Skip to content

Hex2VecEmbedder

Bases: CountEmbedder

Hex2Vec Embedder.

Source code in srai/embedders/hex2vec/embedder.py 
 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
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
class Hex2VecEmbedder(CountEmbedder):
    """Hex2Vec Embedder."""

    DEFAULT_ENCODER_SIZES = [150, 75, 50]

    def __init__(
        self,
        encoder_sizes: Optional[List[int]] = None,
        expected_output_features: Optional[List[str]] = None,
    ) -> None:
        """
        Initialize Hex2VecEmbedder.

        Args:
            encoder_sizes (Optional[List[int]], optional): Sizes of the encoder layers.
                The input layer size shouldn't be included - it's inferred from the data.
                The last element is the embedding size. Defaults to [150, 75, 50].
            expected_output_features (Optional[List[str]], optional): List of expected output
                features. Defaults to None.
        """
        super().__init__(expected_output_features)
        import_optional_dependencies(
            dependency_group="torch", modules=["torch", "pytorch_lightning"]
        )
        if encoder_sizes is None:
            encoder_sizes = Hex2VecEmbedder.DEFAULT_ENCODER_SIZES
        self._assert_encoder_sizes_correct(encoder_sizes)
        self._encoder_sizes = encoder_sizes
        self._model: Optional[Hex2VecModel] = None
        self._is_fitted = False

    def transform(
        self,
        regions_gdf: gpd.GeoDataFrame,
        features_gdf: gpd.GeoDataFrame,
        joint_gdf: gpd.GeoDataFrame,
    ) -> pd.DataFrame:
        """
        Create region embeddings.

        Args:
            regions_gdf (gpd.GeoDataFrame): Region indexes and geometries.
            features_gdf (gpd.GeoDataFrame): Feature indexes, geometries and feature values.
            joint_gdf (gpd.GeoDataFrame): Joiner result with region-feature multi-index.

        Returns:
            pd.DataFrame: Embedding and geometry index for each region in regions_gdf.

        Raises:
            ValueError: If features_gdf is empty and self.expected_output_features is not set.
            ValueError: If any of the gdfs index names is None.
            ValueError: If joint_gdf.index is not of type pd.MultiIndex or doesn't have 2 levels.
            ValueError: If index levels in gdfs don't overlap correctly.
        """
        import torch

        self._check_is_fitted()
        counts_df = self._get_raw_counts(regions_gdf, features_gdf, joint_gdf)
        counts_tensor = torch.from_numpy(counts_df.values)
        embeddings = self._model(counts_tensor).detach().numpy()  # type: ignore
        return pd.DataFrame(embeddings, index=counts_df.index)

    def fit(
        self,
        regions_gdf: gpd.GeoDataFrame,
        features_gdf: gpd.GeoDataFrame,
        joint_gdf: gpd.GeoDataFrame,
        neighbourhood: Neighbourhood[T],
        negative_sample_k_distance: int = 2,
        batch_size: int = 32,
        learning_rate: float = 0.001,
        trainer_kwargs: Optional[Dict[str, Any]] = None,
    ) -> None:
        """
        Fit the model to the data.

        Args:
            regions_gdf (gpd.GeoDataFrame): Region indexes and geometries.
            features_gdf (gpd.GeoDataFrame): Feature indexes, geometries and feature values.
            joint_gdf (gpd.GeoDataFrame): Joiner result with region-feature multi-index.
            neighbourhood (Neighbourhood[T]): The neighbourhood to use.
                Should be intialized with the same regions.
            negative_sample_k_distance (int, optional): When sampling negative samples,
                sample from a distance > k. Defaults to 2.
            batch_size (int, optional): Batch size. Defaults to 32.
            learning_rate (float, optional): Learning rate. Defaults to 0.001.
            trainer_kwargs (Optional[Dict[str, Any]], optional): Trainer kwargs. Defaults to None.

        Raises:
            ValueError: If features_gdf is empty and self.expected_output_features is not set.
            ValueError: If any of the gdfs index names is None.
            ValueError: If joint_gdf.index is not of type pd.MultiIndex or doesn't have 2 levels.
            ValueError: If index levels in gdfs don't overlap correctly.
            ValueError: If negative_sample_k_distance < 2.
        """
        import pytorch_lightning as pl
        from torch.utils.data import DataLoader

        trainer_kwargs = self._prepare_trainer_kwargs(trainer_kwargs)

        counts_df = self._get_raw_counts(regions_gdf, features_gdf, joint_gdf)
        num_features = len(counts_df.columns)
        self._model = Hex2VecModel(
            layer_sizes=[num_features, *self._encoder_sizes], learning_rate=learning_rate
        )
        dataset = NeighbourDataset(counts_df, neighbourhood, negative_sample_k_distance)
        dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True)

        trainer = pl.Trainer(**trainer_kwargs)
        trainer.fit(self._model, dataloader)
        self._is_fitted = True

    def fit_transform(
        self,
        regions_gdf: gpd.GeoDataFrame,
        features_gdf: gpd.GeoDataFrame,
        joint_gdf: gpd.GeoDataFrame,
        neighbourhood: Neighbourhood[T],
        negative_sample_k_distance: int = 2,
        batch_size: int = 32,
        learning_rate: float = 0.001,
        trainer_kwargs: Optional[Dict[str, Any]] = None,
    ) -> pd.DataFrame:
        """
        Fit the model to the data and return the embeddings.

        Args:
            regions_gdf (gpd.GeoDataFrame): Region indexes and geometries.
            features_gdf (gpd.GeoDataFrame): Feature indexes, geometries and feature values.
            joint_gdf (gpd.GeoDataFrame): Joiner result with region-feature multi-index.
            neighbourhood (Neighbourhood[T]): The neighbourhood to use.
                Should be intialized with the same regions.
            negative_sample_k_distance (int, optional): When sampling negative samples,
                sample from a distance > k. Defaults to 2.
            batch_size (int, optional): Batch size. Defaults to 32.
            learning_rate (float, optional): Learning rate. Defaults to 0.001.
            trainer_kwargs (Optional[Dict[str, Any]], optional): Trainer kwargs. Defaults to None.

        Returns:
            pd.DataFrame: Region embeddings.

        Raises:
            ValueError: If features_gdf is empty and self.expected_output_features is not set.
            ValueError: If any of the gdfs index names is None.
            ValueError: If joint_gdf.index is not of type pd.MultiIndex or doesn't have 2 levels.
            ValueError: If index levels in gdfs don't overlap correctly.
            ValueError: If negative_sample_k_distance < 2.
        """
        self.fit(
            regions_gdf,
            features_gdf,
            joint_gdf,
            neighbourhood,
            negative_sample_k_distance,
            batch_size,
            learning_rate,
            trainer_kwargs,
        )
        return self.transform(regions_gdf, features_gdf, joint_gdf)

    def _prepare_trainer_kwargs(self, trainer_kwargs: Optional[Dict[str, Any]]) -> Dict[str, Any]:
        if trainer_kwargs is None:
            trainer_kwargs = {}
        if "max_epochs" not in trainer_kwargs:
            trainer_kwargs["max_epochs"] = 3
        return trainer_kwargs

    def _get_raw_counts(
        self, regions_gdf: pd.DataFrame, features_gdf: pd.DataFrame, joint_gdf: pd.DataFrame
    ) -> pd.DataFrame:
        return super().transform(regions_gdf, features_gdf, joint_gdf).astype(np.float32)

    def _check_is_fitted(self) -> None:
        if not self._is_fitted or self._model is None:
            raise ModelNotFitException("Model not fitted. Call fit() or fit_transform() first.")

    def _assert_encoder_sizes_correct(self, encoder_sizes: List[int]) -> None:
        if len(encoder_sizes) < 1:
            raise ValueError("Encoder sizes must have at least one element - embedding size.")
        if any(size <= 0 for size in encoder_sizes):
            raise ValueError("Encoder sizes must be positive integers.")

__init__ 

__init__(encoder_sizes: Optional[List[int]] = None, expected_output_features: Optional[List[str]] = None) -> None

Initialize Hex2VecEmbedder.

PARAMETER DESCRIPTION
encoder_sizes

Sizes of the encoder layers. The input layer size shouldn't be included - it's inferred from the data. The last element is the embedding size. Defaults to [150, 75, 50].

TYPE: Optional[List[int]] DEFAULT: None

expected_output_features

List of expected output features. Defaults to None.

TYPE: Optional[List[str]] DEFAULT: None

Source code in srai/embedders/hex2vec/embedder.py 
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
def __init__(
    self,
    encoder_sizes: Optional[List[int]] = None,
    expected_output_features: Optional[List[str]] = None,
) -> None:
    """
    Initialize Hex2VecEmbedder.

    Args:
        encoder_sizes (Optional[List[int]], optional): Sizes of the encoder layers.
            The input layer size shouldn't be included - it's inferred from the data.
            The last element is the embedding size. Defaults to [150, 75, 50].
        expected_output_features (Optional[List[str]], optional): List of expected output
            features. Defaults to None.
    """
    super().__init__(expected_output_features)
    import_optional_dependencies(
        dependency_group="torch", modules=["torch", "pytorch_lightning"]
    )
    if encoder_sizes is None:
        encoder_sizes = Hex2VecEmbedder.DEFAULT_ENCODER_SIZES
    self._assert_encoder_sizes_correct(encoder_sizes)
    self._encoder_sizes = encoder_sizes
    self._model: Optional[Hex2VecModel] = None
    self._is_fitted = False

fit 

fit(regions_gdf: gpd.GeoDataFrame, features_gdf: gpd.GeoDataFrame, joint_gdf: gpd.GeoDataFrame, neighbourhood: Neighbourhood[T], negative_sample_k_distance: int = 2, batch_size: int = 32, learning_rate: float = 0.001, trainer_kwargs: Optional[Dict[str, Any]] = None) -> None

Fit the model to the data.

PARAMETER DESCRIPTION
regions_gdf

Region indexes and geometries.

TYPE: gpd.GeoDataFrame

features_gdf

Feature indexes, geometries and feature values.

TYPE: gpd.GeoDataFrame

joint_gdf

Joiner result with region-feature multi-index.

TYPE: gpd.GeoDataFrame

neighbourhood

The neighbourhood to use. Should be intialized with the same regions.

TYPE: Neighbourhood[T]

negative_sample_k_distance

When sampling negative samples, sample from a distance > k. Defaults to 2.

TYPE: int DEFAULT: 2

batch_size

Batch size. Defaults to 32.

TYPE: int DEFAULT: 32

learning_rate

Learning rate. Defaults to 0.001.

TYPE: float DEFAULT: 0.001

trainer_kwargs

Trainer kwargs. Defaults to None.

TYPE: Optional[Dict[str, Any]] DEFAULT: None

RAISES DESCRIPTION
ValueError

If features_gdf is empty and self.expected_output_features is not set.
ValueError

If any of the gdfs index names is None.
ValueError

If joint_gdf.index is not of type pd.MultiIndex or doesn't have 2 levels.
ValueError

If index levels in gdfs don't overlap correctly.
ValueError

If negative_sample_k_distance < 2.
Source code in srai/embedders/hex2vec/embedder.py 
 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
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
def fit(
    self,
    regions_gdf: gpd.GeoDataFrame,
    features_gdf: gpd.GeoDataFrame,
    joint_gdf: gpd.GeoDataFrame,
    neighbourhood: Neighbourhood[T],
    negative_sample_k_distance: int = 2,
    batch_size: int = 32,
    learning_rate: float = 0.001,
    trainer_kwargs: Optional[Dict[str, Any]] = None,
) -> None:
    """
    Fit the model to the data.

    Args:
        regions_gdf (gpd.GeoDataFrame): Region indexes and geometries.
        features_gdf (gpd.GeoDataFrame): Feature indexes, geometries and feature values.
        joint_gdf (gpd.GeoDataFrame): Joiner result with region-feature multi-index.
        neighbourhood (Neighbourhood[T]): The neighbourhood to use.
            Should be intialized with the same regions.
        negative_sample_k_distance (int, optional): When sampling negative samples,
            sample from a distance > k. Defaults to 2.
        batch_size (int, optional): Batch size. Defaults to 32.
        learning_rate (float, optional): Learning rate. Defaults to 0.001.
        trainer_kwargs (Optional[Dict[str, Any]], optional): Trainer kwargs. Defaults to None.

    Raises:
        ValueError: If features_gdf is empty and self.expected_output_features is not set.
        ValueError: If any of the gdfs index names is None.
        ValueError: If joint_gdf.index is not of type pd.MultiIndex or doesn't have 2 levels.
        ValueError: If index levels in gdfs don't overlap correctly.
        ValueError: If negative_sample_k_distance < 2.
    """
    import pytorch_lightning as pl
    from torch.utils.data import DataLoader

    trainer_kwargs = self._prepare_trainer_kwargs(trainer_kwargs)

    counts_df = self._get_raw_counts(regions_gdf, features_gdf, joint_gdf)
    num_features = len(counts_df.columns)
    self._model = Hex2VecModel(
        layer_sizes=[num_features, *self._encoder_sizes], learning_rate=learning_rate
    )
    dataset = NeighbourDataset(counts_df, neighbourhood, negative_sample_k_distance)
    dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True)

    trainer = pl.Trainer(**trainer_kwargs)
    trainer.fit(self._model, dataloader)
    self._is_fitted = True

fit_transform 

fit_transform(regions_gdf: gpd.GeoDataFrame, features_gdf: gpd.GeoDataFrame, joint_gdf: gpd.GeoDataFrame, neighbourhood: Neighbourhood[T], negative_sample_k_distance: int = 2, batch_size: int = 32, learning_rate: float = 0.001, trainer_kwargs: Optional[Dict[str, Any]] = None) -> pd.DataFrame

Fit the model to the data and return the embeddings.

PARAMETER DESCRIPTION
regions_gdf

Region indexes and geometries.

TYPE: gpd.GeoDataFrame

features_gdf

Feature indexes, geometries and feature values.

TYPE: gpd.GeoDataFrame

joint_gdf

Joiner result with region-feature multi-index.

TYPE: gpd.GeoDataFrame

neighbourhood

The neighbourhood to use. Should be intialized with the same regions.

TYPE: Neighbourhood[T]

negative_sample_k_distance

When sampling negative samples, sample from a distance > k. Defaults to 2.

TYPE: int DEFAULT: 2

batch_size

Batch size. Defaults to 32.

TYPE: int DEFAULT: 32

learning_rate

Learning rate. Defaults to 0.001.

TYPE: float DEFAULT: 0.001

trainer_kwargs

Trainer kwargs. Defaults to None.

TYPE: Optional[Dict[str, Any]] DEFAULT: None

RETURNS DESCRIPTION
pd.DataFrame

pd.DataFrame: Region embeddings.
RAISES DESCRIPTION
ValueError

If features_gdf is empty and self.expected_output_features is not set.
ValueError

If any of the gdfs index names is None.
ValueError

If joint_gdf.index is not of type pd.MultiIndex or doesn't have 2 levels.
ValueError

If index levels in gdfs don't overlap correctly.
ValueError

If negative_sample_k_distance < 2.
Source code in srai/embedders/hex2vec/embedder.py 
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
def fit_transform(
    self,
    regions_gdf: gpd.GeoDataFrame,
    features_gdf: gpd.GeoDataFrame,
    joint_gdf: gpd.GeoDataFrame,
    neighbourhood: Neighbourhood[T],
    negative_sample_k_distance: int = 2,
    batch_size: int = 32,
    learning_rate: float = 0.001,
    trainer_kwargs: Optional[Dict[str, Any]] = None,
) -> pd.DataFrame:
    """
    Fit the model to the data and return the embeddings.

    Args:
        regions_gdf (gpd.GeoDataFrame): Region indexes and geometries.
        features_gdf (gpd.GeoDataFrame): Feature indexes, geometries and feature values.
        joint_gdf (gpd.GeoDataFrame): Joiner result with region-feature multi-index.
        neighbourhood (Neighbourhood[T]): The neighbourhood to use.
            Should be intialized with the same regions.
        negative_sample_k_distance (int, optional): When sampling negative samples,
            sample from a distance > k. Defaults to 2.
        batch_size (int, optional): Batch size. Defaults to 32.
        learning_rate (float, optional): Learning rate. Defaults to 0.001.
        trainer_kwargs (Optional[Dict[str, Any]], optional): Trainer kwargs. Defaults to None.

    Returns:
        pd.DataFrame: Region embeddings.

    Raises:
        ValueError: If features_gdf is empty and self.expected_output_features is not set.
        ValueError: If any of the gdfs index names is None.
        ValueError: If joint_gdf.index is not of type pd.MultiIndex or doesn't have 2 levels.
        ValueError: If index levels in gdfs don't overlap correctly.
        ValueError: If negative_sample_k_distance < 2.
    """
    self.fit(
        regions_gdf,
        features_gdf,
        joint_gdf,
        neighbourhood,
        negative_sample_k_distance,
        batch_size,
        learning_rate,
        trainer_kwargs,
    )
    return self.transform(regions_gdf, features_gdf, joint_gdf)

transform 

transform(regions_gdf: gpd.GeoDataFrame, features_gdf: gpd.GeoDataFrame, joint_gdf: gpd.GeoDataFrame) -> pd.DataFrame

Create region embeddings.

PARAMETER DESCRIPTION
regions_gdf

Region indexes and geometries.

TYPE: gpd.GeoDataFrame

features_gdf

Feature indexes, geometries and feature values.

TYPE: gpd.GeoDataFrame

joint_gdf

Joiner result with region-feature multi-index.

TYPE: gpd.GeoDataFrame

RETURNS DESCRIPTION
pd.DataFrame

pd.DataFrame: Embedding and geometry index for each region in regions_gdf.
RAISES DESCRIPTION
ValueError

If features_gdf is empty and self.expected_output_features is not set.
ValueError

If any of the gdfs index names is None.
ValueError

If joint_gdf.index is not of type pd.MultiIndex or doesn't have 2 levels.
ValueError

If index levels in gdfs don't overlap correctly.
Source code in srai/embedders/hex2vec/embedder.py 
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
def transform(
    self,
    regions_gdf: gpd.GeoDataFrame,
    features_gdf: gpd.GeoDataFrame,
    joint_gdf: gpd.GeoDataFrame,
) -> pd.DataFrame:
    """
    Create region embeddings.

    Args:
        regions_gdf (gpd.GeoDataFrame): Region indexes and geometries.
        features_gdf (gpd.GeoDataFrame): Feature indexes, geometries and feature values.
        joint_gdf (gpd.GeoDataFrame): Joiner result with region-feature multi-index.

    Returns:
        pd.DataFrame: Embedding and geometry index for each region in regions_gdf.

    Raises:
        ValueError: If features_gdf is empty and self.expected_output_features is not set.
        ValueError: If any of the gdfs index names is None.
        ValueError: If joint_gdf.index is not of type pd.MultiIndex or doesn't have 2 levels.
        ValueError: If index levels in gdfs don't overlap correctly.
    """
    import torch

    self._check_is_fitted()
    counts_df = self._get_raw_counts(regions_gdf, features_gdf, joint_gdf)
    counts_tensor = torch.from_numpy(counts_df.values)
    embeddings = self._model(counts_tensor).detach().numpy()  # type: ignore
    return pd.DataFrame(embeddings, index=counts_df.index)