Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- function [grad_W, grad_b] = Compute_Gradients(X, Y, P, W, lambda)
- d = size(X, 1);
- N = size(X, 2);
- n_b = 1;
- ones_n_b = double(zeros(n_b, 1) + 1);
- random_data_indices = randsample(N, n_b, true);
- X_Batch = X(:, random_data_indices)
- Y_Batch = Y(:, random_data_indices)
- P_Batch = P(:, random_data_indices)
- G_Batch = -(Y_Batch - P_Batch)
- grad_W = (1 / n_b) * G_Batch * X_Batch' + 2 * lambda * W;
- grad_b = (1 / n_b) * G_Batch * ones_n_b;
- end
- function P = Evaluate_Classifier(X, W, b)
- N = size(X, 2);
- K = size(W, 1);
- P = zeros(K, N);
- for column = 1:N
- s = W * X(:,column) + b;
- p = exp(s);
- p = p / sum(p);
- P(:,column) = p;
- end
- end
Advertisement
