ACDC: Online Unsupervised Cross-Domain Adaptation

TL;DR

This paper introduces ACDC, a novel framework for online unsupervised cross-domain adaptation that effectively manages multiple data streams with evolving characteristics, improving target domain accuracy.

Contribution

ACDC is a comprehensive, self-evolving neural network framework that integrates feature extraction, domain adaptation, and prediction modules for real-time cross-domain learning.

Findings

Achieves over 10% accuracy improvement in target domain

Handles covariate shift and concept drift effectively

Requires minimal hyper-parameter tuning

Abstract

We consider the problem of online unsupervised cross-domain adaptation, where two independent but related data streams with different feature spaces -- a fully labeled source stream and an unlabeled target stream -- are learned together. Unique characteristics and challenges such as covariate shift, asynchronous concept drifts, and contrasting data throughput arises. We propose ACDC, an adversarial unsupervised domain adaptation framework that handles multiple data streams with a complete self-evolving neural network structure that reacts to these defiances. ACDC encapsulates three modules into a single model: A denoising autoencoder that extracts features, an adversarial module that performs domain conversion, and an estimator that learns the source stream and predicts the target stream. ACDC is a flexible and expandable framework with little hyper-parameter tunability. Our experimental results under the prequential test-then-train protocol indicate an improvement in target accuracy over the baseline methods, achieving more than a 10\% increase in some cases.