Added outlier detection funcitons

[michaelkiper]

Overview

This PR adds in a set of stand-alone functions for outlier detection for input data. Specifically, it offers the below 4 methods:

• z-score outliers
• k-means cluster distance percentiles
• Mahalanobis distance percentiles
• Local Outlier Factor (k-NN using other distance metrics besides euclidean)

The main show_outliers function outputs a PNG (see example below) that highlights all the outliers, and provides indices of those outliers.

Related Tickets

resolves #28

[Copilot]

Pull Request Overview

This PR adds outlier detection functionality to the cover-class repository and fixes a data type conversion issue in the training pipeline.

Key changes:

• Adds a new outlier_detection.py module with multiple outlier detection methods (z-score, k-means, Mahalanobis distance, and LOF)
• Fixes train label conversion to explicitly use LongTensor type
• Documents the new outlier detection feature in README with usage examples

Reviewed Changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated 5 comments.

File	Description
src/cover_class/outlier_detection.py	New module implementing four outlier detection methods with visualization capabilities
src/cover_class/train.py	Wraps train_labels with LongTensor conversion for proper type handling in dataloader
README.md	Adds documentation section for outlier detection feature with usage examples

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.