Efficient Fine-Tuning of Transformers for Human Activity Recognition
Analysis
This research explores parameter-efficient fine-tuning techniques, specifically LoRA and QLoRA, for Human Activity Recognition (HAR) using Transformer models. The work likely aims to reduce computational costs associated with training while maintaining or improving performance on HAR tasks.
Key Takeaways
Reference / Citation
View Original"The research integrates LoRA and QLoRA into Transformer models for Human Activity Recognition."