Categorical Mask and Spatial Label Refinement

Implements 'FiberMargin', a deterministic training-free operator for repairing categorical masks and spatial labels from coordinates and labels alone. Its primary multiclass operator uses rotated space-filling-curve charts and two-sided class enclosure at one fixed geometric transport range. A class-balanced, isolation-protected chart-disagreement rule provides pointwise repair decisions and audit scores. An auxiliary nearest-neighbour ballot handles binary masks. The 'C++' engine supports two- and three-dimensional coordinates, removes constant axes independently within each specimen, and reuses one deterministic CPU worker budget without nested process pools. Reproducible mask corruptions, planar and volumetric simulators, damage-aware evaluation, and compact licensed human dorsolateral prefrontal cortex and colorectal cancer benchmarks support assessment.

Lightweight Hardware and GPU Compute Detection

Detects central processing unit and graphics processing unit hardware and reports the apparent availability of 'CUDA', 'Metal', 'ROCm', and 'OpenCL' compute backends. Detection uses operating-system information, documented platform interfaces, and optional command-line utilities, without requiring a GPU framework, 'Python', or a vendor software development kit. Backend interpretation follows the official 'CUDA' < https://docs.nvidia.com/cuda/cuda-driver-api/>, 'Metal' < https://developer.apple.com/documentation/metal>, 'ROCm' < https://rocm.docs.amd.com/>, and 'OpenCL' < https://registry.khronos.org/OpenCL/> documentation. Missing hardware, drivers, libraries, and utilities are handled safely.

Competitive Gene Sets Test with the Mann-Whitney-Wilcoxon Test

Friendly implementation of the Mann-Whitney-Wilcoxon test for competitive gene set enrichment analysis.

Consensus Clustering of Omic Data

Procedures to perform consensus clustering starting from a dissimilarity matrix or a data matrix. It's allowed to select if the subsampling has to be by samples or features. In case of computational heavy load, the procedures can run in parallel.

Estimate Procedure in Case of First Order Autocorrelation

Solve first order autocorrelation problems using an iterative method. This procedure estimates both autocorrelation and beta coefficients recursively until we reach the convergence (8th decimal as default). The residuals are computed after estimating Beta using EGLS approach and Rho is estimated using the previous residuals.

Knowledge Discovery by Accuracy Maximization

A self-guided, weakly supervised learning algorithm for feature extraction from noisy and high-dimensional data. It facilitates the identification of patterns that reflect underlying group structures across all samples in a dataset. The method incorporates a novel strategy to integrate spatial information, improving the clarity of results in spatially resolved data.

Generalized Weighted Quantile Sum Regression

Fits Weighted Quantile Sum (WQS) regression (Carrico et al. (2014) ), a random subset implementation of WQS (Curtin et al. (2019) ), a repeated holdout validation WQS (Tanner et al. (2019) ) and a WQS with 2 indices (Renzetti et al. (2023) ) for continuous, binomial, multinomial, Poisson, quasi-Poisson and negative binomial outcomes.

32-Bit Floats

R comes with a suite of utilities for linear algebra with "numeric" (double precision) vectors/matrices. However, sometimes single precision (or less!) is more than enough for a particular task. This package extends R's linear algebra facilities to include 32-bit float (single precision) data. Float vectors/matrices have half the precision of their "numeric"-type counterparts but are generally faster to numerically operate on, for a performance vs accuracy trade-off. The internal representation is an S4 class, which allows us to keep the syntax identical to that of base R's. Interaction between floats and base types for binary operators is generally possible; in these cases, type promotion always defaults to the higher precision. The package ships with copies of the single precision 'BLAS' and 'LAPACK', which are automatically built in the event they are not available on the system.

Implement Fleming-Viot-Dependent Dirichlet Processes

A Bayesian Nonparametric model for the study of time-evolving frequencies, which has become renowned in the study of population genetics. The model consists of a Hidden Markov Model (HMM) in which the latent signal is a distribution-valued stochastic process that takes the form of a finite mixture of Dirichlet Processes, indexed by vectors that count how many times each value is observed in the population. The package implements methodologies presented in Ascolani, Lijoi and Ruggiero (2021) and Ascolani, Lijoi and Ruggiero (2023) that make it possible to study the process at the time of data collection or to predict its evolution in future or in the past.

Spatial KWD for Large Spatial Maps

Contains efficient implementations of Discrete Optimal Transport algorithms for the computation of Kantorovich-Wasserstein distances between pairs of large spatial maps (Bassetti, Gualandi, Veneroni (2020), ). All the algorithms are based on an ad-hoc implementation of the Network Simplex algorithm. The package has four main helper functions: compareOneToOne() (to compare two spatial maps), compareOneToMany() (to compare a reference map with a list of other maps), compareAll() (to compute a matrix of distances between a list of maps), and focusArea() (to compute the KWD distance within a focus area). In non-convex maps, the helper functions first build the convex-hull of the input bins and pad the weights with zeros.