arXiv subjects

Jonathan Lin

Publications and source records attributed to Jonathan Lin.

At least 19 recordsLinked to original sources

arXiv2026

Covariate-localized False Discovery Rates

We introduce a flexible model for covariate-dependent multiple testing which can be encoded using a nonparametric Gaussian mixture model. Weight-localized predictive recursion (PRx), a new development in the methodology of Newton's predictive recursion algorithm, is then leveraged to estimate the components of this mixture model, allowing for recovery of the covariate-localized false discovery rate $\text{Pr}(H_i = 0|z_i,x_i)$ using a single, unified algorithm. This quantity represents the most direct extension of Efron's local false discovery rate to the covariate-dependent setting, and admits provable Bayesian FDR control properties under simple rejection rules. We introduce several procedures for estimating and thresholding the local false discovery rate, and show using various simulations and a real-data example that our procedures lead to increased power, tighter Bayesian FDR control, and more interpretable rejections. We furthermore show that this holds for fixed and randomized hypothesis labels, indicating that our proposed methods perform well under both frequentist and Bayesian interpretations of multiple testing.

stat.ME↗

arXiv2026

The Photonic Lantern Nuller: from concept to laboratory and on-sky demonstrations

This thesis work presents the conceptual design and experimental characterization of the Photonic Lantern Nuller instrument, which uses a multimode-to-single-mode demultiplexing waveguide to cancel out starlight while maintaining planet light, allowing for the direct characterization of planets at a telescope's diffraction limit. The PLN was experimentally characterized in the lab, where it was further enhanced using common-path wavefront sensing and control techniques, and then demonstrated on sky at the Subaru Telescope. Highlights include measured in-lab null-depths of $\sim 10^{-4}$ in three out of four ports simultaneously and on-sky null-depths of approximately $\sim 10^{-1}$ (limited by jitter and atmospheric residuals). We provide an overview of these results and discuss avenues for future work.

astro-ph.IM↗

arXiv2026

Quantum-optimal coronagraphy with spatial mode sorting for direct exoplanet observations

Conventional coronagraphs struggle to reach the theoretical limit of exoplanet detection at close separations to the star, particularly when the telescope has a complex aperture or when the star is partially resolved. Coronagraphy or nulling using spatial mode-sorting can reach the theoretical limit, but the optimal solution has so far only been calculated for an idealized unresolved star, whose signal lies entirely in the piston mode of the telescope. This work aims to enable the calculation of optimal nulling modes for realistic observational scenarios as a function of the size of the star and planet parameters, with the goal of improving coronagraphic performance at ~lambda/D working angles given partially resolved stars and complex telescope apertures. We perform numerical calculations using tools from quantum information theory and explore the behavior of optimal mode-sorting measurements. The optimal measurement for measuring a planet parameter is calculable from the density matrix describing the state of the system. The spatial mode that maximizes the classical signal-to-noise ratio is approximately quantum optimal to leading order in the stellar leakage and the planet flux ratio. We present optimal modes for measuring planets with known signals, and we characterize the tradeoffs inherent to coronagraphs targeting more than one planet location. Example coronagraph designs are presented for three cases of scientific interest: 1) the optimal extension of the fiber nuller architecture for detecting and spectrally characterizing planets across an arbitrary field-of-view using high-resolution spectroscopy, 2) following up planets detected by the visible coronagraph of the Habitable Worlds Observatory at more challenging infrared wavelengths, and 3) detecting and localizing planets at close working angles with the Planetary Camera and Spectrograph on the Extremely Large Telescope.

astro-ph.IM↗

arXiv2026

Fast Semiparametric Density Regression with Weight-localized Predictive Recursion

Predictive recursion (PR) is a fast algorithm for nonparametric estimation of a mixing density, with connections to sequential Bayesian updating under a Dirichlet process prior and rigorous frequentist consistency guarantees. Extending PR to the regression setting, where one seeks to estimate how a mixing density varies with covariate, is nontrivial: dependent Dirichlet process priors, the natural Bayesian generalization, gives no simple recursive updating formula. We introduce PRx, which overcomes this challenge through combining kernel-based weight localization with the recursive scheme of the original PR algorithm. The algorithm scales linearly in sample size and covariate dimension, completing in seconds to minutes where MCMC-based competitors require hours. Exactly as with ordinary PR, the algorithm produces as a byproduct a likelihood score, the PRMLx, whose maximizer is shown to be a consistent estimator for unmixed parameters. In simulations and case studies PRx produces conditional density estimates competitive with established Bayesian procedures at a fraction of the computational cost, and can also be adapted for a wide range of statistical applications including Bayesian model comparison and covariate-dependent multiple testing.

stat.ME↗

arXiv2026

Choice of optical transformation for photonic circuit wavefront sensors

Coronagraph designs which use photonic integrated circuits have the highest theoretical throughput for off-axis signals, and therefore the highest potential exoplanet yield for future high-contrast direct imaging campaigns. Using the rejected starlight, the photonic integrated circuit may also provide simultaneous wavefront sensing, allowing for the correction of non-common-path aberrations. This work considers how a photonic circuit should be configured to maximize its sensitivity to phase aberrations. Two cases are considered: in the first, the photonic circuit is coupled directly to an electric field in a piecewise manner, while in the second, the circuit is coupled to the field via an optical mode sorter. In either case, this work constructs a unitary matrix which can be applied by a photonic circuit to produce maximum sensitivity.

astro-ph.IM↗

arXiv2026

How to Train Your Resistive Network: Generalized Equilibrium Propagation and Analytical Learning

Machine learning is a powerful method of extracting meaning from data; unfortunately, current digital hardware is extremely energy-intensive. There is interest in an alternative analog computing implementation that could match the performance of traditional machine learning while being significantly more energy-efficient. However, it remains unclear how to train such analog computing systems while adhering to locality constraints imposed by the physical (as opposed to digital) nature of these systems. Local learning algorithms such as Equilibrium Propagation and Coupled Learning have been proposed to address this issue. In this paper, we develop an algorithm to exactly calculate gradients using a graph theoretic and analytical framework for Kirchhoff's laws. We also introduce Generalized Equilibrium Propagation, a framework encompassing a broad class of Hebbian learning algorithms, including Coupled Learning and Equilibrium Propagation, and show how our algorithm compares. We demonstrate our algorithm using numerical simulations and show that we can train resistor networks without the need for a replica or readout over all resistors, only at the output layer. We also show that under the analytical gradient approach, it is possible to update only a subset of the resistance values without a strong degradation in performance.

cs.LG↗

arXiv2026

TCBench: A Benchmark for Tropical Cyclone Track and Intensity Forecasting at the Global Scale

TCBench is a benchmark for evaluating global, short to medium-range (1-5 days) forecasts of tropical cyclone (TC) track and intensity. To allow a fair and model-agnostic comparison, TCBench builds on the IBTrACS observational dataset and formulates TC forecasting as predicting the time evolution of an existing tropical system conditioned on its initial position and intensity. TCBench includes state-of-the-art physics-based (TIGGE) and Artificial Intelligence Weather Prediction (AIWP) models (AIFS, Pangu-Weather, FourCastNet v2, GenCast, FNV3). If not readily available (e.g., from the NOAA website as is done with TIGGE), TC tracks are consistently derived from model outputs using the TempestExtremes library. TCBench provides deterministic and probabilistic storm-following metrics. On 2023 test cases, AIWP models skillfully forecast TC tracks, while skillful intensity forecasts require additional steps such as post-processing or task-specific training. Designed for accessibility, TCBench helps AI practitioners tackle domain-relevant TC challenges and equips tropical meteorologists with data-driven tools and workflows to improve prediction and TC process understanding. By lowering barriers to reproducible, process-aware evaluation of extreme events, TCBench aims to democratize data-driven TC forecasting.

cs.CE↗

arXiv2025

On-sky Demonstration of Subdiffraction-limited Astronomical Measurement Using a Photonic Lantern

Resolving fine details of astronomical objects provides critical insights into their underlying physical processes. This drives in part the desire to construct ever-larger telescopes and interferometer arrays and to observe at shorter wavelength to lower the diffraction limit of angular resolution. Alternatively, one can aim to overcome the diffraction limit by extracting more information from a single telescope's aperture. A promising way to do this is spatial mode-based imaging, which projects focal-plane field onto a set of spatial modes before detection, retaining focal-plane phase information crucial at small angular scales but typically lost in intensity imaging. However, the practical implementation of mode-based imaging in astronomy from the ground has been challenged by atmospheric turbulence. Here, we present the first on-sky demonstration of a subdiffraction-limited, mode-based measurement using a photonic lantern (PL)-fed spectrometer installed on the SCExAO instrument at the Subaru Telescope. We introduce a novel calibration strategy that mitigates time-varying wavefront error and misalignment effects, leveraging simultaneously recorded focal-plane images and using a spectral-differential technique that self-calibrates the data. Observing the classical Be star $\beta$ CMi, we detected spectral-differential spatial signals and reconstructed images of its H$\alpha$-emitting disk. We achieved an unprecedented H$\alpha$ photocenter precision of 50$\mu$as in about 10-minute observation with a single telescope, measuring the disk's near-far side asymmetry for the first time. This work demonstrates the high precision, efficiency, and practicality of photonic mode-based imaging techniques to recover subdiffraction-limited information, opening new avenues for high angular resolution spectroscopic studies in astronomy.

astro-ph.IM↗

arXiv2025

A unifying approach to self-organizing systems interacting via conservation laws

We present a unified framework for embedding and analyzing dynamical systems using generalized projection operators rooted in local conservation laws. By representing physical, biological, and engineered systems as graphs with incidence and cycle matrices, we derive dual projection operators that decompose network fluxes and potentials. This formalism aligns with principles of non-equilibrium thermodynamics and captures a broad class of systems governed by flux-forcing relationships and local constraints. We extend this approach to collective dynamics through the PRojective Embedding of Dynamical Systems (PrEDS), which lifts low-dimensional dynamics into a high-dimensional space, enabling both replication and recovery of the original dynamics. When systems fall within the PrEDS class, their collective behavior can be effectively approximated through projection onto a mean-field space. We demonstrate the versatility of PrEDS across diverse domains, including resistive and memristive circuits, adaptive flow networks (e.g., slime molds), elastic string networks, and particle swarms. Notably, we establish a direct correspondence between PrEDS and swarm dynamics, revealing new insights into optimization and self-organization. Our results offer a general theoretical foundation for analyzing complex networked systems and for designing systems that self-organize through local interactions.

cond-mat.soft↗

arXiv2025

Experimental and on-sky demonstration of spectrally dispersed wavefront sensing using a photonic lantern

Adaptive optics systems are critical in any application where highly resolved imaging or beam control must be performed through a dynamic medium. Such applications include astronomy and free-space optical communications, where light propagates through the atmosphere, as well as medical microscopy and vision science, where light propagates through biological tissue. Recent works have demonstrated common-path wavefront sensors for adaptive optics using the photonic lantern, a slowly varying waveguide that can efficiently couple multi-moded light into single-mode fibers. We use the SCExAO astrophotonics platform at the 8-m Subaru Telescope to show that spectral dispersion of lantern outputs can improve correction fidelity, culminating with an on-sky demonstration of real-time wavefront control. To our best knowledge, this is the first such result for either a spectrally dispersed or a photonic lantern wavefront sensor. Combined with the benefits offered by lanterns in precision spectroscopy, our results suggest the future possibility of a unified wavefront sensing spectrograph using compact photonic devices.

astro-ph.IM↗

arXiv2025

Implicit Electric Field Conjugation with the Photonic Lantern Nuller

The Photonic Lantern Nuller (PLN) is an instrument concept designed to characterize exoplanets within a single beam-width from its host star. The PLN leverages the spatial symmetry of a mode-selective photonic lantern (MSPL) to create nulled ports, which cancel out on-axis starlight but allow off-axis exoplanet light to couple. The null-depths are limited by wavefront aberrations in the system as well as by imperfections in the lantern. We show that the implicit electric field conjugation algorithm can be used to reduce the stellar coupling through the PLN by orders of magnitude while maintaining the majority of the off-axis light, leading to deeper null depths (~10^{-4}) and thus higher sensitivity to potential planet signals. We discuss a theory for the tradeoff we observed between the different ports, where iEFC improves the nulls of some ports at the expense of others, and show that targeting one port alone can lead to deeper starlight rejection through that port than when targeting all ports at once. We also observe different levels of stability depending on the port and discuss the implications for practically implementing this technique for science observations.

astro-ph.IM↗

arXiv2025

RACCOON: A Retrieval-Augmented Generation Approach for Location Coordinate Capture from News Articles

Geocoding involves automatic extraction of location coordinates of incidents reported in news articles, and can be used for epidemic intelligence or disaster management. This paper introduces Retrieval-Augmented Coordinate Capture Of Online News articles (RACCOON), an open-source geocoding approach that extracts geolocations from news articles. RACCOON uses a retrieval-augmented generation (RAG) approach where candidate locations and associated information are retrieved in the form of context from a location database, and a prompt containing the retrieved context, location mentions and news articles is fed to an LLM to generate the location coordinates. Our evaluation on three datasets, two underlying LLMs, three baselines and several ablation tests based on the components of RACCOON demonstrate the utility of RACCOON. To the best of our knowledge, RACCOON is the first RAG-based approach for geocoding using pre-trained LLMs.

cs.CL↗

arXiv2024

Assessing the Role of Volumetric Brain Information in Multiple Sclerosis Progression

Multiple sclerosis is a chronic autoimmune disease that affects the central nervous system. Understanding multiple sclerosis progression and identifying the implicated brain structures is crucial for personalized treatment decisions. Deformation-based morphometry utilizes anatomical magnetic resonance imaging to quantitatively assess volumetric brain changes at the voxel level, providing insight into how each brain region contributes to clinical progression with regards to neurodegeneration. Utilizing such voxel-level data from a relapsing multiple sclerosis clinical trial, we extend a model-agnostic feature importance metric to identify a robust and predictive feature set that corresponds to clinical progression. These features correspond to brain regions that are clinically meaningful in MS disease research, demonstrating their scientific relevance. When used to predict progression using classical survival models and 3D convolutional neural networks, the identified regions led to the best-performing models, demonstrating their prognostic strength. We also find that these features generalize well to other definitions of clinical progression and can compensate for the omission of highly prognostic clinical features, underscoring the predictive power and clinical relevance of deformation-based morphometry as a regional identification tool.

stat.AP↗

arXiv2024

Nonlinear techniques for few-mode wavefront sensors

We present several nonlinear wavefront sensing techniques for few-mode sensors, all of which are empirically calibrated and agnostic to the choice of wavefront sensor. The first class of techniques involves a straightforward extension of the linear phase retrieval scheme to higher order; the resulting Taylor polynomial can then be solved using the method of successive approximations, though we discuss alternate methods such as homotopy continuation. In the second class of techniques, a model of the WFS intensity response is created using radial basis function interpolation. We consider both forward models, which map phase to intensity and can be solved with nonlinear least-squares methods such as the Levenberg-Marquardt algorithm, as well as backwards models which directly map intensity to phase and do not require a solver. We provide demonstrations for both types of techniques in simulation using a quad-cell sensor and a photonic lantern wavefront sensor as examples. Next, we demonstrate how the nonlinearity of an arbitrary sensor may studied using the method of numerical continuation, and apply this technique both to the quad-cell sensor and a photonic lantern sensor. Finally, we briefly consider the extension of nonlinear techniques to polychromatic sensors.

astro-ph.IM↗

arXiv2024

Coupled-mode theory for astrophotonics

Coupled-mode theory (CMT) is a powerful tool for simulating near-harmonic systems. In telecommunications, variations of the theory have been used extensively to study waveguides, both analytically and through numerical modelling. Analogous mathematical techniques to the CMT are also widely used in quantum mechanics. The purpose of this work is to collect different formulations of the CMT and their underlying connections to quantum mechanical techniques, and to showcase their utility in modelling slowly varying waveguides including directional couplers and photonic lanterns. My choice of example waveguides is motivated by the astronomical applications of such devices in starlight nulling, wavefront sensing, and high-resolution spectroscopy. I first provide a brief review of the standard form of the CMT, applicable for waveguides with fixed eigenmodes. Next, I show that the CMT also applies for slowly varying waveguides, and demonstrate the close relation between the CMT and several well-known approximation methods from quantum mechanics, as well as concepts like geometric phase. Finally, I present a verification of my analysis, in the form of the numerical package cbeam.

astro-ph.IM↗

arXiv2024

Spectral characterization of a 3-port photonic lantern for application to spectroastrometry

Spectroastrometry, which measures wavelength-dependent shifts in the center of light, is well-suited for studying objects whose morphology changes with wavelength at very high angular resolutions. Photonic lantern (PL)-fed spectrometers have potential to enable measurement of spectroastrometric signals because the relative intensities between the PL output SMFs contain spatial information on the input scene. In order to use PL output spectra for spectroastrometric measurements, it is important to understand the wavelength-dependent behaviors of PL outputs and develop methods to calibrate the effects of time-varying wavefront errors in ground-based observations. We present experimental characterizations of the 3-port PL on the SCExAO testbed at the Subaru Telescope. We develop spectral response models of the PL and verify the behaviors with lab experiments. We find sinusoidal behavior of astrometric sensitivity of the 3-port PL as a function of wavelength, as expected from numerical simulations. Furthermore, we compare experimental and numerically simulated coupling maps and discuss their potential use for offsetting pointing errors. We then present a method of building PL spectral response models (solving for the transfer matrices as a function of wavelength) using coupling maps, which can be used for further calibration strategies.

astro-ph.IM↗

arXiv2024

Machine learning models for Si nanoparticle growth in nonthermal plasma

Nanoparticles (NPs) formed in nonthermal plasmas (NTPs) can have unique properties and applications. However, modeling their growth in these environments presents significant challenges due to the non-equilibrium nature of NTPs, making them computationally expensive to describe. In this work, we address the challenges associated with accelerating the estimation of parameters needed for these models. Specifically, we explore how different machine learning models can be tailored to improve prediction outcomes. We apply these methods to reactive classical molecular dynamics data, which capture the processes associated with colliding silane fragments in NTPs. These reactions exemplify processes where qualitative trends are clear, but their quantification is challenging, hard to generalize, and requires time-consuming simulations. Our results demonstrate that good prediction performance can be achieved when appropriate loss functions are implemented and correct invariances are imposed. While the diversity of molecules used in the training set is critical for accurate prediction, our findings indicate that only a fraction (15-25\%) of the energy and temperature sampling is required to achieve high levels of accuracy. This suggests a substantial reduction in computational effort is possible for similar systems.

physics.comp-ph↗

arXiv2024

On the Potential of Spectroastrometry with Photonic Lanterns

We investigate the potential of photonic lantern (PL) fiber fed spectrometers for two-dimensional spectroastrometry. Spectroastrometry, a technique for studying small angular scales by measuring centroid shifts as a function of wavelength, is typically conducted using long-slit spectrographs. However, slit-based spectroastrometry requires observations with multiple position angles to measure two-dimensional spectroastrometric signals. In a typical configuration of PL-fed spectrometers, light from the focal plane is coupled into the few-moded PL, which is then split into several single-mode outputs, with the relative intensities containing astrometric information. The single-moded beams can be fed into a high-resolution spectrometer to measure wavelength-dependent centroid shifts. We perform numerical simulations of a standard 6-port PL and demonstrate its capability of measuring spectroastrometric signals. The effects of photon noise, wavefront errors, and chromaticity are investigated. When the PL is designed to have large linear responses to tip-tilts at the wavelengths of interest, the centroid shifts can be efficiently measured. Furthermore, we provide mock observations of detecting accreting protoplanets. PL spectroastrometry is potentially a simple and efficient technique for detecting spectroastrometric signals.

astro-ph.IM↗