Answers are grounded in this paper and cite their sources.
Power amplifier (PA) modeling is crucial for linearizing the PA in digital predistortion (DPD). As the number of operational states of PA increases, the number of nonlinear characteristics of PA that need to be learned also increases correspondingly. This escalation results in dual challenges: exponential growth in model parameters and substantial increases in required training data volume for multistate PA modeling. To address these challenges, a PA modeling method empowered by continuous learning with gradient weight integration is proposed, which can eliminate the need for model expansion of multistate PA modeling and only requires the training data of the current state to build the multistate PA model. The experimental results show that the proposed PA modeling method can achieve a 91.6% optimization in storage cost during the multistate PA modeling process and maintains comparable multistate PA modeling accuracy.
No takes yet. Share an insight, caveat, or question.
Li et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: