ACDC: Online Unsupervised Cross-Domain Adaptation
This addresses domain adaptation for streaming data with covariate shift and concept drifts, but it is incremental as it builds on existing adversarial methods.
The paper tackles online unsupervised cross-domain adaptation with labeled source and unlabeled target streams, proposing ACDC, an adversarial framework that improves target accuracy by over 10% in some cases.
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.