Distributed Learning as a Service: The Developer's Perspective
Authors: Tianyue Chu, Filippo Vannella, Dimitra Tsigkari, Paula Delgado-Santos, Fernando López, Pablo Gomez Guerrero, Sotirios Spantideas, David Solans Noguero
Organizations: Telef´onica Scientific Research, Telef´onica, Madrid, Spain · National and Kapodistrian University of Athens, Psachna, Evia, Greece
Application developers of distributed learning services face challenges that a typical federated learning loop does not address. Specifically, the model updates can still leak private data, devices might not be able to participate in the training due to limited resources, a single aggregator might not be able to scale, and the transmissions of model weights induce a considerable bandwidth cost. This paper demonstrates DLaaS (Distributed Learning as a Service) from the developer's vantage point. Using a single admin dashboard, the developer initiates a distributed/federated learning job and is able to activate Differential Privacy (DP), Split Learning (SL), Hierarchical Aggregation (HA), and Knowledge Distillation (KD) as declarative options, with no change to the clients' code. We demonstrate the complete service lifecycle on an industrial smart-home Wake-up Word (WuW) task, using the "Ok Aura" dataset. Once the developer initiates a distributed learning job by toggling DP, SL, HA, and KD in the admin dashboard, the system dispatches the job to a set of Android clients and Dockerized helper aggregators. In the demonstration, these mechanisms run live across configurations. Then, the clients train the model locally and return their updates. The trained model is served to a consumer-side Android application that performs on-device WuW detection on a live microphone stream. In particular, the conference attendees will be invited to speak the trigger phrase and monitor in real time the per-class confidence and inference latency. Finally, we release the source code and short video walkthroughs of these configurations.
Figures & tables
Fig. 1: DLaaS architecture annotated for the demonstration as a job flows through the system: (1) the developer composes the job at the admin front-end (DP mechanism and its (ε,δ) values, the SL, HA, and KD toggles) and submits the specifications; (2) the server cluster compiles the given specs into a per-round plan and runs the server-side core logic (central DP, SL, KD, model aggregation, and device scheduling), driving the distributed learning round loop; (3) the helper tier performs partial aggregation over client clusters and can host the SL server-side part; and (4) each client device runs local training, local DP, and sends the trained model back to the server.
Fig. 2: (1) Compose — the server-side admin interface (Django admin). (a) the admin home, where projects are created, scheduled, and managed; (b) a project’s training configuration, selecting the model, dataset, and number of FL rounds; and (c) the DP mechanism with its (ε,δ) configuration and the SL, HA, and KD module toggles. The whole job is composed from this one interface.
Fig. 4: (3) Local training — the Android client. Left: the authenticated app on the emulator. Right: the Logcat lines reporting per-epoch training loss and accuracy (timestamp, PID, and tag columns omitted for readability).
Fig. 5: (4) Detect — the consumer app. It downloads the trained global model and reports live on the “Ok Aura” detection task with per-class confidence and inference latency.
This work focuses on the question of learning from a large number of devices with each device holding only a single sample of data. Several real-world applications exist to this one sample per client setup up including learning from fitness trackers, data/app usage aggregators, body-worn sensing devices, and daily event monitors to name a few. When a client has only one sample, the standard federated learning paradigm breaks down as a local update based on that single point is far from being useful, especially in the earlier rounds for estimation of the model coefficients. This utility is further weakened by the privacy-inducing noise applied at every round. This work caters to this problem to enable such clients to collaboratively contribute to effectively learn a global model without leaking the privacy of their data. The proposed approach injects a single, carefully calibrated noisy perturbation to transform the sample at each client, followed by a post-processed representation which is shared with the server. These representations aggregated at the server are processed to obtain an unbiased gradient update that in expectation matches the non-private centralized gradient while preserving data privacy. This approach is different than traditional private federated learning, where the communication payloads involve model coefficients as opposed to privately transformed data samples. This method enables devices with extremely limited data to collaborate and learn accurate, privacy-preserving models without requiring large local datasets or sacrificing individual privacy.
Praneeth Vepakomma, Amirhossein Reisizadeh, Samuel Horváth +1
Massachusetts Institute of Technology (MIT) · Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)
Federated learning (FL) enables training of a global model while keeping raw data on end-devices. Despite this, FL has shown to leak private user information and thus in practice, it is often coupled with methods such as differential privacy (DP) and secure vector sum to provide formal privacy guarantees to its participants. In realistic cross-device deployments, the data are highly heterogeneous, so vanilla federated learning converges slowly and generalizes poorly. Clustered federated learning (CFL) mitigates this by segregating users into clusters, leading to lower intra-cluster data heterogeneity. Nevertheless, coupling CFL with DP remains challenging: the injected DP noise makes individual client updates excessively noisy, and the server is unable to initialize cluster centroids with the less noisy aggregated updates. To address this challenge, we propose PINA, a two-stage framework that first lets each client fine-tune a lightweight low-rank adaptation (LoRA) adapter and privately share a compressed sketch of the update. The server leverages these sketches to construct robust cluster centroids. In the second stage, PINA introduces a normality-driven aggregation mechanism that improves convergence and robustness. Our method retains the benefits of clustered FL while providing formal privacy guarantees against an untrusted server. Extensive evaluations show that our proposed method outperforms state-of-the-art DP-FL algorithms by an average of 2.9% in accuracy for privacy budgets (epsilon in {2, 8}).
Jie Xu, Haaris Mehmood, Rogier Van Dalen +2
Samsung R&D Institute UK (SRUK) · Samsung AI Centre Cambridge
We present Federated Learning Playground, an interactive browser-based platform inspired by and extends TensorFlow Playground that teaches core Federated Learning (FL) concepts. Users can experiment with heterogeneous client data distributions, model hyperparameters, and aggregation algorithms directly in the browser without coding or system setup, and observe their effects on client and global models through real-time visualizations, gaining intuition for challenges such as non-IID data, local overfitting, and scalability. The playground serves as an easy to use educational tool, lowering the entry barrier for newcomers to distributed AI while also offering a sandbox for rapidly prototyping and comparing FL methods. By democratizing exploration of FL, it promotes broader understanding and adoption of this important paradigm.
Bryan Shan, Alysa Ziying Tan, Han Yu
Nanyang Technological University (NTU), Singapore · Alibaba-NTU Singapore Joint Research Institute, NTU, Singapore · Alibaba Group, Hangzhou, China