Historic Supernova Light Echo Identification with Machine Learning
Bibliographic record
Abstract
Advances in machine learning for visual recognition and ultra-low surface brightness imaging have made it possible to detect older and fainter historic supernova light echoes (SN LEs). We are particularly interested in the historic core-collapse SN (CCSN) Crab (SN 1054), as it is the only CCSN with records of direct-light observations in the last 1000 years. We have improved the SN LE machine-learning Python package ALED (Automated Light Echo Detection), created by Bhullar et al. 2021, by adding false positive masks as an additional input. ALED is visual recognition software that identifies and locates LEs in difference images. Before the invention of ALED, LE images had to be categorized by visual inspection, which was a very time-consuming task. Additionally, we have developed a method for manufacturing and augmenting LE training sets, which has previously not been applied to LEs. We manufactured Dragonfly Telephoto Array (DTA) LEs by extracting LEs from Canada-France-Hawaii Telescope difference images and overlaying them on DTA difference images. The DTA is a promising tool for LE detection because of its ability to observe ultra-low surface brightness structures. Additionally, we augmented the only existing DTA LE image by overlaying it on other DTA images. Both of these procedures provided options for further augmentation, such as changing the LE's brightness and width. We also created a process to mask the bright star difference artifacts in DTA images. These stars are typically mislabeled as LEs, and hence masking them makes LE identification simpler. We have created an effective DTA training set for ALED, which is prepared to search for LEs around the historic CCSN Crab (SN 1054), once more DTA images in that region are procured.
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 | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| 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 | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.002 |
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".