Optimization and analysis of distributed averaging with memory
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
This paper analyzes the rate of convergence of a distributed averaging scheme making use of memory at each node. In conventional distributed averaging, each node computes an update based on its current state and the current states of their neighbours. Previous work observed the trajectories at each node converge smoothly and demonstrated via simulation that a predictive framework can lead to faster rates of convergence. This paper provides theoretical guarantees for a distributed averaging algorithm with memory. We analyze a scheme where updates are computed as a convex combination of two terms: (i) the usual update using only current states, and (ii) a local linear predictor term that makes use of a node's current and previous states. Although this scheme only requires one additional memory register, we prove that this approach can lead to dramatic improvements in the rate of convergence. For example, on the N-node chain topology, our approach leads to a factor of N improvement over the standard approach, and on the two-dimensional grid, our approach achieves a factor of ¿N improvement. Our analysis is direct and involves relating the eigenvalues of a conventional (memoryless) averaging matrix to the eigenvalues of the averaging matrix implementing the proposed scheme via a standard linearization of the quadratic eigenvalue problem. The success of our approach relies on each node using the optimal parameter for combining the two update terms. We derive a closed form expression for the optimal parameter as a function of the second largest eigenvalue of a memoryless averaging matrix, which can easily be computed in a decentralized fashion using existing methods, making our approach amenable to a practical implementation.
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.
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| 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 it