| 24 | |
| 25 | |
| 26 | int main() |
| 27 | { |
| 28 | // The mlp takes column vectors as input and gives column vectors as output. The dlib::matrix |
| 29 | // object is used to represent the column vectors. So the first thing we do here is declare |
| 30 | // a convenient typedef for the matrix object we will be using. |
| 31 | |
| 32 | // This typedef declares a matrix with 2 rows and 1 column. It will be the |
| 33 | // object that contains each of our 2 dimensional samples. (Note that if you wanted |
| 34 | // more than 2 features in this vector you can simply change the 2 to something else) |
| 35 | typedef matrix<double, 2, 1> sample_type; |
| 36 | |
| 37 | |
| 38 | // make an instance of a sample matrix so we can use it below |
| 39 | sample_type sample; |
| 40 | |
| 41 | // Create a multi-layer perceptron network. This network has 2 nodes on the input layer |
| 42 | // (which means it takes column vectors of length 2 as input) and 5 nodes in the first |
| 43 | // hidden layer. Note that the other 4 variables in the mlp's constructor are left at |
| 44 | // their default values. |
| 45 | mlp::kernel_1a_c net(2,5); |
| 46 | |
| 47 | // Now let's put some data into our sample and train on it. We do this |
| 48 | // by looping over 41*41 points and labeling them according to their |
| 49 | // distance from the origin. |
| 50 | for (int i = 0; i < 1000; ++i) |
| 51 | { |
| 52 | for (int r = -20; r <= 20; ++r) |
| 53 | { |
| 54 | for (int c = -20; c <= 20; ++c) |
| 55 | { |
| 56 | sample(0) = r; |
| 57 | sample(1) = c; |
| 58 | |
| 59 | // if this point is less than 10 from the origin |
| 60 | if (sqrt((double)r*r + c*c) <= 10) |
| 61 | net.train(sample,1); |
| 62 | else |
| 63 | net.train(sample,0); |
| 64 | } |
| 65 | } |
| 66 | } |
| 67 | |
| 68 | // Now we have trained our mlp. Let's see how well it did. |
| 69 | // Note that if you run this program multiple times you will get different results. This |
| 70 | // is because the mlp network is randomly initialized. |
| 71 | |
| 72 | // each of these statements prints out the output of the network given a particular sample. |
| 73 | |
| 74 | sample(0) = 3.123; |
| 75 | sample(1) = 4; |
| 76 | cout << "This sample should be close to 1 and it is classified as a " << net(sample) << endl; |
| 77 | |
| 78 | sample(0) = 13.123; |
| 79 | sample(1) = 9.3545; |
| 80 | cout << "This sample should be close to 0 and it is classified as a " << net(sample) << endl; |
| 81 | |
| 82 | sample(0) = 13.123; |
| 83 | sample(1) = 0; |