Skip to content

Graph Algorithms

Graph Algorithms

WIP: These classes are currently actively being developed and are subject to change in both API and functionality over time.

Graph Algorithms

Docs TBD

ProximityGraph

Build a proximity graph of the nearest neighbors based on feature space.

Source code in src/sageworks/algorithms/graph/light/proximity_graph.py 
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
class ProximityGraph:
    """
    Build a proximity graph of the nearest neighbors based on feature space.
    """

    def __init__(self, n_neighbors: int = 5):
        """Initialize the ProximityGraph with the specified parameters.

        Args:
            n_neighbors (int): Number of neighbors to consider (default: 5)
        """
        self.n_neighbors = n_neighbors
        self.nx_graph = nx.Graph()

    def build_graph(
        self,
        df: pd.DataFrame,
        features: list,
        id_column: str,
        target: str,
        store_features=True,
    ) -> nx.Graph:
        """
        Processes the input DataFrame and builds a proximity graph.

        Args:
            df (pd.DataFrame): The input DataFrame containing feature columns.
            features (list): List of feature column names to be used for building the proximity graph.
            id_column (str): Name of the ID column in the DataFrame.
            target (str): Name of the target column in the DataFrame.
            store_features (bool): Whether to store the features as node attributes (default: True).

        Returns:
            nx.Graph: The proximity graph as a NetworkX graph.
        """
        # Drop NaNs from the DataFrame using the provided utility
        df = drop_nans(df)

        # Initialize FeatureSpaceProximity with the input DataFrame and the specified features
        knn_spider = FeatureSpaceProximity(
            df,
            features=features,
            id_column=id_column,
            target=target,
            neighbors=self.n_neighbors,
        )

        # Use FeatureSpaceProximity to get all neighbor indices and distances
        indices, distances = knn_spider.get_neighbor_indices_and_distances()

        # Compute max distance for scaling (to [0, 1])
        max_distance = distances.max()

        # Initialize an empty graph
        self.nx_graph = nx.Graph()

        # Use the ID column for node IDs instead of relying on the DataFrame index
        node_ids = df[id_column].values

        # Add nodes with their features as attributes using the ID column
        for node_id in node_ids:
            if store_features:
                self.nx_graph.add_node(
                    node_id, **df[df[id_column] == node_id].iloc[0].to_dict()
                )  # Use .iloc[0] for correct node attributes
            else:
                self.nx_graph.add_node(node_id)

        # Add edges with weights based on inverse distance
        for i, neighbors in enumerate(indices):
            one_edge_added = False
            for j, neighbor_idx in enumerate(neighbors):
                if i != neighbor_idx:
                    # Compute the weight of the edge (inverse of distance)
                    weight = 1.0 - (distances[i][j] / max_distance)  # Scale to [0, 1]

                    # Map back to the ID column instead of the DataFrame index
                    src_node = node_ids[i]
                    dst_node = node_ids[neighbor_idx]

                    # Add the edge to the graph (if the weight is greater than 0.1)
                    if weight > 0.1 or not one_edge_added:
                        self.nx_graph.add_edge(src_node, dst_node, weight=weight)
                        one_edge_added = True

        # Return the NetworkX graph
        return self.nx_graph

    def get_neighborhood(self, node_id: Union[str, int], radius: int = 1) -> nx.Graph:
        """
        Get a subgraph containing nodes within a given number of hops around a specific node.

        Args:
            node_id: The ID of the node to center the neighborhood around.
            radius: The number of hops to consider around the node (default: 1).

        Returns:
            nx.Graph: A subgraph containing the specified neighborhood.
        """
        # Use NetworkX's ego_graph to extract the neighborhood within the given radius
        if node_id in self.nx_graph:
            return nx.ego_graph(self.nx_graph, node_id, radius=radius)
        else:
            raise ValueError(f"Node ID '{node_id}' not found in the graph.")

__init__(n_neighbors=5)

Initialize the ProximityGraph with the specified parameters.

Parameters:

Name Type Description Default
n_neighbors int

Number of neighbors to consider (default: 5)

 5
Source code in src/sageworks/algorithms/graph/light/proximity_graph.py 
15
16
17
18
19
20
21
22
def __init__(self, n_neighbors: int = 5):
    """Initialize the ProximityGraph with the specified parameters.

    Args:
        n_neighbors (int): Number of neighbors to consider (default: 5)
    """
    self.n_neighbors = n_neighbors
    self.nx_graph = nx.Graph()

build_graph(df, features, id_column, target, store_features=True)

Processes the input DataFrame and builds a proximity graph.

Parameters:

Name Type Description Default
df DataFrame

The input DataFrame containing feature columns.

 required
features list

List of feature column names to be used for building the proximity graph.

 required
id_column str

Name of the ID column in the DataFrame.

 required
target str

Name of the target column in the DataFrame.

 required
store_features bool

Whether to store the features as node attributes (default: True).

 True

Returns:

Type Description
Graph

nx.Graph: The proximity graph as a NetworkX graph.
Source code in src/sageworks/algorithms/graph/light/proximity_graph.py 
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
def build_graph(
    self,
    df: pd.DataFrame,
    features: list,
    id_column: str,
    target: str,
    store_features=True,
) -> nx.Graph:
    """
    Processes the input DataFrame and builds a proximity graph.

    Args:
        df (pd.DataFrame): The input DataFrame containing feature columns.
        features (list): List of feature column names to be used for building the proximity graph.
        id_column (str): Name of the ID column in the DataFrame.
        target (str): Name of the target column in the DataFrame.
        store_features (bool): Whether to store the features as node attributes (default: True).

    Returns:
        nx.Graph: The proximity graph as a NetworkX graph.
    """
    # Drop NaNs from the DataFrame using the provided utility
    df = drop_nans(df)

    # Initialize FeatureSpaceProximity with the input DataFrame and the specified features
    knn_spider = FeatureSpaceProximity(
        df,
        features=features,
        id_column=id_column,
        target=target,
        neighbors=self.n_neighbors,
    )

    # Use FeatureSpaceProximity to get all neighbor indices and distances
    indices, distances = knn_spider.get_neighbor_indices_and_distances()

    # Compute max distance for scaling (to [0, 1])
    max_distance = distances.max()

    # Initialize an empty graph
    self.nx_graph = nx.Graph()

    # Use the ID column for node IDs instead of relying on the DataFrame index
    node_ids = df[id_column].values

    # Add nodes with their features as attributes using the ID column
    for node_id in node_ids:
        if store_features:
            self.nx_graph.add_node(
                node_id, **df[df[id_column] == node_id].iloc[0].to_dict()
            )  # Use .iloc[0] for correct node attributes
        else:
            self.nx_graph.add_node(node_id)

    # Add edges with weights based on inverse distance
    for i, neighbors in enumerate(indices):
        one_edge_added = False
        for j, neighbor_idx in enumerate(neighbors):
            if i != neighbor_idx:
                # Compute the weight of the edge (inverse of distance)
                weight = 1.0 - (distances[i][j] / max_distance)  # Scale to [0, 1]

                # Map back to the ID column instead of the DataFrame index
                src_node = node_ids[i]
                dst_node = node_ids[neighbor_idx]

                # Add the edge to the graph (if the weight is greater than 0.1)
                if weight > 0.1 or not one_edge_added:
                    self.nx_graph.add_edge(src_node, dst_node, weight=weight)
                    one_edge_added = True

    # Return the NetworkX graph
    return self.nx_graph

get_neighborhood(node_id, radius=1)

Get a subgraph containing nodes within a given number of hops around a specific node.

Parameters:

Name Type Description Default
node_id Union[str, int]

The ID of the node to center the neighborhood around.

 required
radius int

The number of hops to consider around the node (default: 1).

 1

Returns:

Type Description
Graph

nx.Graph: A subgraph containing the specified neighborhood.
Source code in src/sageworks/algorithms/graph/light/proximity_graph.py 
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
def get_neighborhood(self, node_id: Union[str, int], radius: int = 1) -> nx.Graph:
    """
    Get a subgraph containing nodes within a given number of hops around a specific node.

    Args:
        node_id: The ID of the node to center the neighborhood around.
        radius: The number of hops to consider around the node (default: 1).

    Returns:
        nx.Graph: A subgraph containing the specified neighborhood.
    """
    # Use NetworkX's ego_graph to extract the neighborhood within the given radius
    if node_id in self.nx_graph:
        return nx.ego_graph(self.nx_graph, node_id, radius=radius)
    else:
        raise ValueError(f"Node ID '{node_id}' not found in the graph.")

Questions?

The SuperCowPowers team is happy to answer any questions you may have about AWS and SageWorks. Please contact us at sageworks@supercowpowers.com or on chat us up on Discord