True iff `a` and `b` differ at most in byte order.

Return the set of inline backtick-wrapped tokens in `text`. Only matches single-backtick spans (inline code), not triple-backtick fences. Spl

Given the position just after ``engine.<name>(`` in *text*, return the set of kwarg names used in the call. Crude but sufficient for skill con

Return the first ``[...]`` substring that parses as a JSON list. Tolerates surrounding prose, markdown fences, or repeated arrays. Returns ``

Alternating optimization for Radar. Objective: min_{W,R} ||X - WX - R||^2 + alpha * ||W||_F^2 + gamma * ||R||_{2,1}

Alternating optimization for ANOMALOUS. Objective: min_{W,R} ||X - XWX - R||^2 + alpha * ||W||_F^2 + lambda_r * ||R||_{

Build n_estimators sets of hyperspheres. Parameters ---------- X : numpy array of shape (n_samples, n_features)

Fit k-Shape on a single channel and return per-timestamp scores. Parameters ---------- ts : np.ndarray of shape (n_timestamps

Default ABOD method. Use all training points with high complexity O(n^3). For internal use only.

Fast ABOD method. Only use n_neighbors for angle calculation. Internal use only

A non-Tree-bearing detector for the compat_load pass-through test.

Format the KB complexity field as a one-line readable string. The KB stores complexity as a dict ``{"time": ..., "space": ...}``. Older entri

Format the KB paper field as a short human-facing reference. The KB stores paper as a dict ``{"id": ..., "short": ...}`` where ``short`` is e

Format the KB requires list as comma-joined ``pyod[extra]`` hints. Each token must be in ``_REQUIRES_TO_EXTRA``; unknown tokens raise KeyErro

Identifies competent base detectors based on correlation scores Parameters ---------- scores : numpy array, shape (n_clf,)

Computes the critical values of a given distance matrix. Parameters ---------- dist_matrix : array-like, shape (n_samples, n_features)

Get local region for each test instance Parameters ---------- X_test_norm : numpy array, shape (n_samples, n_features)

Return the minimum time series length required. Returns ------- min_length : int

Return the minimum time series length required. Returns ------- min_length : int

Minimum time series length required.

Return the minimum time series length required. Need enough timestamps beyond the window to produce several error vectors for a stabl

Get number of jobs for the computation. See sklearn/utils/__init__.py for more information. This function reimplements the logic of joblib to

Get parameter names for the estimator See http://scikit-learn.org/stable/modules/generated/sklearn.base.BaseEstimator.html and sklear

Computes the set of r-neighbours. Parameters ---------- dist_matrix : array-like, shape (n_samples, n_features) The distance matr

Utility function to decide the version of sklearn. PyOD will result in different behaviors with different sklearn version Returns ------

Import scripts/regen_skill.py as a module by file path.

initialize unsupervised detectors if no predefined detectors is provided. Parameters ---------- X : numpy array of sh

True iff `inst` is `sklearn.tree._tree.Tree`. Imported lazily so loading non-tree artifacts does not require sklearn to be importable on this

Yield the per-tree sklearn estimators of an IForest or sklearn.IsolationForest. PyOD's IForest proxies ``estimators_`` through to its ``detect

Load model and processor/tokenizer on first use.

Mirror of joblib.load that swaps in a custom unpickler class. Re-uses joblib's `_validate_fileobject_and_memmap` so compressed files and mmap

Return overlapping log-mel context windows for one waveform. Output shape is ``(n_windows, n_mels * context)``. Clips shorter than one contex

Check if string samples look like audio file paths.

Check if string samples look like image file paths.

Apply the robust median absolute deviation (MAD) to measure the distances of data points from the median. Returns --

Compute Mahalanobis distance scores from prediction errors. Parameters ---------- errors : np.ndarray of shape (n_valid, n_ch

Make and configure a copy of the `base_estimator_` attribute. sklearn/base.py Warning: This method should be used to properly instan

Return `state` unchanged if no realignment is needed, otherwise return a new state dict with the `nodes` ndarray realigned to the running skle

Internal function for calculating the distance with numba. Do not use. Parameters ---------- X : array of shape (n_samples, n_features)

Route via an LLM client (internal; see plan_detection).

Internal method to add subplot of inliers and outliers. Parameters ---------- X_inliers : numpy array of shape (n_samples, n_

Pretty print the dictionary 'params' See http://scikit-learn.org/stable/modules/generated/sklearn.base.BaseEstimator.html and sklearn/base.py

Transform embeddings using fitted preprocessing.

Probe whether pythresh can load its bundled `.pkl` model resources. pythresh's META threshold loads pickled sklearn models from ``pythresh/mo

Rename a field in nodes_arr to simulate a future upstream rename. ``old`` must exist in the current dtype; the returned array has ``new`` inst

Render a deduplicated bullet list across multiple modalities.

Materialise the runtime scaler for this fit call. - ``self.scaler is None`` (the new default) -> a fresh ``MinMaxScaler()`` so each

Change one field's dtype kind/itemsize to simulate an incompatible upstream redefinition.

Score a single channel against pre-learned centroids. Parameters ---------- ts : np.ndarray of shape (n_timestamps,)

Sets fixed random_state parameters for an estimator. Internal use only. Modified from sklearn/base.py Finds all parameters ending ``random_st

Set random seed for reproducibility

Shift-align x by the given lag, zero-padding where needed. Parameters ---------- x : np.ndarray of shape (m,) shift : int Pos

This function is called internally to calculate the shared nearest neighbors (SNN). SNN is reported to be more robust than k nearest n

Internal function for fast snn calculation Parameters ---------- ind : int Indices return by kNN. ref_set_ : int, optional (

This function is called internally to perform subspace outlier detection algorithm. Returns ------- anomaly_

Fit on one channel: initialize centroids, stream with updates. Parameters ---------- windows : np.ndarray of shape (n_windows

Remove content inside well-formed KB-DERIVED blocks. KB-DERIVED blocks already came from the KB by definition; we only want to validate hand-

Suggest substitute detector names for failed slots. Calls ``plan_detection`` with ``exclude_detectors`` set to the union of failed an

Run minimax training loop. Parameters ---------- windows : np.ndarray, shape (n_windows, window_size, n_channels) dev

Train the autoencoder. Parameters ---------- train_loader : torch.utils.data.DataLoader DataLoader for the traini

Build and train the LSTM model using PyTorch. Parameters ---------- inputs : np.ndarray of shape (n_pairs, window_size, n_cha

Check the estimator and the n_estimator attribute, set the `base_estimator_` attribute.

Check the value of alpha and beta and clustering algorithm.

Recursively populate out[prefix] with d.keys(), then recurse.

Validate a single ``state.attr[..]..`` occurrence against ground truth. Returns (ok, problem_message) where problem_message is None on success.

Internal function to calculate weighted cosine using optimized numba code. Parameters ---------- curr_pt : numpy array of shape (n_sa

Load the packaged od-expert skill and on-demand references. In a live Claude Code or Codex session this Markdown is injected into the agent c

Source: https://github.com/ChongYou/subspace-clustering/blob/master/cluster/selfrepresentation.py An active support based algorit

Analyzes the dataset and generates a prompt for GPT-4o to get standardized tags describing the dataset. Returns ----

find the angle between two 3D vectors. Parameters ---------- v1 : list, first vector v2 : list, second vector Returns ------

Parse natural-language feedback into a structured action. Maps a free-text string to a structured-feedback dict using a small set of keyword

Build and return an unfitted detector from a plan. Parameters ---------- plan : dict (DetectionPlan) Output of plan_detection().

Build a detector using a factory preset. Presets are class-method factories that wire common defaults for a modality (e.g., text or image). C

Need to define model in this method. self.feature_size is the number of features in the input data.

Dry-run: return 0 if no file would change, 1 otherwise. Used by --check mode and by tests. Does NOT mutate any file: it reads the file, compu

Compute consensus from successful detector results. Rank-normalizes scores per detector via ``rankdata``, averages to get consensus scores, t

Estimate per-feature contribution to anomaly scores. For each feature column, computes the Pearson correlation between the column's absolute

Add reject option into the unsupervised detector. This comes with guarantees: an estimate of the expected rejection rate (retur

Create the generator of the GAN for a given latent size. Parameters ---------- latent_size : int The size of the latent spac

Utility function for visualizing the synthetic samples generated by generate_data_cluster function. Parameters ---------- X_train : n

Not supported (transductive detector).

Predict raw anomaly score of X using the fitted detector. The anomaly score of an input sample is computed based on different detector

Predict raw anomaly scores for time series X. Parameters ---------- X : array-like of shape (n_timestamps,) or (n_timestamps,

Not supported (transductive detector).

Predict raw anomaly score of X using the fitted detector. The anomaly score of an input sample is computed based on different detecto

Not supported (transductive detector).

Not supported (transductive detector).