Method learn

src/RaceNet/Network.cpp:42–64

}

void Network::learn(std::vector<double> expectedOutput) {
    Y = Matrix<double>({expectedOutput}); // row matrix

    // Error E = 1/2 (expectedOutput - computedOutput)^2
    // Then, we need to calculate the partial derivative of E with respect to W and B

    // compute gradients
    dEdB[hiddenLayersCount] = H[hiddenLayersCount + 1].subtract(Y).multiply(
            H[hiddenLayersCount].dot(W[hiddenLayersCount]).add(B[hiddenLayersCount]).applyFunction(sigmoidePrime));
    for (int i = hiddenLayersCount - 1; i >= 0; i--) {
        dEdB[i] = dEdB[i + 1].dot(W[i + 1].transpose()).multiply(H[i].dot(W[i]).add(B[i]).applyFunction(sigmoidePrime));
    }

    for (int i = 0; i < hiddenLayersCount + 1; i++) {
        dEdW[i] = H[i].transpose().dot(dEdB[i]);
    }

    // update weights
    for (int i = 0; i < hiddenLayersCount + 1; i++) {
        W[i] = W[i].subtract(dEdW[i].multiply(learningRate));
        B[i] = B[i].subtract(dEdB[i].multiply(learningRate));
    }
}

void Network::printToFile(Matrix<double> &m, std::ofstream &file) {
    int h = m.getHeight();