Evaluating machine learning architectures for sound event detection for signals with variable signal-to-noise-ratios in the Beaufort Sea
Bibliographic record
Abstract
This paper explores the challenging polyphonic sound event detection problem using machine learning architectures applied to data recorded in the Beaufort Sea during the Canada Basin Acoustic Propagation Experiment. Four candidate architectures were investigated and evaluated on nine classes of signals broadcast from moored sources that were recorded on a vertical line array of hydrophones over the course of the yearlong experiment. These signals represent a high degree of variability with respect to time-frequency characteristics, changes in signal-to-noise ratio (SNR) associated with varying signal levels as well as fluctuating ambient sound levels, and variable distributions, which resulted in class imbalances. Within this context, binary relevance, which decomposes the multi-label learning task into a number of independent binary learning tasks, was examined as an alternative to the conventional multi-label classification (MLC) approach. Binary relevance has several advantages, including flexible, lightweight model configurations that support faster model inference. In the experiments presented, binary relevance outperformed conventional MLC approach on classes with the most imbalance and lowest SNR. A deeper investigation of model performance as a function of SNR showed that binary relevance significantly improved recall within the low SNR range for all classes studied.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 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 source (direct Gemma or distilled Codex), 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".