DAS: A DRL-Based Scheme for Workload Allocation and Worker Selection in Distributed Coded Machine Learning
Bibliographic record
Abstract
Machine Learning (ML) has been widely applied to successfully address a variety of different problems across diverse domains, such as robotics, healthcare, and finance. However, high-complexity ML algorithms often require overlong computation time, which significantly impacts their feasibility. Distributed Machine Learning (DML) has been used to tackle the slow computation problem with high-complexity ML algorithms. Nevertheless, with DML, the computation results from all participating computing devices need to be collected in order to complete an ML task. When part of the participating devices, known as the stragglers, cannot return their results in time, the overall computation time will be extended. Distributed Coded Machine Learning (DCML) is a promising solution to mitigate the negative impact of the stragglers. With DCML, redundancy is injected into an ML task so that only a subset of the results from participating devices are required to finish the ML task. In DCML, how to select proper participating devices, referred to as workers, and how to allocate appropriate workloads to the selected workers are two challenging problems. In this paper, we consider a DCML scenario where numerous computing devices are available for an ML task. These devices are willing to offer their computation capacity in exchange for compensation. To encourage the computing devices to participate in the distributed computation, a reverse auction-based incentive mechanism is employed. With the objective of minimizing both the completion time of the ML task and the compensation for participating devices, we propose a Deep reinforcement learning based workload Allocation and worker Selection scheme for DCML, DAS. To our knowledge, this is the first attempt to simultaneously tackle both the workload allocation and worker selection issues in DCML. Our experimental results indicate that DAS outperforms the state-of-the-art schemes in terms of completion time and compensation.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.004 | 0.002 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".