Stacking-Based Ensemble Learning Method for the Recognition of the Pedestrian Crossing Intention

Table 1

Pseudocode of the stacking algorithm.

---

Input: training set S_train = {(x_1,y₁), (x_2,y₂),…, (x_m,y_m)};

  Base classifier: L₁, L_2, … L_T;

  Meta classifier: L (Bi-LSTM).

Process:

 for t = 1,2, …., T do

  h_t = L_t (S_train)% train the base classifier separately using the training set

 end for

 N = Ø; % create new datasets

 for i = 1, 2 m do

  for t = 1,2, …, T do

   z_it = h_t (x_i)% use the classifier h_t to test the validation set

  end for

  N = N∪{(z_i1, z_i2,… z_iT), y_i}

 end for

  h’ = L (N); % training a meta-classifier based on the Bi-LSTM algorithm with the newly combined dataset

Output: H(x) = h^’ (h₁ (x), h₂ (x)…, h_T (x))

---