[MNT, ENH, DOC] Rework similarity search (#2473)

* WIP remake module structure

* Update _brute_force.py

* Update test__commons.py

* WIP mock and test

* Add test for base subsequence

* Fix subsequence_search tests

* debug brute force mp

* more debug of subsequence tests

* more debug of subsequence tests

* Add functional LSH neighbors

* add notebook for sim search tasks

* Updated series similarity search

* Fix mistake addition in transformers and fix base classes

* Fix registry and api reference

* Update documentation and fix some leftover bugs

* Update documentation and add default test params

* Fix identifiers and test data shape for all_estimators tests

* Fix missing params

* Fix n_jobs params and tags, add some docs

* Fix numba test bug and update testing data for sim search

* Fix imports, testing data tests, and impose predict/_predict interface to all sim search estimators

* Fix args

* Fix extract test

* update docs api and notebooks

* remove notes

* Patrick comments

* Adress comments and clean index code

* Fix Patrick comments

* Fix variable suppression mistake

* Divide base class into task specific

* Fix typo in imports

* Empty commit for CI

* Fix typo again

* Add  check_inheritance exception for similarity search

* Revert back to non per type base classes

* Factor check index and typo in test

---------

Co-authored-by: Patrick Schäfer <[email protected]>
Co-authored-by: Matthew Middlehurst <[email protected]>
Co-authored-by: baraline <[email protected]>

Author: 3 people

aeon/similarity_search/__init__.py

@@ -1,7 +1,5 @@
 """Similarity search module."""
 
-__all__ = ["BaseSimilaritySearch", "QuerySearch", "SeriesSearch"]
+__all__ = ["BaseSimilaritySearch"]
 
-from aeon.similarity_search.base import BaseSimilaritySearch
-from aeon.similarity_search.query_search import QuerySearch
-from aeon.similarity_search.series_search import SeriesSearch
+from aeon.similarity_search._base import BaseSimilaritySearch

aeon/similarity_search/_base.py

@@ -0,0 +1,81 @@
+"""Base class for similarity search."""
+
+__maintainer__ = ["baraline"]
+__all__ = [
+    "BaseSimilaritySearch",
+]
+
+
+from abc import abstractmethod
+from typing import Union
+
+import numpy as np
+from numba.typed import List
+
+from aeon.base import BaseAeonEstimator
+
+
+class BaseSimilaritySearch(BaseAeonEstimator):
+    """Base class for similarity search applications."""
+
+    _tags = {
+        "requires_y": False,
+        "fit_is_empty": False,
+    }
+
+    @abstractmethod
+    def __init__(self):
+        super().__init__()
+
+    @abstractmethod
+    def fit(
+        self,
+        X: Union[np.ndarray, List],
+        y=None,
+    ):
+        """
+        Fit estimator to X.
+
+        State change:
+            Changes state to "fitted".
+
+        Writes to self:
+            _is_fitted : flag is set to True.
+
+        Parameters
+        ----------
+        X : Series or Collection, any supported type
+            Data to fit transform to, of python type as follows:
+                Series: 2D np.ndarray shape (n_channels, n_timepoints)
+                Collection: 3D np.ndarray shape (n_cases, n_channels, n_timepoints)
+                or list of 2D np.ndarray, case i has shape (n_channels, n_timepoints_i)
+        y: ignored, exists for API consistency reasons.
+
+        Returns
+        -------
+        self : a fitted instance of the estimator
+        """
+        ...
+
+    @abstractmethod
+    def predict(
+        self,
+        X: Union[np.ndarray, None] = None,
+    ):
+        """
+        Predict method.
+
+        Can either work with new series or with None (for case when predict can be made
+        using the data given in fit against itself) depending on the estimator.
+
+        Parameters
+        ----------
+        X : Series or Collection, any supported type
+            Data to fit transform to, of python type as follows:
+                Series: 2D np.ndarray shape (n_channels, n_timepoints)
+                Collection: 3D np.ndarray shape (n_cases, n_channels, n_timepoints)
+                or list of 2D np.ndarray, case i has shape (n_channels, n_timepoints_i
+                None : If None type is accepted, it means that the predict function will
+                work only with the data given in fit. (e.g. self matrix profile instead)
+        """
+        ...