Distributed Reinforcement Learning Framework for Resource Allocation in Disaster Response
Bibliographic record
Abstract
Making decisions during a disaster can be challenging when human lives and infrastructures are exposed. An important factor to consider when allocating resources, in these situations, is the critical infrastructures' interdependencies. i2Sim, the Infrastructures' Interdependencies Simulator, is a tool build for that purpose. As a layered architecture, i2Sim includes a dedicated decision-making layer. The use of Reinforcement Learning (RL), a machine learning approach based on an agent learning from experience, has been successfully tested with some dimensionality constraints. This paper introduces two improvements to our previous tests aiming at increasing speed and allowing larger dimensionality problems. The first addition is an improved reward scheme for speeding up convergence while guaranteeing it. The correct application of shaping rewards requires a deep understanding of the problem and extensive convergence tests. The second improvement added is the implementation of a scheduler programmed to trigger multiple instances of the same model using different parameters. This scheduler partitions the state space, enabling the agent's training to be done in parallel via a distributed RL algorithm. With this idea, the state/action matrix representing knowledge is partitioned for training, assigned to computing nodes, populated with knowledge (trained) and collected/reconstructed for use. This work has tested on an IBM cluster with 24 computing nodes. The test model is an aggregated model of the City of Vancouver configured for a disaster with numerous consequences over the different critical infrastructures. Based on the model's configuration, 24 scenarios were identified, created and solved simultaneously. The scheduler automates the training by setting up model and learning parameters, looping execution of instances and gathering results from all nodes. The results verify a proof of concept and enable applicability to new models with highly increased dimensionalities.
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 | 0.000 | 0.000 |
| 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 | 0.000 |
| 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 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".