GWFAST forecasts the signal-to-noise ratios and parameter estimation capabilities of networks of gravitational-wave detectors, based on the Fisher information matrix approximation. It is designed for applications to third-generation gravitational-wave detectors. It is based on Automatic Differentiation, which makes use of the library JAX (ascl:2111.002). This allows efficient parallelization and numerical accuracy. The code includes a module for parallel computation on clusters.

Many fields in science and engineering measure data that inherently live on non-Euclidean geometries, such as the sphere. Techniques developed in the Euclidean setting must be extended to other geometries. Due to recent interest in geometric deep learning, analogues of Euclidean techniques must also handle general manifolds or graphs. Often, data are only observed over partial regions of manifolds, and thus standard whole-manifold techniques may not yield accurate predictions. In this thesis, a new wavelet basis is designed for datasets like these.

Although many definitions of spherical convolutions exist, none fully emulate the Euclidean definition. A novel spherical convolution is developed, designed to tackle the shortcomings of existing methods. The so-called sifting convolution exploits the sifting property of the Dirac delta and follows by the inner product of a function with the translated version of another. This translation operator is analogous to the Euclidean translation in harmonic space and exhibits some useful properties. In particular, the sifting convolution supports directional kernels; has an output that remains on the sphere; and is efficient to compute. The convolution is entirely generic and thus may be used with any set of basis functions. An application of the sifting convolution with a topographic map of the Earth demonstrates that it supports directional kernels to perform anisotropic filtering.

Slepian wavelets are built upon the eigenfunctions of the Slepian concentration problem of the manifold - a set of bandlimited functions which are maximally concentrated within a given region. Wavelets are constructed through a tiling of the Slepian harmonic line by leveraging the existing scale-discretised framework. A straightforward denoising formalism demonstrates a boost in signal-to-noise for both a spherical and general manifold example. Whilst these wavelets were inspired by spherical datasets, like in cosmology, the wavelet construction may be utilised for manifold or graph data.

EXCEED-DM (EXtended Calculation of Electronic Excitations for Direct detection of Dark Matter) provides a complete framework for computing DM-electron interaction rates. Given an electronic configuration, EXCEED-DM computes the relevant electronic matrix elements, then particle physics specific rates from these matrix elements. This allows for separation between approximations regarding the electronic state configuration, and the specific calculation being performed.

APERO (A PipelinE to Reduce Observations) performs data reduction for the Canada-France-Hawaii Telescope's near-infrared spectropolarimeter SPIRou and offers different recipes or modules for performing specific tasks. APERO can individually run recipes or process a set of files, such as cleaning a data file of detector effects, collecting all dark files and creating a master dark image to use for correction, and creating a bad pixel mask for identifying and dealing with bad pixels. It can extract out flat images to measure the blaze and produced blaze correction and flat correction images, extract dark frames to provide correction for the thermal background after extraction of science or calibration frames, and correct extracted files for leakage coming from a FP (for OBJ_FP files only). It can also take a hot star and calculate telluric transmission, and then use the telluric transmission to calculate principle components (PCA) for correcting input images of atmospheric absorption, among many other tasks.

ODNet uses a convolutional neural network to examine frames of a given observation, using the flux of a targeted star along time, to detect occultations. This is particularly useful to reliably detect asteroid occultations for the Unistellar Network, which consists of 10,000 digital telescopes owned by citizen scientists that is regularly used to record asteroid occultations. ODNet is not costly in term of computing power, opening the possibility for embedding the code on the telescope directly. ODNet's models were developed and trained using TensorFlow version 2.4.

BiGONLight (Bi-local geodesic operators framework for numerical light propagation) encodes the Bi-local Geodesic Operators formalism (BGO) to study light propagation in the geometric optics regime in General Relativity. The parallel transport equations, the optical tidal matrix, and the geodesic deviation equations for the bilocal operators are expressed in 3+1 form and encoded in BiGONLight as Mathematica functions. The bilocal operators are used to obtain all possible optical observables by combining them with the observer and emitter four-velocities and four-accelerations. The user can choose the position of the source and the observer anywhere along the null geodesic with any four-velocities and four-accelerations.

Korg computes stellar spectra from 1D model atmospheres and linelists assuming local thermodynamic equilibrium and implements both plane-parallel and spherical radiative transfer. The code is generally faster than other codes, and is compatible with automatic differentiation libraries and easily extensible, making it ideal for statistical inference and parameter estimation applied to large data sets.

H-FISTA (Hierarchical Fast Iterative Shrinkage Thresholding Algorithm) retrieves the phases of the wavefield from intensity measurements for pulsar spectroscopy. The code accepts input data in ASCII format as produced by PSRchive's (ascl:1105.014) psrflux function, a FITS file, or a pickle. If using a notebook, any custom reader can be used as long as the data ends up in a NumPy array. H-FISTA obtains sparse models of the wavefield in a hierarchical approach with progressively increasing depth. Once the tail of the noise distribution is reached, the hierarchy terminates with a final unregularized optimization, resulting in a fully dense model of the complex wavefield that permits the discovery of faint signals by appropriate averaging.

PDFchem models the cold ISM at moderate and large scales using functions connecting the quantities of the local and the observed visual extinctions and the local number density with probability density functions. For any given observed visual extinction sampled with thousands of clouds, the algorithm instantly computes the average abundances of the most important species and performs radiative transfer calculations to estimate the average emission of the most commonly observed lines.

The Python module 2DFFTUtils implements tasks associated with measuring spiral galaxy pitch angle with 2DFFT (ascl:1608.015). Since most of the 2DFFT utilities are implemented in one place, it makes preparing images for 2DFFT and dealing with 2DFFT data interactively or in scripts event easier.

gsf fits photometric data points, simultaneously with grism spectra if provided, to get posterior probability of galaxy physical properties, such as stellar mass, dust attenuation, metallicity, as well as star formation and metallicity enrichment histories. Designed for extra-galactic science, this flexible, python-based SED fitting code involves a Markov-Chain Monte-Carlo (MCMC) process, and may take more time (depending on the number of parameters and length of MCMC chains) than other SED fitting codes based on chi-square minimization.

fastSHT performs spherical harmonic transforms on a large number of spherical maps. It converts massive SHT operations to a BLAS level 3 problem and uses the highly optimized matrix multiplication toolkit to accelerate the computation. GPU acceleration is supported and can be very effective. The core code is written in Fortran, but a Python wrapper is provided and recommended.

BlackJAX is a sampling library designed for ease of use, speed, and modularity and works on CPU as well as GPU. It is not a probabilistic programming library (PLL), though it integrates well with PPLs as long as they can provide a (potentially unnormalized) log-probability density function compatible with JAX. BlackJAX is written in pure Python and depends on XLA via JAX (ascl:2111.002). It can be used by those who have a logpdf and need a sampler or need more than a general-purpose sampler. It is also useful for building a sample on GPU and for users who want to learn how sampling algorithms work.

ovejero conducts hierarchical inference of strongly-lensed systems with Bayesian neural networks. It requires lenstronomy (ascl:1804.012) and fastell (ascl:9910.003) to run lens models with elliptical mass distributions. The code trains Bayesian Neural Networks (BNNs) to predict posteriors on strong gravitational lensing images and can integrate with forward modeling tools in lenstronomy to allow comparison between BNN outputs and more traditional methods. ovejero also provides hierarchical inference tools to generate population parameter estimates and unbiased posteriors on independent test sets.

The Population Monte-Carlo (PMC) sampling code pmclib performs fast end efficient parallel iterative importance sampling to compute integrals over the posterior including the Bayesian evidence.

mgcnn is a Convolutional Neural Network (CNN) architecture for classifying standard and modified gravity (MG) cosmological models based on the weak-lensing convergence maps they produce. It is implemented in Keras using TensorFlow as the backend. The code offers three options for the noise flag, which correspond to noise standard deviations, and additional options for the number of training iterations and epochs. Confusion matrices and evaluation metrics (loss function and validation accuracy) are saved as numpy arrays in the generated output/ directory after each iteration.

baobab generates images of strongly-lensed systems, given some configurable prior distributions over the parameters of the lens and light profiles as well as configurable assumptions about the instrument and observation conditions. Wrapped around lenstronomy (ascl:1804.012), baobab supports prior distributions ranging from artificially simple to empirical. A major use case for baobab is the generation of training and test sets for hierarchical inference using Bayesian neural networks (BNNs); the code can generate the training and test sets using different priors.

unTimely Catalog Explorer searches for and visualizes detections in the unTimely Catalog, a full-sky, time-domain catalog of detections based on WISE and NEOWISE image data acquired between 2010 and 2020. The tool searches the catalog by coordinates to create finder charts for each epoch with overplotted catalog positions and light curves using the unTimely photometry, to overplot these light curves with AllWISE multi-epoch and NEOWISE-R single exposure (L1b) photometry, and to create image blinks with overlaid catalog positions in GIF format.

PAHDecomp models mid-infrared spectra of galaxies; it is based on the popular PAHFIT code (ascl:1210.009). In contrast to PAHFIT, this model decomposes the continuum into a star-forming component and an obscured nuclear component based on Bayesian priors on the shape of the star-forming component (using templates + prior on extinction), making this tool ideally suited for modeling the spectra of heavily obscured galaxies. PAHDecomp successfully recovers properties of Compact Obscured Nuclei (CONs) where the inferred nuclear optical depth strongly correlates with the surface brightness of HCN-vib emission in the millimeter. This is currently set up to run on the short low modules of Spitzer IRS data (5.2 - 14.2 microns) but will be ideal for JWST/MIRI MRS data in the future.

AMBER (Abundance Matching Box for the Epoch of Reionization) models the cosmic dawn. The semi-numerical code allows users to directly specify the reionization history through the redshift midpoint, duration, and asymmetry input parameters. The reionization process is further controlled through the minimum halo mass for galaxy formation and the radiation mean free path for radiative transfer. The parallelized code is over four orders of magnitude faster than radiative transfer simulations and will efficiently enable large-volume models, full-sky mock observations, and parameter-space studies.

The analytic propagator Kepler-Collisions calculates collisions for Keplerian systems. The algorithm maintains a list of collision possibilities and jumps from one collision to the next; since collisions are rare in astronomical scales, jumping from collision to collision and calculating each one is more efficient than calculating all the time steps that are between collisions.

PTAfast calculates the overlap reduction function in Pulsar Timing Array produced by the stochastic gravitational wave background for arbitrary polarizations, propagation velocities, and pulsar distances.

cuvarbase provides a Python library for performing period finding (Lomb-Scargle, Phase Dispersion Minimization, Conditional Entropy, Box-least squares) on astronomical time-series datasets. Speedups over CPU implementations depend on the algorithm, dataset, and GPU capabilities but are typically ~1-2 orders of magnitude and are especially high for BLS and Lomb-Scargle.

paltas conducts simulation-based inference on strong gravitational lensing images. It builds on lenstronomy (ascl:1804.012) to create large datasets of strong lensing images with realistic low-mass halos, Hubble Space Telescope (HST) observational effects, and galaxy light from HST's COSMOS field. paltas also includes the capability to easily train neural posterior estimators of the parameters of the lensing system and to run hierarchical inference on test populations.

Cloud Killer recovers surface albedo maps by using reflected light photometry to map the clouds and surface of unresolved exoplanets. For light curves with negligible photometric uncertainties, the minimal top-of-atmosphere albedo at a location is a good estimate of its surface albedo. On synthetic data, it shows little bias, good precision, and accuracy, but slightly underestimated uncertainties; exoplanets with large, changing cloud structures observed near quadrature phases are good candidates for Cloud Killer cloud removal.

LensingETC optimizes observing strategies for multi-filter imaging campaigns of galaxy-scale strong lensing systems. It uses the lens modelling software lenstronomy (ascl:1804.012) to simulate and model mock imaging data, forecasts the lens model parameter uncertainties, and optimizes observing strategies.

PGOPHER simulates and fits rotational, vibrational, and electronic spectra. It handles linear molecules and symmetric and asymmetric tops, including effects due to unpaired electrons and nuclear spin, with a separate mode for vibrational structure. The code performs many sorts of transitions, including Raman, multiphoton, and forbidden transitions. It can simulate multiple species and states simultaneously, including special effects such as perturbations and state dependent predissociation. Fitting can be to line positions, intensities, or band contours. PGOPHER uses a standard graphical user interface and makes comparison with, and fitting to, spectra from various sources easy. In addition to overlaying numerical spectra, it is also possible to overlay pictures from pdf files and even plate spectra to assist in checking that published constants are being used correctly.

tvguide determines whether stars and galaxies are observable by TESS. It uses an object's right ascension and declination and estimates the pointing of TESS's cameras using predicted spacecraft ephemerides to determine whether and for how long the object is observable with TESS. tvguide returns a file with two columns, the first the minimum number of sectors the target is observable for and the second the maximum.

The Faiss library performs efficient similarity search and clustering of dense vectors. It contains algorithms that search in sets of vectors of any size, up to ones that possibly do not fit in RAM. It also contains supporting code for evaluation and parameter tuning. Faiss is written in C++ with complete wrappers for Python/numpy. Some of the most useful algorithms are implemented on the GPU.

BornRaytrace uses neural data compression of weak lensing map summary statistics to simulate weak gravitational lensing effects. It can raytrace through overdensity Healpix maps to return a convergence map, include shear-kappa transformation on the full sphere, and also include intrinsic alignments (NLA model).

MCCD (Multi-CCD) generates a Point Spread Function (PSF) model based on stars observations in the field of view. After defining the MCCD model parameters and running and validating the training, the model can recover the PSF at any position in the field of view. Written in Python, MCCD also calculates various statistics and can plot a random test star and its model reconstruction.

The photometric redshift-aided classification pipeline SHEEP uses ensemble learning to classify astronomical sources into galaxies, quasars and stars. It uses tabular data and also allows the use of sparse data. The approach uses SDSS and WISE photometry, but SHEEP can also be used with other types of tabular data, such as radio fluxes or magnitudes.

The simulation and analysis framework ixpeobssim was specifically developed for the Imaging X-ray Polarimetry Explorer (IXPE). It produces realistic simulated observations, in the form of event lists in FITS format, that also contain a strict superset of the information included in the publicly released IXPE data products. The framework's core simulation capabilities are complemented by post-processing applications that support the spatial, spectral, and temporal models needed for analysis of typical polarized X-ray sources, allowing implementation of complex, polarization-aware analysis pipelines. Where applicable, the data formats are consistent with the common display and analysis tools used by the community, e.g., the binned count spectra can be fed into XSPEC (ascl:9910.005), along with the corresponding response functions, for doing standard spectral analysis. All ixpeobssim simulation and analysis tools are fully configurable via the command line.

POSYDON (POpulation SYnthesis with Detailed binary-evolution simulatiONs) incorporates full stellar structure and evolution modeling for single and binary-star population synthesis. The code is modular and allows the user to specify initial population properties and adopt choices that determine how stellar evolution proceeds. Populations are simulated with the use of MESA (ascl:1010.083) evolutionary tracks for single, non-interacting, and interacting binaries organized in grids. Machine-learning methods are incorporated and applied on the grids for classification and various interpolation calculations, and the development of irregular grids guided by active learning, for computational efficiency.

LavAtmos performs gas-melt equilibrium calculations for a given temperature and melt composition. The thermodynamics of the melt are modeled by the MELTS code as presented in the Thermoengine package (ascl:2208.006). In combination with atmospheric chemistry codes, LavAtmos enables the characterization of interior compositions through atmospheric signatures.

PySME is a partial reimplementation of Spectroscopy Made Easy (SME, ascl:1202.013), which fits an observed spectrum of a star with a model spectrum. The IDL routines of SME used to call a dynamically linked library of compiled C++ and Fortran programs have been rewritten in Python. In addition, an object oriented paradigm and continuous integration practices, including build automation, self-testing, and frequent builds, have been added.

PETSc (Portable, Extensible Toolkit for Scientific Computation) provides a suite of data structures and routines for the scalable (parallel) solution of scientific applications modeled by partial differential equations, and is intended for use in large-scale application projects. The toolkit includes a large suite of parallel linear, nonlinear equation solvers and ODE integrators that are easily used in application codes written in C, C++, Fortran and Python. PETSc provides many of the mechanisms needed within parallel application codes, such as simple parallel matrix and vector assembly routines that allow the overlap of communication and computation. In addition, PETSc (pronounced PET-see) includes support for managing parallel PDE discretizations.

Solar-MACH (Solar MAgnetic Connection HAUS) derives and visualizes the spatial configuration and solar magnetic connection of different observers (i.e., spacecraft or planets) in the heliosphere at different times. It provides publication-ready figures for analyzing Solar Energetic Particle events (SEPs) or solar transients such as Coronal Mass Ejections (CMEs). Solar-MACH is available as a Python package; a Streamlit-enabled tool that runs in a browser is also available (solar-mach.github.io)

Blacklight postprocesses general-relativistic magnetohydrodynamic simulation data and produces outputs for analyzing data sets, including maps of auxiliary quantities and false-color renderings. The code can use Athena++ (ascl:1912.005) outputs directly, and also supports files in HARM (ascl:1209.005) and iHARM3d (ascl:2210.013) format. Written in C++, Blacklight offers support for adaptive mesh refinement input, slow-light calculations, and adaptive ray tracing.

iharm3D implements the HARM algorithm (ascl:1209.005) with modifications and enables a second-order, conservative, shock-capturing scheme for general-relativistic magnetohydrodynamics (GRMHD). Written in C, it simulates black hole accretion systems in arbitrary stationary spacetimes.

pixmappy provides a Python interface to gbdes pixel map (astrometry) solutions. It reads the YAML format astrometry solutions produced by gbdes (ascl:2210.011) and issues a PixelMap instance, which is a map from one 2d coordinate system ("pixel") to another ("world") 2d system. A PixelMap instance can be used as a function mapping one (or many) coordinate pairs. An inverse method does reverse mapping, and the local jacobian of the map is available also. The type of mapping that can be expressed is very flexible, and PixelMaps can be compounded into chains of tranformations.

gbdes derives photometric and astrometric calibration solutions for complex multi-detector astronomical imagers. The package includes routines to filter catalogs down to useful stellar objects, collect metadata from the catalogs and create a config file holding FITS binary tables describing exposures, instruments, fields, and other available information in the data, and uses a friends-of-friends matching algorithm to link together all detections of common objects found in distinct exposures. gbdes also calculates airmasses and parallactic angles for each exposure, calculates and saves the expected differential chromatic refraction (DCR) needed for precision astrometry, optimizes the parameters of a photometric model to maximize agreement between magnitudes measured in different exposures of the same source, and optimizing the parameters of an astrometric model to maximize agreement among the exposures and any reference catalogs, and performs other tasks. The solutions derived and used by gbdes are stored in YAML format; gbdes uses the Python code pixmappy (ascl:2210.012) to read the astrometric solution files and execute specified transformations.

The time series reconstruction method of massive astronomical catalogs reconstructs all celestial objects' time series data for astronomical catalogs with great accuracy. In addition, the program, which requires a Spark cluster, solves the boundary source leakage problem on the premise of ensuring accuracy, and the user can set different parameters for different data sets to filter the error records in the catalogs.

NEMESIS (Non-linear optimal Estimator for MultivariatE spectral analySIS) is the general purpose correlated-k/LBL retrieval code developed from the RADTRAN project (ascl:2210.008). Originally based on the correlated-k approximation, NEMESIS also works in line-by-line (LBL) mode. It has been designed to be generally applicable to any planet and with any observing mode and so is suitable for both solar-system studies and also exoplanetary studies.

RADTRAN calculates the transmission, absorption or emission spectra emitted by planetary atmospheres using either line-by-line integration, spectral band models, or 'correlated-K' approaches. Part of the NEMESIS project (ascl:2210.009), the code also incorporates both multiple scattering and single scattering calculations. RADTRAN is general purpose and not hard-wired to any specific planet.

COMET (Clustering Observables Modelled by Emulated perturbation Theory) provides emulated predictions of large-scale structure observables from models that are based on perturbation theory. It substantially speeds up these analytic computations without any relevant sacrifice in accuracy, enabling an extremely efficient exploration of large-scale structure likelihoods. At its core, COMET exploits an evolution mapping approach which gives it a high degree of flexibility and allows it to cover a wide cosmology parameter space at continuous redshifts up to z∼3z \sim 3z∼3. Among others, COMET supports parameters for cold dark matter density (ωc\omega_cωc​), baryon density (ωb\omega_bωb​), Scalar spectral index (nsn_sns​), Hubble expansion rate (hhh) and Curvature density (ΩK\Omega_KΩK​). The code can obtain the real-space galaxy power spectrum at one-loop order multipoles (monopole, quadrupole, hexadecapole) of the redshift-space, power spectrum at one-loop order, the linear matter power spectrum (with and without infrared resummation), Gaussian covariance matrices for the real-space power spectrum, and redshift-space multipoles and χ2\chi^2χ2's for arbitrary combinations of multipoles. COMET provides an easy-to-use interface for all of these computations.

ExoRad 2.0, a generic point source radiometric model, interfaces with any instrument to provide an estimate of several Payload performance metrics. For each target and for each photometric and spectroscopic channel, the code provides estimates of signals in pixels, saturation times, and read, photon, and dark current noise. ExoRad also provides estimates for the zodiacal background, inner sanctum, and sky foreground.

PSFr empirically reconstructs an oversampled version of the point spread function (PSF) from astronomical imaging observations. The code provides a light-weighted API of a refined version of an algorithm originally implemented in lenstronomy (ascl:1804.012). It provides user support with different artifacts in the data and supports the masking of pixels, or the treatment of saturation levels. PSFr has been used to reconstruct the PSF from multiply imaged lensed quasar images observed by the Hubble Space Telescope in a crowded lensing environment and more recently with James Webb Space Telescope (JWST) imaging data for a wide dynamical flux range.

Finder_charts creates multi-band finder charts from image data of various partial- and all-sky surveys such as DSS, 2MASS, WISE, UKIDSS, VHS, Pan-STARRS, and DES. It also creates a WISE time series of image data acquired between 2010 and 2021. All images are reprojected so that north is up and east is to the left. The resulting finder charts can be overplotted with corresponding catalog positions. All catalog entries within the specified field of view can be saved in a variety of formats, including ipac, csv, and tex, as can the finder charts in png, pdf, eps, and other common graphics formats. Finder_charts consists of a single Python module, which depends only on well-known packages, making it easy to install.

NIRDust uses K-band (2.2 micrometers) spectra to measure the temperature of the dust heated by an Active Galactic Nuclei (AGN) accretion disk. The package provides several functionalities to pre-process spectra and fit the hot dust component of a AGN K-band spectrum with a blackbody function. NIRDust needs a minimum of two spectra to run: a target spectrum, where the dust temperature will be estimated, and a reference spectrum, where the emission is considered to be purely stellar. The reference spectrum will be used by NIRDust to model the stellar emission from the target spectrum.