The scaling laws have become the de facto guidelines for designing large language models (LLMs), but they were studied under the assumption of unlimited computing resources for both training and inference. As LLMs are increasingly used as personalized intelligent assistants, their customization (i.e., learning through fine-tuning) and deployment onto resource-constrained edge devices will become more and more prevalent. An urgent but open question is how a resource-constrained computing environment would affect the design choices for a personalized LLM. We study this problem empirically in this work. In particular, we consider the tradeoffs among a number of key design factors and their intertwined impacts on learning efficiency and accuracy. The factors include the learning methods for LLM customization, the amount of personalized data used for learning customization, the types and sizes of LLMs, the compression methods of LLMs, the amount of time afforded to learn, and the difficulty levels of the target use cases. Through extensive experimentation and benchmarking, we draw a number of surprisingly insightful guidelines for deploying LLMs onto resource-constrained devices. For example, an optimal choice between parameter learning and RAG may vary depending on the difficulty of the downstream task, the longer fine-tuning time does not necessarily help the model, and a compressed LLM may be a better choice than an uncompressed LLM to learn from limited personalized data.

After a large language model (LLM) is deployed on edge devices, it is desirable for these devices to learn from user-generated conversation data to generate user-specific and personalized responses in real-time. However, user-generated data usually contains sensitive and private information, and uploading such data to the cloud for annotation is not preferred if not prohibited. While it is possible to obtain annotation locally by directly asking users to provide preferred responses, such annotations have to be sparse to not affect user experience. In addition, the storage of edge devices is usually too limited to enable large-scale fine-tuning with full user-generated data. It remains an open question how to enable on-device LLM personalization, considering sparse annotation and limited on-device storage. In this paper, we propose a novel framework to select and store the most representative data online in a self-supervised way. Such data has a small memory footprint and allows infrequent requests of user annotations for further fine-tuning. To enhance fine-tuning quality, multiple semantically similar pairs of question texts and expected responses are generated using the LLM. Our experiments show that the proposed framework achieves the best user-specific content-generating capability (accuracy) and fine-tuning speed (performance) compared with vanilla baselines. To the best of our knowledge, this is the very first on-device LLM personalization framework.

The Internet of Things (IoT) has been widely applied in personal health monitoring on biosignals. Conventional detection methods in the field count on a variety of heuristic criteria by utilizing extracted features, which are carefully selected through extensive clinical trials and experts’ experiences. Recently, deep learning (DL) gains rapidly growing attention in health monitoring. The most significant advantage of DL-based methods is that DL could execute feature engineering automatically with only labeled data, which results in a great reduction in the expertise involved and manual works in the detection method’s design. However, individual differences among various patients (subjects) can lead to accuracy degradation of the pretrained deep model. Simply fine-tuning the deep model with the patient-specific data cannot alleviate the problem since the pretrained model may not generalize well to new data. To address the problem, we propose a metalearning-based personalization method to generate the personalized neural network for each patient to conduct patient-specific detection. Specifically, the proposed metalearning method leverages a novel patientwise training tasks formatting strategy to train the neural network that ends up with a well-generalized model initialization containing across-patient knowledge. The well-generalized model initialization would then be utilized to perform a quick adaptation to the specific patient’s data domain. In this way, a new patient could be immediately assigned with a personalized neural network using limited labeled data. Experimental results show that the proposed metalearning-based personalization method achieves 8.2%, 2.5%, and 6.4% higher accuracy when compared with the existing DL detection methods in VF detection, AF detection, and human activity recognition, respectively.

Federated learning (FL) has been widely adopted in IoT-enabled health monitoring on biosignals thanks to its advantages in data privacy preservation. However, the global model trained from FL generally performs unevenly across subjects since biosignal data is inherent with complex temporal dynamics. The morphological characteristics of biosignals with the same label can vary significantly among different subjects (i.e., intersubject variability) while biosignals with varied temporal patterns can be collected on the same subject (i.e., intrasubject variability). To address the challenges, we present the personalized meta-federated learning (PMFed) framework for personalized IoT-enabled health monitoring. Specifically, in the FL stage, a novel momentum-based model aggregating strategy is introduced to aggregate clients’ models based on domain similarity in the meta-FL paradigm to obtain a well-generalized global model while speeding up the convergence. In the model personalizing stage, an adaptive model personalization mechanism is devised to adaptively tailor the global model based on the subject-specific biosignal features while preserving the learned cross-subject representations. We develop an IoT-enabled computing framework to evaluate the effectiveness of PMFed over three real-world health monitoring tasks. Experimental results show that the PMFed excels at detection performances in terms of F1 and accuracy by up to 9.4% and 8.7%, and reduces training overhead and throughput by up to 56.3% and 63.4% when compared with the SOTA FL algorithms.