Pico is an algorithm that quickly computes the CMB scalar, tensor and lensed power spectra, the matter transfer function and the WMAP 5 year likelihood. It is intended to accelerate parameter estimation codes; Pico can compute the CMB power spectrum and matter transfer function, as well as any computationally expensive likelihoods, in a few milliseconds. It is extremely fast and accurate over a large volume of parameter space and its accuracy can be improved by using a larger training set. More generally, Pico allows using massively parallel computing resources, including distributed computing projects such as Cosmology@Home, to speed up the slow steps in inherently sequential calculations.

The Void IDentification and Examination toolkit (VIDE) identifies voids using a modified version of the parameter-free void finder ZOBOV (ascl:1304.005); a Voronoi tessellation of the tracer particles is used to estimate the density field followed by a watershed algorithm to group Voronoi cells into zones and subsequently voids. Output is a summary of void properties in plain ASCII; a Python API is provided for analysis tasks, including loading and manipulating void catalogs and particle members, filtering, plotting, computing clustering statistics, stacking, comparing catalogs, and fitting density profiles.

sic (Sparse Inpainting Code) generates Gaussian, isotropic CMB realizations, masks them, and recovers the large-scale masked data using sparse inpainting; it is written in Fortran90.

pycola is a multithreaded Python/Cython N-body code, implementing the Comoving Lagrangian Acceleration (COLA) method in the temporal and spatial domains, which trades accuracy at small-scales to gain computational speed without sacrificing accuracy at large scales. This is especially useful for cheaply generating large ensembles of accurate mock halo catalogs required to study galaxy clustering and weak lensing. The COLA method achieves its speed by calculating the large-scale dynamics exactly using LPT while letting the N-body code solve for the small scales, without requiring it to capture exactly the internal dynamics of halos.

This software trains artificial neural networks to find non-linear functionals of data that maximize Fisher information: information maximizing neural networks (IMNNs). As compressing large data sets vastly simplifies both frequentist and Bayesian inference, important information may be inadvertently missed. Likelihood-free inference based on automatically derived IMNN summaries produces summaries that are good approximations to sufficient statistics. IMNNs are robustly capable of automatically finding optimal, non-linear summaries of the data even in cases where linear compression fails: inferring the variance of Gaussian signal in the presence of noise, inferring cosmological parameters from mock simulations of the Lyman-α forest in quasar spectra, and inferring frequency-domain parameters from LISA-like detections of gravitational waveforms. In this final case, the IMNN summary outperforms linear data compression by avoiding the introduction of spurious likelihood maxima.

SSSpaNG is a data-driven Gaussian Process model of the spectra of APOGEE red clump stars, whose parameters are inferred using Gibbs sampling. By pooling information between stars to infer their covariance it permits clear identification of the correlations between spectral pixels. Harnessing this correlation structure, a complete spectrum for each red clump star can be inferred, inpainting missing regions and de-noising by a factor of at least 2-3 for low-signal-to-noise stars.

spinsfast is a fast spin-s spherical harmonic transform algorithm, which is flexible and exact for band-limited functions. It permits the computation of several distinct spin transforms simultaneously. Specifically, only one set of special functions is computed for transforms of quantities with any spin, namely the Wigner d matrices evaluated at π/2, which may be computed with efficient recursions. For any spin, the computation scales as O(L^3), where L is the band limit of the function.

The CompressedFisher library tests whether Fisher forecasts using simulated components are converged. The library contains tools to compute standard Fisher estimates, estimate the level of bias due to the finite number of simulations, and compute the compressed Fisher information. Typical usage of CompressedFisher requires two ensembles of simulations: one set of simulations is given at the fiducial parameters (𝜃) to estimate the covariance matrix. The second is a set of simulated derivatives; these can either be in the form of realizations of the derivatives themselves or simulations evaluate at a set of point in the neighborhood of the fiducial point that the code can use to estimate the derivatives.

LtU-ILI (Learning the Universe Implicit Likelihood Inference) performs machine learning parameter inference. Given labeled training data or a stochastic simulator, the LtU-ILI piepline automatically trains state-of-the-art neural networks to learn the data-parameter relationship and produces robust, well-calibrated posterior inference. The package comes with a wide range of customizable complexity, including posterior-, likelihood-, and ratio-estimation methods for ILI, including sequential learning analogs, and various neural density estimators, including mixture density networks, conditional normalizing flows, and ResNet-like ratio classifiers. It offers fully-customizable, exotic embedding networks, including CNNs and Graph Neural Networks, and a unified interface for multiple ILI backends such as sbi, pydelfi, and lampe. LtU-ILI also handles multiple marginal and multivariate posterior coverage metrics, and offers Jupyter and command-line interfaces and a parallelizable configuration framework for efficient hyperparameter tuning and production runs.