def fit(self, attribute_list, class_list, row_sampler, col_sampler,
bin_structure):
# when we start to fit a tree, we first conduct row and column sampling
col_sampler.shuffle()
row_sampler.shuffle()
class_list.sampling(row_sampler.row_mask)
# then we create the root node, initialize histogram(Gradient sum and Hessian sum)
root_node = TreeNode(name=1,
depth=1,
feature_dim=attribute_list.feature_dim)
root_node.Grad_setter(class_list.grad.sum())
root_node.Hess_setter(class_list.hess.sum())
self.root = root_node
# every time a new node is created, we put it into self.name_to_node
self.name_to_node[root_node.name] = root_node
# put it into the alive_node, and fill the class_list, all data are assigned to root node initially
self.alive_nodes.append(root_node)
for i in range(class_list.dataset_size):
class_list.corresponding_tree_node[i] = root_node
# then build the tree util there is no alive tree_node to split
self.build(attribute_list, class_list, col_sampler, bin_structure)
self.clean_up()
