, patterns);
file_sep_line_enter(data, file, delimiter, isheader);
ndata_check(data);
N = 4;
int arch[4] = { inp, 4, 2, oup }; //
int actfcn[4] = { 0, 3, 3, 1 };// {Null,'tansig','purelin'}
net = nnet_init(N, arch, actfcn);
set_trainfcn(net, "traingda");
set_training_ratios(net, 1.0, 0.0, 0.0);
set_trainmethod(net, "online", patterns);
set_max_epoch(net, 500);
set_target_mse(net, 1e-05f);// Target MSE error
set_learning_rate(net, 0.01f);// learning rate
//set_momentum(net, 0.9);// No momentum term
set_norm_method(net,1);
set_mnmx(net, 0, 1, 0, 1);
train(net, patterns, data->data, data->target);
tdata = ndata_init(inp, oup, tsize);
file_sep_line_enter(tdata, tfile, delimiter, isheader);
tmse = nnet_test(net, tsize, tdata->data, tdata->target);
printf("\n Test MSE %f \n", tmse);
nnet_save(net, "gene.nnet");
ndata_free(data);
ndata_free(tdata);
nnet_free(net);
return 0;
}
