hidden-outlier-generation 1.0.0

Python implementation of the bisect algorithm for hidden outlier generation.

Hidden Outlier Generator with Bisection Algorithm

Description

This repository hosts a Python-based hidden outlier generator leveraging the bisect algorithm.

Installation

To install, clone this repository:

```bash git clone https://github.com/dschulmeist/hidden-outlier-generation ```

Usage

Here's how to import and use the main function to generate synthetic hidden outliers:

```python from hog_bisect.bisect import BisectHOGen

Initialize the generator

generator = BisectHOGen(data, outlier_detection_method=pyod.models.lof.LOF, seed=42)

Generate hidden outliers

outliers = generator.fit_generate(gen_points=100) ```

Features

Methods and Classes

fit_generate()

Generate synthetic hidden outliers using a bisection algorithm.

• Args:

  • gen_points (int): Number of synthetic points to generate.
  • check_fast (bool): Fast check flag.
  • fixed_interval_length (bool): Flag for fixed interval length.
  • get_origin_type (str): Method to determine the origin.
  • verbose (bool): Verbose flag.
  • n_jobs (int): Number of parallel jobs.

• Returns:

  • ndarray: An array containing generated hidden outliers.

BisectHOGen

A class for generating synthetic hidden outliers.

Code Structure

BisectHOGen Class

The BisectHOGen class initializes the generator and contains utility methods.

```python class BisectHOGen: ... ```

Function Definitions

Functions like outlier_check, inference, interval_check, and bisect are defined to perform various tasks in the outlier detection and generation process.

```python def outlier_check(...): ... def inference(...): ... def interval_check(...): ... def bisect(...): ... ```

Hidden Outlier Generator with Bisection Algorithm

Description

This repository hosts a Python-based hidden outlier generator leveraging the bisect algorithm.

Utility Functions

random_unif_on_sphere(number, dimensions, r, random_state=5)

Generates uniformly distributed random points on a sphere.

• Args:

  • number (int): Number of points to generate.
  • dimensions (int): The dimensions of the sphere.
  • r (float): Radius of the sphere.
  • random_state (int, optional): Random seed.

• Returns:

  • ndarray: An array containing the generated points on the sphere.

gen_powerset(dims)

Generates the power set of dimensions, which are sets containing all possible combinations of dimensions.

• Args:

  • dims (int): Number of dimensions.

• Returns:

  • set: The power set of dimensions.

subspace_grab(indices, data)

Grabs a subspace of the data based on the specified indices.

• Args:

  • indices (list or tuple): Indices of the attributes for the subspace.
  • data (ndarray): The original data.

• Returns:

  • ndarray: The subspace data.

gen_rand_subspaces(dims, upper_limit, include_all_attr=True, seed=5)

Generates random subspaces based on given dimensions.

• Args:

  • dims (int): Number of dimensions.
  • upper_limit (int): Upper limit for the number of subspaces.
  • include_all_attr (bool, optional): Whether to include all attributes.
  • seed (int, optional): Random seed.

• Returns:

  • set: The generated subspaces.

fit_model(subspace, data, outlier_detection_method, tempdir)

Fits an outlier detection model for a given subspace.

• Args:

  • subspace (tuple): The subspace to fit the model on.
  • data (ndarray): The dataset.
  • outlier_detection_method (class): The outlier detection model class.
  • tempdir (str): Temporary directory for storing data.

• Returns:

  • tuple: The subspace and the fitted model.

fit_in_all_subspaces(...)

Fits models for all possible subspaces of the given data.

• Args:

  • outlier_detection_method (class): The outlier detection model class.
  • data (ndarray): The dataset.
  • tempdir (str): Temporary directory for storing data.
  • subspace_limit (int): 2^subspace_limit will be the maximum amount of subspaces fitted.
  • seed (int, optional): Seed for random number generator.
  • n_jobs (int): Number of cores to use.

• Returns:

  • dict: Dictionary of models fitted on different subspaces.