absorb.md — A knowledge graph of what AI thinkers are actually saying

paper / ben_j_todd / Jun 12

Missing Flares Indicate Repeating Partial Tidal Disruption Events (rpTDEs)

Analysis of TDE 2022dbl and TDE 2020vdq, which exhibited two UV/optical flares but missed an expected third flare, strongly suggests they are repeating partial tidal disruption events (rpTDEs). The absence of the third flare, with luminosity limits significantly below previous events, and inconsistencies with binary TDE models, support this classification. Modeling indicates these events likely result from a main-sequence star on a bound orbit with a deep tidal encounter, suggesting many rpTDEs may produce only two observable flares, with implications for their use in astrophysical studies.

astrophysicstidal disruption eventsrecurrent flarescelestial mechanicsgalactic nuclei

“TDE 2022dbl and TDE 2020vdq are classified as repeating partial tidal disruption events (rpTDEs) due to the absence of a third expected flare and their observed flare energetics.”

paper / ben_j_todd / Jun 12

AUCp: A Novel Metric for Unsupervised Abnormality Detection Model Selection Using Unlabeled Validation Data

This paper introduces AUCp, a novel metric for model selection in abnormality detection tasks that utilizes unlabeled validation data. Unlike existing methods that rely on labeled validation sets or evaluate image reconstruction realism, AUCp focuses directly on detection performance by treating all unlabeled test samples as positive pseudo-ground truths and calculating a traditional AUC score. This approach has demonstrated mathematical and empirical evidence of improving disease detection for unsupervised and self-supervised methods, outperforming conventional metrics.

medical-image-analysisunsupervised-learningabnormality-detectionself-supervised-learningmodel-selectionauc-metrics

“AUCp is a novel metric for model selection in abnormality detection that does not require a labeled validation set.”

paper / ben_j_todd / Jun 12

Robust Estimation of Cellular Microenvironment Parameters Using IMPULSED Diffusion MRI and Neural Network Mapping

This study investigates the robust estimation of cellular microenvironment parameters ($d$, $V_{in}$, $D_{ex}$) from IMPULSED diffusion MRI signals using a mapping-based framework. Sensitivity analysis at SNR 30 indicated these parameters are reliably derivable. A four-layer neural network demonstrated superior performance over regression models, yielding low mean absolute errors. In vitro validation confirmed the feasibility of noninvasive assessment for monitoring therapeutic responses.

diffusion-mricellular-parameter-estimationuncertainty-quantificationneural-networksmedical-physicstumor-microenvironmentradiation-therapy

“Diffusion MRI can robustly estimate cellular microenvironment parameters such as cell diameter ($d$), intracellular volume fraction ($V_{in}$), and extracellular diffusion coefficient ($D_{ex}$), with these parameters exhibiting relative uncertainty below 1.0.”

paper / ben_j_todd / Jun 12

Phase-Shift Interferometry Turns a 123-Meter Optical Cavity into a Sub-Micron Absolute Strainmeter

This work demonstrates a dual-laser phase-sensing technique that converts the free spectral range variation of a 123-meter high-finesse cavity into a continuous absolute length measurement with sub-micron precision. By phase-locking a second laser at a fixed frequency offset from a resonance-locked laser, cavity length drift produces a measurable phase shift in transmission, enabling a cavity-based strainmeter with seismic sensitivity down to 10^{-10} m/m. The method resolves earth tides and distant earthquakes, positioning long-baseline optical cavities as viable precision geophysical instruments without requiring independent length calibration.

opticslaser-technologycavity-sensingseismic-monitoringgeophysics

“A dual-laser technique using relative phase shift can continuously measure absolute cavity length with sub-micron precision over 123 meters.”