wordsmithing

Author: pmelchior

paper/paper.md

@@ -46,53 +46,57 @@ because galaxies exhibit complex morphologies, which cannot be described by trad
 there are so many sources, they routinely overlap with each other, either due to physical interactions or due to their
 close alignment along the line of sight. To extract all information of interest and avoid biases from incorrect modeling
 assumptions, it is therefore necessary to simultaneously model full scenes comprising many sources instead of analyzing
-each source separately, and each of the source models may itself need to be composed of multiple, morphological complex
+each source separately, and each of the source models may itself need to be composed of multiple,
+morphologically complex
 components.
 
 # Statement of need
 
 `scarlet2` is a Python package for full-scene modeling in observational astronomy. It inherits modeling assumptions from
 `scarlet` [@scarlet], namely that a scene comprises multiple sources, each source comprises multiple
-components, and
-each component is determined by a spectrum model and a morphology model, whose outer product represents the light
+components, and each component is determined by a spectrum model and a morphology model, whose outer product
+represents the light
 emission in a sky region as a hyperspectral data cube (wavelength $\times$ height $\times$ width). `scarlet2` retains
 the object-oriented paradigm and many classes and functions from `scarlet`, but augments standard Python with the `jax`
 library [@jax2018github] for automatic differentiation and just-in-time compilation.
 
 `scarlet2` acts as a flexible, modular, and extendable modeling language for celestial sources that combines parametric
 and non-parametric models to describe complex scenarios such as multi-source blending, strong-lensing systems,
 supernovae and their host galaxies, etc. As a modeling language, `scarlet2` is agnostic about the optimization or
-inference method the user wants to employ, but it provides methods to optimize the likelihood function or sample from
+inference method the user wants to employ, but it also provides methods to optimize the likelihood function or
+sample from
 the posterior, which utilize the `optax` package [@deepmind2020jax] or the `numpyro` inference framework
 [@pyro-2019; @phan-2019], respectively. The likelihood of multiple
-observations (at different resolutions, wavelengths, or observing epochs) times can be combined for a joint model of
+observations (at different resolutions, wavelengths, or observing epochs) can be combined for a joint model of
 static and transient sources. To match the coordinates from different observations, `scarlet2` utilizes the `Astropy`
 package [@astropy]. `scarlet2` can also interface with deep learning methods. Besides being natively portable
-to GPUs,
-parameters can be specified with neural networks as data-driven priors, which helps break the degeneracies that arise
+to GPUs, parameters can be specified with neural networks as data-driven priors, which helps break the
+degeneracies that arise
 when multiple components are fit simultaneously [@sampson-2024].
 
 ![Scene with seven detected sources in multi-band images from the Hyper Suprime-Cam Subaru Strategic Program.
 Each source is modelled with a non-parametric spectrum and morphology (1st panel), the entire scene is then convolved
 with the telescope's point spread function (2nd panel) and compared to the observations (3rd panel).
-The residuals (4th panel) reveal the presence of undetected sources and source components (e.g. in the center of source
+The residuals (4th panel) reveal the presence of previously undetected sources and source components (e.g. in the center of source
 #1).](scarlet2_model.png)
 
 To support the wide range of scientific studies that will be made with large sky surveys, `scarlet2` was designed with
-flexibility and ease of use in mind. Several publications have developed and demonstrated new capabilities, including
-modeling of interstellar dust embedded in distant galaxies
-[@siegel-2025] and of transient sources such as active galactic nuclei [@ward-2025] and tidal disruption
-events
-[@yao-2025].
-Future developments will integrate into cloud-based science platforms, provide support for users to make effective
+flexibility and ease of use in mind. Several publications have developed and demonstrated its capabilities,
+including
+modeling of interstellar dust embedded in distant galaxies [@siegel-2025] and of transient sources such as
+active galactic
+nuclei [@ward-2025] and tidal disruption events [@yao-2025].
+Future developments will integrate `scarlet2` into cloud-based science platforms, provide support for users to
+make effective
 modeling choices and to validate their inference results, and create a robust processing pipeline for joint pixel-level
 analyses of surveys from the Vera C. Rubin Observatory, the Euclid mission, the Nancy Grace Roman Space Telescope, and
 the La Silla Schmidt Southern Survey.
 
 # Acknowledgements
 
 We acknowledge contributions from
-the [LINCC Frameworks Incubator Program](https://lsstdiscoveryalliance.org/programs/lincc-frameworks/incubators/), in
+the [LINCC Frameworks Incubator Program](https://lsstdiscoveryalliance.org/programs/lincc-frameworks/incubators/),
+in
 particular from software engineers Max West, Drew Oldag, and Sean McGuire, in adopting comprehensive software workflows
 through the Python Project Template [@oldag-2024] and creating a user-focused recommendation and validation
 suite.