gpmp::ml::Trainers Class Reference

Training Algorithms. More...

#include <trainers.hpp>

Public Member Functions
std::vector< double >	gradientdesc (const std::vector< std::vector< double >> &X, const std::vector< double > &y, double alpha, int num_iters)
	Perform gradient descent for linear regression. More...
std::vector< double >	stoch_gradientdesc (const std::vector< std::vector< double >> &X, const std::vector< double > &y, double alpha, int num_iters)
	Perform stochastic gradient descent for linear regression. More...
std::vector< double >	minibatch_gradientdesc (const std::vector< std::vector< double >> &X, const std::vector< double > &y, double alpha, int num_iters, int batch_size)
	Perform mini-batch gradient descent for linear regression. More...

Detailed Description

Training Algorithms.

Examples

mlp.cpp.

Definition at line 54 of file trainers.hpp.

Member Function Documentation

gradientdesc()

std::vector< double > gpmp::ml::Trainers::gradientdesc	(	const std::vector< std::vector< double >> &	X,
		const std::vector< double > &	y,
		double	alpha,
		int	num_iters
	)

Perform gradient descent for linear regression.

Given features X, target y, learning rate alpha, and number of iterations num_iters, this function optimizes the parameters theta using gradient descent

Parameters

X	Features matrix (each row represents a training example)
y	Target vector
alpha	Learning rate
num_iters	Number of iterations

Returns

std::vector<double> Optimized parameters theta

Definition at line 37 of file trainers.cpp.

                                                {
    int m = X.size();                  // Number of training examples
    int n = X[0].size();               // Number of features
    std::vector<double> theta(n, 0.0); // Initialize parameters
 
    // Perform gradient descent
    for (int iter = 0; iter < num_iters; iter++) {
        std::vector<double> delta_theta(n,
                                        0.0); // Initialize change in parameters
 
        // Compute predictions and errors
        for (int i = 0; i < m; i++) {
            double prediction = 0;
            for (int j = 0; j < n; j++) {
                prediction += theta[j] * X[i][j];
            }
            double error = prediction - y[i];
 
            // Update delta_theta
            for (int j = 0; j < n; j++) {
                delta_theta[j] += error * X[i][j];
            }
        }
 
        // Update parameters
        for (int j = 0; j < n; j++) {
            theta[j] -= (alpha / m) * delta_theta[j];
        }
    }
 
    return theta;
}

minibatch_gradientdesc()

std::vector< double > gpmp::ml::Trainers::minibatch_gradientdesc	(	const std::vector< std::vector< double >> &	X,
		const std::vector< double > &	y,
		double	alpha,
		int	num_iters,
		int	batch_size
	)

Perform mini-batch gradient descent for linear regression.

Given features X, target y, learning rate alpha, number of iterations num_iters, and batch size batch_size, this function optimizes the parameters theta using mini-batch gradient descent

Parameters

X	Features matrix (each row represents a training example)
y	Target vector
alpha	Learning rate
num_iters	Number of iterations
batch_size	Size of mini-batch

Returns

std::vector<double> Optimized parameters theta

Definition at line 103 of file trainers.cpp.

                    {
    int m = X.size();                  // Number of training examples
    int n = X[0].size();               // Number of features
    std::vector<double> theta(n, 0.0); // Initialize parameters
 
    std::default_random_engine generator;
    std::uniform_int_distribution<int> distribution(0, m - 1);
 
    for (int iter = 0; iter < num_iters; iter++) {
        // Select random batch indices
        std::vector<int> batch_indices;
        for (int i = 0; i < batch_size; i++) {
            batch_indices.push_back(distribution(generator));
        }
 
        // Compute gradient for the batch
        std::vector<double> delta_theta(n, 0.0);
        for (int idx : batch_indices) {
            double prediction = 0;
            for (int j = 0; j < n; j++) {
                prediction += theta[j] * X[idx][j];
            }
            double error = prediction - y[idx];
 
            for (int j = 0; j < n; j++) {
                delta_theta[j] += (error * X[idx][j]);
            }
        }
 
        // Update parameters using the gradient of the batch
        for (int j = 0; j < n; j++) {
            theta[j] -= (alpha / batch_size) * delta_theta[j];
        }
    }
 
    return theta;
}

stoch_gradientdesc()

std::vector< double > gpmp::ml::Trainers::stoch_gradientdesc	(	const std::vector< std::vector< double >> &	X,
		const std::vector< double > &	y,
		double	alpha,
		int	num_iters
	)

Perform stochastic gradient descent for linear regression.

Given features X, target y, learning rate alpha, and number of iterations num_iters, this function optimizes the parameters theta using stochastic gradient descent

Parameters

X	Features matrix (each row represents a training example)
y	Target vector
alpha	Learning rate
num_iters	Number of iterations

Returns

std::vector<double> Optimized parameters theta

Definition at line 73 of file trainers.cpp.

                   {
    int m = X.size();                  // Number of training examples
    int n = X[0].size();               // Number of features
    std::vector<double> theta(n, 0.0); // Initialize parameters
 
    std::default_random_engine generator;
    std::uniform_int_distribution<int> distribution(0, m - 1);
 
    for (int iter = 0; iter < num_iters; iter++) {
        int random_index =
            distribution(generator); // Choose a random training example
        double prediction = 0;
 
        for (int j = 0; j < n; j++) {
            prediction += theta[j] * X[random_index][j];
        }
 
        double error = prediction - y[random_index];
 
        for (int j = 0; j < n; j++) {
            theta[j] -= (alpha * error * X[random_index][j]);
        }
    }
 
    return theta;
}

---

The documentation for this class was generated from the following files:

• include/openGPMP/ml/trainers.hpp
• modules/ml/trainers.cpp