embiggen/embedders/ensmallen_embedders/walklets_skipgram.py
"""Module providing WalkletsSkipGram model implementation."""
from typing import Optional, Dict, Any
from embiggen.embedders.ensmallen_embedders.walklets import WalkletsEnsmallen
class WalkletsSkipGramEnsmallen(WalkletsEnsmallen):
"""Class providing WalkletsSkipGram implemeted in Rust from Ensmallen."""
def __init__(
self,
embedding_size: int = 100,
epochs: int = 30,
clipping_value: float = 6.0,
number_of_negative_samples: int = 10,
walk_length: int = 128,
iterations: int = 10,
window_size: int = 4,
return_weight: float = 1.0,
explore_weight: float = 1.0,
max_neighbours: Optional[int] = 100,
learning_rate: float = 0.01,
learning_rate_decay: float = 0.9,
central_nodes_embedding_path: Optional[str] = None,
contextual_nodes_embedding_path: Optional[str] = None,
normalize_by_degree: bool = False,
stochastic_downsample_by_degree: Optional[bool] = False,
normalize_learning_rate_by_degree: Optional[bool] = False,
use_scale_free_distribution: Optional[bool] = True,
random_state: int = 42,
dtype: str = "f32",
ring_bell: bool = False,
enable_cache: bool = False
):
"""Create new abstract Node2Vec method.
Parameters
--------------------------
embedding_size: int = 100
Dimension of the embedding.
epochs: int = 30
Number of epochs to train the model for.
clipping_value: float = 6.0
Value at which we clip the dot product, mostly for numerical stability issues.
By default, `6.0`, where the loss is already close to zero.
number_of_negative_samples: int = 10
The number of negative classes to randomly sample per batch.
This single sample of negative classes is evaluated for each element in the batch.
walk_length: int = 128
Maximal length of the walks.
iterations: int = 10
Number of iterations of the single walks.
window_size: int = 4
Window size for the local context.
On the borders the window size is trimmed.
return_weight: float = 1.0
Weight on the probability of returning to the same node the walk just came from
Having this higher tends the walks to be
more like a Breadth-First Search.
Having this very high (> 2) makes search very local.
Equal to the inverse of p in the Node2Vec paper.
explore_weight: float = 1.0
Weight on the probability of visiting a neighbor node
to the one we're coming from in the random walk
Having this higher tends the walks to be
more like a Depth-First Search.
Having this very high makes search more outward.
Having this very low makes search very local.
Equal to the inverse of q in the Node2Vec paper.
max_neighbours: Optional[int] = 100
Number of maximum neighbours to consider when using approximated walks.
By default, None, we execute exact random walks.
This is mainly useful for graphs containing nodes with high degrees.
learning_rate: float = 0.01
The learning rate to use to train the Node2Vec model. By default 0.01.
learning_rate_decay: float = 0.9
Factor to reduce the learning rate for at each epoch. By default 0.9.
central_nodes_embedding_path: Optional[str] = None
Path where to mmap and store the central nodes embedding.
If provided, we expect the path to contain the substring `{window_size}` which
will be replaced with the i-th window size embedding that is being computed.
This is necessary to embed large graphs whose embedding will not
fit into the available main memory.
contextual_nodes_embedding_path: Optional[str] = None
Path where to mmap and store the central nodes embedding.
If provided, we expect the path to contain the substring `{window_size}` which
will be replaced with the i-th window size embedding that is being computed.
This is necessary to embed large graphs whose embedding will not
fit into the available main memory.
normalize_by_degree: bool = False
Whether to normalize the random walk by the node degree
of the destination node degrees.
stochastic_downsample_by_degree: Optional[bool] = False
Randomly skip samples with probability proportional to the degree of the central node. By default false.
normalize_learning_rate_by_degree: Optional[bool] = False
Divide the learning rate by the degree of the central node. By default false.
use_scale_free_distribution: Optional[bool] = True
Sample negatives proportionally to their degree. By default true.
dtype: str = "f32"
The data type to be employed, by default f32.
random_state: int = 42
The random state to reproduce the training sequence.
ring_bell: bool = False,
Whether to play a sound when embedding completes.
enable_cache: bool = False
Whether to enable the cache, that is to
store the computed embedding.
"""
super().__init__(
embedding_size=embedding_size,
epochs=epochs,
clipping_value=clipping_value,
number_of_negative_samples=number_of_negative_samples,
walk_length=walk_length,
iterations=iterations,
window_size=window_size,
return_weight=return_weight,
explore_weight=explore_weight,
max_neighbours=max_neighbours,
learning_rate=learning_rate,
learning_rate_decay=learning_rate_decay,
central_nodes_embedding_path=central_nodes_embedding_path,
contextual_nodes_embedding_path=contextual_nodes_embedding_path,
normalize_by_degree=normalize_by_degree,
stochastic_downsample_by_degree=stochastic_downsample_by_degree,
normalize_learning_rate_by_degree=normalize_learning_rate_by_degree,
use_scale_free_distribution=use_scale_free_distribution,
dtype=dtype,
random_state=random_state,
ring_bell=ring_bell,
enable_cache=enable_cache
)
def parameters(self) -> Dict[str, Any]:
"""Returns parameters for smoke test."""
removed = [
"alpha",
]
return dict(
**{
key: value
for key, value in super().parameters().items()
if key not in removed
}
)
@classmethod
def model_name(cls) -> str:
"""Returns name of the model."""
return "Walklets SkipGram"