Overcoming Concept Shift in Domain-Aware Settings through Consolidated Internal Distributions

We develop an algorithm to improve the predictive performance of a pre-trained model under \textit{concept shift} without retraining from scratch when only unannotated samples initial concepts are accessible. this problem as domain adaptation problem, where source data is inaccessible during adaptation. The core idea based on consolidating intermediate internal distribution, learned represent data, after adapting model. provide theoretical analysis and conduct extensive experiments five benchmark datasets demonstrate that proposed method effective.