nbodykit provides algorithms for analyzing cosmological datasets from N-body simulations and large-scale structure surveys, and takes advantage of the abundance and availability of large-scale computing resources. The package provides a unified treatment of simulation and observational datasets by insulating algorithms from data containers, and reduces wall-clock time by scaling to thousands of cores. All algorithms are parallel and run with Message Passing Interface (MPI); the code is designed to be deployed on large super-computing facilities. nbodykit offers an interactive user interface that performs as well in a Jupyter notebook as on super-computing machines.

pyRSD computes the theoretical predictions of the redshift-space power spectrum of galaxies. It also includes functionality for fitting data measurements and finding the optimal model parameters, using both MCMC and nonlinear optimization techniques.

HorizonGRound forward models general relativistic effects from the tracer luminosity function. It also compares relativistic corrections with the local primordial non-Gaussianity signature in ultra-large-scale clustering statistics. The package includes several recipes along with the data required to run them.

Zeus is a pure-Python implementation of the Ensemble Slice Sampling method. Ensemble Slice Sampling improves upon Slice Sampling by bypassing some of that method's difficulties; it also exploits an ensemble of parallel walkers, thus making it immune to linear correlations. Zeus offers fast and robust Bayesian inference and efficient Markov Chain Monte Carlo without hand-tuning. The code provides excellent performance in terms of autocorrelation time and convergence rate, can scale to multiple CPUs without any extra effort, and includes convergence diagnostics.

hankl implements the FFTLog algorithm in lightweight Python code. The FFTLog algorithm can be thought of as the Fast Fourier Transform (FFT) of a logarithmically spaced periodic sequence (= Hankel Transform). hankl consists of two modules, the General FFTLog module and the Cosmology one. The latter is suited for modern cosmological application and relies heavily on the former to perform the Hankel transforms. The accuracy of the method usually improves as the range of integration is enlarged; FFTlog prefers an interval that spans many orders of magnitude. Resolution is important, as low resolution introduces sharp features which in turn causes ringing.

pocoMC performs Bayesian inference, including model comparison, for challenging scientific problems. The code utilizes a normalizing flow to precondition the target distribution by removing any correlations between its parameters. pocoMC then generates posterior samples, used for parameter estimation, with a powerful adaptive Sequential Monte Carlo algorithm manifesting a sampling efficiency that can be orders of magnitude higher than without precondition. Furthermore, pocoMC also provides an unbiased estimate of the model evidence that can be used for the task of Bayesian model comparison. The code is designed to excel in demanding parameter estimation problems that include multimodal and highly non–Gaussian target distributions.

SUNBIRD trains neural-network-based models for galaxy clustering. It also incorporates pre-trained emulators for different summary statistics, including galaxy two-point correlation function, density-split clustering statistics, and old-galaxy cross-correlation function. These models have been trained on mock galaxy catalogs, and were calibrated to work for specific samples of galaxies. SUNBIRD implements routines with PyTorch to train new neural-network emulators.