Upper confidence bound multi-armed bandits for partially observed Hawkes processes
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
We consider the problem of estimating and ranking a set of self-excited point processes when an action must be taken to observe the events of each process. This situation arises in a number of real-world applications, for example, when crime goes unreported in some regions, or COVID-19 cases are undetected due to a lack of testing resources. Often times, such self-excited events may bear implicit causality. Therefore, we start with Hawkes Processes to model how one event triggers the other. In the scenario of undersampling, we propose Hawkes Process Multi-armed Bandits for learning such point processes to quickly learn the riskiest point processes, while carefully balancing exploitation of known (observed) point processes and exploration of unknown processes. By considering the cumulative number of events of each process as a reward, we derive an upper confidence bound on the counting process to inform actions in the form of which processes to observe in upcoming MAB rounds, based on the history of the partially observed point processes. We then derive a regret bound that scales logarithmically with the number of rounds of observation. We test our model on simulated datasets, crime report data in Vancouver and Los Angeles, and earthquake event data from Alaska, California, and worldwide. Our model outperforms several state-of-the-art MAB algorithms that can be adapted to non-stationary point process estimation across the datasets and performance metrics.
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 | 0.005 | 0.052 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.005 |
| Science and technology studies | 0.002 | 0.002 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.004 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
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