Over-the-Air Federated Multi-Task Learning Over MIMO Multiple Access Channels

نویسندگان

چکیده

With the explosive growth of data and wireless devices, federated learning (FL) over medium has emerged as a promising technology for large-scale distributed intelligent systems. Yet, urgent demand ubiquitous intelligence will generate large number concurrent FL tasks, which may seriously aggravate scarcity communication resources. By exploiting analog superposition electromagnetic waves, over-the-air computation (AirComp) is an appealing solution to alleviate burden required by FL. However, sharing frequency-time resources in inevitably brings about problem inter-task interference, poses new challenge that needs be appropriately addressed. In this paper, we study multi-task (OA-FMTL) multiple-input multiple-output (MIMO) multiple access (MAC) channel. We propose novel model aggregation method alignment local gradients different alleviates straggler due channel heterogeneity. establish communication-learning analysis framework proposed OA-FMTL scheme considering spatial correlation between formulate optimization design transceiver beamforming device selection. To solve problem, develop algorithm using alternating (AO) fractional programming (FP), effectively mitigates impact interference on performance. show use method, selection no longer essential, thereby avoiding heavy computational involved selecting active devices. Numerical results demonstrate validity superb performance scheme.

برای دانلود باید عضویت طلایی داشته باشید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Over-the-Air Computation in MIMO Multi-Access Channels: Beamforming and Channel Feedback

Future Internet-of-Things (IoT) networks are expected to wirelessly connect billions of sensors. It is impractical to collect data from the massive population of sensors using any traditional orthogonal multi-access scheme that leads to excessive latency. To tackle the challenge, a technique called over-theair computation (AirComp) was recently developed to enable a data-fusion to receive a des...

متن کامل

Time-Reversal Space Division Multiple Access over Multi-path Channels

A novel concept of time-reversal division multiple access (TRDMA) is proposed in [1] as a low-complexity, energy-efficient wireless multi-user access method. In this paper, we first generalize the work in [1] as a pre-filtering based space division multiple access (SDMA) communication system, where each user’s signature waveform is derived from the information of CIRs between basestation (BS) a...

متن کامل

Federated Multi-Task Learning

Federated learning poses new statistical and systems challenges in training machinelearning models over distributed networks of devices. In this work, we show thatmulti-task learning is naturally suited to handle the statistical challenges of thissetting, and propose a novel systems-aware optimization method, MOCHA, that isrobust to practical systems issues. Our method and theor...

متن کامل

Reliable Computation over Multiple-Access Channels

In the standard multiple-access problem, a central access point needs to reconstruct the signals observed by each user separately. Suppose now that the access point only reconstructs the sum (or any other function) of these observed signals. For this problem, we develop strategies and information-theoretic performance bounds. It is shown that in general, separating source from channel coding le...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: IEEE Transactions on Wireless Communications

سال: 2023

ISSN: ['1536-1276', '1558-2248']

DOI: https://doi.org/10.1109/twc.2022.3221797