Comparison of machine learning techniques for the classification of echolocation clicks from three species of odontocetes
Bibliographic record
Abstract
A species classifier is presented which decides whether or not short groups of clicks are produced by one or more individuals from the following species: Blainville's beaked whales, short-finned pilot whales, and Risso's dolphins.The system locates individual clicks using the Teager energy operator and then constructs feature vectors for these clicks using cepstral analysis.Two different types of detectors confirm or reject the presence of each species.Gaussian mixture models (GMMs) are used to model time series independent characteristics of the species feature vector distributions.Support vector machines (SVMs) are used to model the boundaries between each species' feature distribution and that of other species.Detection error tradeoff curves for all three species are shown with the following equal error rates: Blainville's beaked whales (GMM 3.32%/SVM 5.54%), pilot whales (GMM 16.18%/SVM 15.00%), and Risso's dolphins (GMM 0.03%/SVM 0.70%). SOMM AIRECe travail concerne la création d'un système pour identifier trois espèces d 'odontocètes par les clics d'écholocation: la baleine à bec de Blainville, la baleine pilote, et le dauphin de Risso.Les clics sont identifiés par l'opérateur d 'énergie Teager-Kaiser, et les vecteurs cepstraux sont construits.Dans un travail de détection, on compare les résultats obtenus avec deux modèles différents : le modèle de mélange gaussiens (MMG) et la machine à vecteurs de support (MVS).Les résultats de la détection sont exprimés par les courbes de DET, « Detection Error Tradeoff».Le point sur les courbes de DET où les probabilités de fausses alarmes et manques de détection sont égales est comme suit : la baleine à bec de Blainville (MMG 3,32%/MVS 5,54%), la baleine pilote (MMG 16,18%/MVS 15,00%) et le dauphin de Risso (MMG 0,03%/MVS 0,70%).
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 | 0.006 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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 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".