double Perceptron::calculate_output(const Vector<double>& inputs) const
{
// Control sentence (if debug)
#ifdef __OPENNN_DEBUG__
const size_t size = inputs.size();
const size_t inputs_number = get_inputs_number();
if(size != inputs_number)
{
std::ostringstream buffer;
buffer << "OpenNN Exception: Perceptron class.\n"
<< "double calculate_output(const Vector<double>&) const method.\n"
<< "Size must be equal to number of inputs.\n";
throw std::logic_error(buffer.str());
}
#endif
// Calculate outputs
return(calculate_activation(calculate_combination(inputs)));
}
inline void gpu_train_batch(FeedForward_Network<activation, error>& network,
arma::Mat<float> inputs, arma::Mat<float> targets, int batch_size, float learning_rate = 0.8f, float momentum = 0.8f) {
network.resize_activation(batch_size);
Raw_FeedForward_Network<activation, error> raw_net = convert_to_raw(network);
Raw_FeedForward_Network<activation, error> * d_network = network_to_gpu(raw_net);
int batches_in_train = targets.n_rows/batch_size - 1;
for (int i = 0; i < batches_in_train; ++i) {
arma::Mat<float> input_slice = inputs.rows(i*batch_size, (i+1) * batch_size - 1);
Raw_Matrix raw_input = to_raw(input_slice);
Raw_Matrix * d_input = matrix_to_gpu(raw_input);
int num_trials = input_slice.n_rows;
calculate_activation(num_trials, network.layer_sizes, d_network, d_input);
//TODO make this memory shared as to not realloc
free_gpu_matrix(d_input);
arma::Mat<float> targets_slice = targets.rows(i*batch_size, (i+1) * batch_size - 1);
Raw_Matrix raw_targets = to_raw(targets_slice);
Raw_Matrix * d_targets = matrix_to_gpu(raw_targets);
backprop(num_trials, network.layer_sizes, d_network, d_targets, learning_rate, momentum);
free_gpu_matrix(d_targets);
}
network_to_cpu_free(d_network, raw_net);
update_from_raw(network, raw_net);
}
inline arma::Mat<float> gpu_predict(FeedForward_Network<activation, error>& network,
arma::Mat<float> inputs) {
network.resize_activation(inputs.n_rows);
Raw_FeedForward_Network<activation, error> raw_net = convert_to_raw(network);
Raw_FeedForward_Network<activation, error> * d_network = network_to_gpu(raw_net);
Raw_Matrix raw_inputs = to_raw(inputs);
Raw_Matrix * d_inputs = matrix_to_gpu(raw_inputs);
int num_trials = inputs.n_rows;
calculate_activation(num_trials, network.layer_sizes, d_network, d_inputs);
free_gpu_matrix(d_inputs);
network_to_cpu_free(d_network, raw_net);
update_from_raw(network, raw_net);
return network.activations.back();
}
double Perceptron::calculate_output(const Vector<double>& inputs, const Vector<double>& parameters) const
{
// Control sentence (if debug)
#ifdef __OPENNN_DEBUG__
const size_t inputs_size = inputs.size();
const size_t inputs_number = get_inputs_number();
if(inputs_size != inputs_number)
{
std::ostringstream buffer;
buffer << "OpenNN Exception: Perceptron class.\n"
<< "double calculate_output(const Vector<double>&, const Vector<double>&) const method.\n"
<< "Size of inputs must be equal to number of inputs.\n";
throw std::logic_error(buffer.str());
}
const size_t parameters_size = parameters.size();
const size_t parameters_number = count_parameters_number();
if(parameters_size != parameters_number)
{
std::ostringstream buffer;
buffer << "OpenNN Exception: Perceptron class.\n"
<< "double calculate_output(const Vector<double>&, const Vector<double>&) const method.\n"
<< "Size of potential parameters (" << parameters_size << ") must be equal to number of parameters (" << parameters_number << ").\n";
throw std::logic_error(buffer.str());
}
#endif
return(calculate_activation(calculate_combination(inputs, parameters)));
}
double Perceptron::calculate_output(const Vector<double>& input)
{
// Control sentence (if debug)
#ifdef _DEBUG
int size = input.get_size();
if(size != inputs_number)
{
std::cerr << "Flood Error: Perceptron class." << std::endl
<< "double calculate_output(const Vector<double>&) method." << std::endl
<< "Size must be equal to number of inputs." << std::endl;
exit(1);
}
#endif
// Calculate output
return(calculate_activation(calculate_combination(input)));
}
