Local Drift

Local drift, the phenomenon of changing data distributions over time, poses a significant challenge to the reliability and fairness of machine learning models. Current research focuses on detecting and mitigating this drift across various data types (text, images, logs) using methods like statistical process control, document embeddings with dimensionality reduction, and novel federated learning algorithms that incorporate drift correction mechanisms. Addressing local drift is crucial for ensuring the continued accuracy and ethical deployment of machine learning systems in diverse applications, ranging from healthcare to infrastructure monitoring.

Papers

September 19, 2024

Is it Still Fair? A Comparative Evaluation of Fairness Algorithms through the Lens of Covariate Drift
Oscar Blessed Deho, Michael Bewong, Selasi Kwashie, Jiuyong Li, Jixue Liu, Lin Liu, Srecko Joksimovic
Camera Lens Fairness Metric Comparative Evaluation Baseline Algorithm Local Drift

June 13, 2024

FLea: Addressing Data Scarcity and Label Skew in Federated Learning via Privacy-preserving Feature Augmentation
Tong Xia, Abhirup Ghosh, Xinchi Qiu, Cecilia Mascolo
Model Overfitting Label Distribution Data Scarcity Feature Augmentation Rat Image Local Drift

February 12, 2024

Out-of-Distribution Detection and Data Drift Monitoring using Statistical Process Control
Ghada Zamzmi, Kesavan Venkatesh, Brandon Nelson, Smriti Prathapan, Paul H. Yi, Berkman Sahiner, Jana G. Delfino
Machine Learning Distribution Detection Drift Detection Data Drift Drift Analysis Statistical Process Control Local Drift

October 17, 2023

Data Drift Monitoring for Log Anomaly Detection Pipelines
Dipak Wani, Samuel Ackerman, Eitan Farchi, Xiaotong Liu, Hau-wen Chang, Sarasi Lalithsena
Anomaly Detection Drift Detection Log Message Log Anomaly Detection Local Drift Anomaly Contamination

September 17, 2023

Detecting covariate drift in text data using document embeddings and dimensionality reduction
Vinayak Sodar, Ankit Sekseria
Language Model Dimensionality Reduction Drift Detection Text Data Document Embeddings Different Word Embeddings Local Drift Text Mining Model

October 15, 2022

The Influence of Multiple Classes on Learning Online Classifiers from Imbalanced and Concept Drifting Data Streams
Agnieszka Lipska, Jerzy Stefanowski
External Influence Data Stream Link Stream Multi Class Group Imbalance Online Classification Local Drift

March 22, 2022

FedDC: Federated Learning with Non-IID Data via Local Drift Decoupling and Correction
Liang Gao, Huazhu Fu, Li Li, Yingwen Chen, Ming Xu, Cheng-Zhong Xu
Raw Data Language Correction Global Model Local Drift

Local Drift

Papers

Is it Still Fair? A Comparative Evaluation of Fairness Algorithms through the Lens of Covariate Drift

FLea: Addressing Data Scarcity and Label Skew in Federated Learning via Privacy-preserving Feature Augmentation

Out-of-Distribution Detection and Data Drift Monitoring using Statistical Process Control

Data Drift Monitoring for Log Anomaly Detection Pipelines

Detecting covariate drift in text data using document embeddings and dimensionality reduction

The Influence of Multiple Classes on Learning Online Classifiers from Imbalanced and Concept Drifting Data Streams

FedDC: Federated Learning with Non-IID Data via Local Drift Decoupling and Correction