Draft causal discovery GES implementation #1517
Draft
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Comment on lines
+215
to
+254
| @model_validator(mode="before") | ||
| def set_defaults(cls, values): | ||
| """ | ||
| Set default values for optional parameters. | ||
|
|
||
| Parameters | ||
| ---------- | ||
| values : dict | ||
| Dictionary of input values | ||
|
|
||
| Returns | ||
| ------- | ||
| dict | ||
| Updated values with defaults set | ||
| """ | ||
| # Ensure that no_descendants and no_ascendents are lists. | ||
| if values.get("no_descendants") is None: | ||
| values["no_descendants"] = [] | ||
| if values.get("no_ascendents") is None: | ||
| values["no_ascendents"] = [] | ||
| return values | ||
|
|
||
| def __init__(self, **data): | ||
| """ | ||
| Initialize the causal discovery model. | ||
|
|
||
| Parameters | ||
| ---------- | ||
| **data : dict | ||
| Keyword arguments matching class attributes | ||
| """ | ||
| super().__init__(**data) | ||
| # Process the data and initialize internal state. | ||
| self.data = self.data.drop(columns=["date"], errors="ignore") | ||
| self.nodes = self.data.columns.tolist() | ||
| self.graph = {node: [] for node in self.nodes} | ||
| self.no_descendants = set(self.no_descendants) | ||
| self.no_ascendents = set(self.no_ascendents) | ||
| self.local_score_cache = {} | ||
| self.inference_cache = {} |
There was a problem hiding this comment.
No need for any of this. Just use default_factory
| dot.render(output_filename, format=output_format, view=view) | ||
| return dot | ||
|
|
||
| def visualize( |
There was a problem hiding this comment.
How about to_graphviz to be aligned with the ModelBuilder method name
Comment on lines
+408
to
+420
| if parents: | ||
| intercept = pm.Normal(f"intercept_{node}", mu=0, sigma=1) | ||
| betas = pm.Normal(f"beta_{node}", mu=0, sigma=1, shape=len(parents)) | ||
| parent_data = self.data[parents].values | ||
| mu_node = intercept + pt.dot(parent_data, betas) | ||
| else: | ||
| intercept = pm.Normal(f"intercept_{node}", mu=0, sigma=1) | ||
| mu_node = intercept | ||
|
|
||
| sigma = pm.HalfNormal(f"sigma_{node}", sigma=1) | ||
| pm.Normal( | ||
| f"obs_{node}", mu=mu_node, sigma=sigma, observed=self.data[node].values | ||
| ) |
There was a problem hiding this comment.
Would the goal be to vectorize this across set of DAG combinations?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
BayesianCausalDiscoveryBasethat leverages Pydantic for parameter validation and encapsulates common functionality for Bayesian causal discovery.BayesianGreedySearchthat performs a greedy search with modular forward and backward phases.Related Issue
Checklist
pre-commit.ci autofixto auto-fix.📚 Documentation preview 📚: https://pymc-marketing--1517.org.readthedocs.build/en/1517/