Aluode/PerceptionLabPortable

1# Authors: The scikit-learn developers
2# SPDX-License-Identifier: BSD-3-Clause
3
4# See _tree.pyx for details.
5
6import numpy as np
7cimport numpy as cnp
8
9from ..utils._typedefs cimport float32_t, float64_t, intp_t, int32_t, uint8_t, uint32_t
10
11from ._splitter cimport Splitter
12from ._splitter cimport SplitRecord
13
14cdef struct Node:
15    # Base storage structure for the nodes in a Tree object
16
17    intp_t left_child                    # id of the left child of the node
18    intp_t right_child                   # id of the right child of the node
19    intp_t feature                       # Feature used for splitting the node
20    float64_t threshold                  # Threshold value at the node
21    float64_t impurity                   # Impurity of the node (i.e., the value of the criterion)
22    intp_t n_node_samples                # Number of samples at the node
23    float64_t weighted_n_node_samples    # Weighted number of samples at the node
24    uint8_t missing_go_to_left     # Whether features have missing values
25
26
27cdef struct ParentInfo:
28    # Structure to store information about the parent of a node
29    # This is passed to the splitter, to provide information about the previous split
30
31    float64_t lower_bound           # the lower bound of the parent's impurity
32    float64_t upper_bound           # the upper bound of the parent's impurity
33    float64_t impurity              # the impurity of the parent
34    intp_t n_constant_features      # the number of constant features found in parent
35
36cdef class Tree:
37    # The Tree object is a binary tree structure constructed by the
38    # TreeBuilder. The tree structure is used for predictions and
39    # feature importances.
40
41    # Input/Output layout
42    cdef public intp_t n_features        # Number of features in X
43    cdef intp_t* n_classes               # Number of classes in y[:, k]
44    cdef public intp_t n_outputs         # Number of outputs in y
45    cdef public intp_t max_n_classes     # max(n_classes)
46
47    # Inner structures: values are stored separately from node structure,
48    # since size is determined at runtime.
49    cdef public intp_t max_depth         # Max depth of the tree
50    cdef public intp_t node_count        # Counter for node IDs
51    cdef public intp_t capacity          # Capacity of tree, in terms of nodes
52    cdef Node* nodes                     # Array of nodes
53    cdef float64_t* value                # (capacity, n_outputs, max_n_classes) array of values
54    cdef intp_t value_stride             # = n_outputs * max_n_classes
55
56    # Methods
57    cdef intp_t _add_node(self, intp_t parent, bint is_left, bint is_leaf,
58                          intp_t feature, float64_t threshold, float64_t impurity,
59                          intp_t n_node_samples,
60                          float64_t weighted_n_node_samples,
61                          uint8_t missing_go_to_left) except -1 nogil
62    cdef int _resize(self, intp_t capacity) except -1 nogil
63    cdef int _resize_c(self, intp_t capacity=*) except -1 nogil
64
65    cdef cnp.ndarray _get_value_ndarray(self)
66    cdef cnp.ndarray _get_node_ndarray(self)
67
68    cpdef cnp.ndarray predict(self, object X)
69
70    cpdef cnp.ndarray apply(self, object X)
71    cdef cnp.ndarray _apply_dense(self, object X)
72    cdef cnp.ndarray _apply_sparse_csr(self, object X)
73
74    cpdef object decision_path(self, object X)
75    cdef object _decision_path_dense(self, object X)
76    cdef object _decision_path_sparse_csr(self, object X)
77
78    cpdef compute_node_depths(self)
79    cpdef compute_feature_importances(self, normalize=*)
80
81
82# =============================================================================
83# Tree builder
84# =============================================================================
85
86cdef class TreeBuilder:
87    # The TreeBuilder recursively builds a Tree object from training samples,
88    # using a Splitter object for splitting internal nodes and assigning
89    # values to leaves.
90    #
91    # This class controls the various stopping criteria and the node splitting
92    # evaluation order, e.g. depth-first or best-first.
93
94    cdef Splitter splitter              # Splitting algorithm
95
96    cdef intp_t min_samples_split       # Minimum number of samples in an internal node
97    cdef intp_t min_samples_leaf        # Minimum number of samples in a leaf
98    cdef float64_t min_weight_leaf         # Minimum weight in a leaf
99    cdef intp_t max_depth               # Maximal tree depth
100    cdef float64_t min_impurity_decrease   # Impurity threshold for early stopping
101
102    cpdef build(
103        self,
104        Tree tree,
105        object X,
106        const float64_t[:, ::1] y,
107        const float64_t[:] sample_weight=*,
108        const uint8_t[::1] missing_values_in_feature_mask=*,
109    )
110
111    cdef _check_input(
112        self,
113        object X,
114        const float64_t[:, ::1] y,
115        const float64_t[:] sample_weight,
116    )
117
118
119# =============================================================================
120# Tree pruning
121# =============================================================================
122
123# The private function allows any external caller to prune the tree and return
124# a new tree with the pruned nodes. The pruned tree is a new tree object.
125#
126# .. warning:: this function is not backwards compatible and may change without
127#              notice.
128cdef void _build_pruned_tree(
129    Tree tree,  # OUT
130    Tree orig_tree,
131    const uint8_t[:] leaves_in_subtree,
132    intp_t capacity
133)
134