/* -------------------------------------------------------------------
			    Problem set #5

			     Main program
------------------------------------------------------------------- */

#include "network.hh"

main(int argc, char* argv[])
{
  // Accept data file as parameter on command line, complain if not there:
  if (argc != 2) {
    cerr << "Usage: " << argv[0] << " <datafile>\n";
    exit(0);
  }

  // Declare matrices input_data and output_data as zero-dimensional
  // The routine read_data below will resize them to the proper size
  matrix input_data(0, 0), output_data(0, 0);

  // Read data from file into matrices input_data and output_data
  read_data(argv[1], input_data, output_data);

  // Set up useful variables
  int input_dimensions = input_data.columns();
  int output_dimensions = output_data.columns();
  int training_examples = input_data.rows();

  // Call this to initialize our model so it can set up weights
  initialize(input_dimensions, output_dimensions);

  // Randomize the weights of our model from -scale to +scale 
  // scale = 1 in the call below:
  randomize_weights(1);

  // An example of how to read a parameter from the 'parameters' file.
  // You can read doubles and ints from the parameter file
  // read_parameter is called with the parameter name as it appears in
  // the parameter file, the variable to set, and a default value in
  // case the parameter doesn't appear in the file
  double learning_rate;
  read_parameter("mu", learning_rate, .1);

  // A useful method of both vector and matrix is zero(), which zeros
  // out all elements.  Call it with something like v.zero();

  // Now you must write the gradient descent code, using the gradient
  // descent algorithm of your choice.  You will use the functions
  // you wrote in one_layer.cc and two_layer.cc.

  // Your code here
}