proteus 1.0.0

• Added a NEWS.md file to track changes to the package.

proteus 1.1.0

• Automated estimation of differentiation variable (deriv).
• Automated detection of cuda compliant device.
• Various improvement and code consolidation.
• Added function for hyper-param random search.
• Added another loss function (score).

proteus 1.1.1

• Added threshold parameter to control the automatic differentiation.
• Besides prediction, now you get also a sampler for each prediction point.

proteus 1.1.2

• Substitution of threshold parameter with min_default.

proteus 1.1.3

• Added option doing block validation in parallel. Improved numeric stability.

proteus 1.1.4

• Solve issued about saturation of workers during parallelization.
• Fixed default parameters setting for stride.
• Added skewed normal to the list of possible distribution.

proteus 1.1.5

• Fixed worker-related issue
• Rewritten main distributions using torch primitives, improving nummerical stability.
• Set crps as the only loss function, removed the others.

proteus 2.0.0

• Removed non-essential runtime dependencies, including modeest, moments, greybox, tictoc, and the tidyverse utility packages.
• Reimplemented descriptive statistics, error metrics, array slicing, timing, and validation with base R utilities.
• Smoothing, Kalman imputation, and parallel cross-validation are now explicit optional features and fail with an actionable installation message when their packages are unavailable.
• Added input validation, regression tests, and a reproducible introductory vignette.