Installation

Installation

R

Installing from CRAN (Stable)

install.packages("smooth")

Installing from GitHub (Development)

if (!require("remotes")) install.packages("remotes")
remotes::install_github("config-i1/smooth")

Dependencies

Depends:

• R >= 3.0.2
• greybox >= 2.0.2

Imports:

• Rcpp, nloptr, zoo, pracma, statmod, MASS, xtable, generics, stats, graphics, methods

LinkingTo:

• Rcpp >= 0.12.3
• RcppArmadillo >= 0.8.100.0.0

System Requirements

Linux: No special requirements; BLAS/LAPACK typically pre-installed.

macOS: gfortran is required. Install via:

# Using Homebrew
brew install gcc

# Or download from CRAN
# https://mac.r-project.org/tools/

See thecoatlessprofessor.com for detailed macOS instructions.

Windows: Standard R build tools (Rtools) should be sufficient.

Troubleshooting

Functions Stop Working After Upgrade

Sometimes after upgrading smooth, C++ functions are cached incorrectly. Solutions:

1. Restart R session
2. If that doesn't work:

   remove.packages("smooth")
   # Delete the "smooth" folder from R packages directory
   # Restart R
   install.packages("smooth")

macOS Compilation Errors

If you see errors about -lgfortran or -lquadmath:

1. Install gfortran from CRAN tools or via Homebrew
2. Ensure gfortran is in your PATH
3. Restart R and retry installation

Verifying Installation

library(smooth)
packageVersion("smooth")

# Test basic functionality
data <- rnorm(100)
model <- adam(data, model="ANN")
print(model)

Python

Installing from PyPI (Stable)

Note: The package is not on PyPI yet. A different stale package of the same name is there. We are working on it.

# Doesn't work yet:
# pip install smooth

Installing from GitHub Release (Pre-built Wheels)

Pre-built wheels are available on the GitHub Releases page. These do not require a C++ compiler or Armadillo to be installed, since the C++ extensions are already compiled.

1. Go to the latest release and download the .whl file matching your platform and Python version. The file names follow this pattern:

   • smooth-<version>-cp<python>-cp<python>-<platform>.whl

   For example:

   • smooth-4.4.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl — Linux x86_64, Python 3.12
   • smooth-4.4.0-cp311-cp311-macosx_14_0_arm64.whl — macOS Apple Silicon, Python 3.11
   • smooth-4.4.0-cp310-cp310-win_amd64.whl — Windows 64-bit, Python 3.10

2. Install the downloaded wheel with pip:

   pip install smooth-4.4.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl

   Replace the filename with the one you downloaded.

Alternatively, you can install directly from the release URL without downloading first:

pip install "https://github.com/config-i1/smooth/releases/download/v4.4.0/smooth-4.4.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl"

Installing from Source (Development)

Important: This package includes C++ extensions that require compilation. You must have a C++ compiler, CMake, and Armadillo installed before running pip install. See System Requirements below.

pip install "git+https://github.com/config-i1/smooth.git@master#subdirectory=python"

Dependencies

Python Version:

• Python >= 3.10

Runtime Dependencies:

• numpy
• pandas
• scipy
• nlopt
• statsmodels
• pybind11

Build Dependencies:

• cmake >= 3.25
• scikit-build-core
• ninja

C++ Compiler:

• Linux: g++ (usually pre-installed, or sudo apt install g++)
• macOS: Xcode Command Line Tools (xcode-select --install)
• Windows: Visual Studio Build Tools with C++ workload

System Requirements

Linux (Debian/Ubuntu):

sudo apt update
sudo apt install libblas-dev liblapack-dev libarmadillo-dev cmake ninja-build

Linux (Fedora/RHEL):

sudo dnf install blas-devel lapack-devel armadillo-devel cmake ninja-build

macOS:

brew install armadillo cmake ninja

Windows: Install Visual Studio Build Tools with C++ workload.

Troubleshooting

Missing nlopt

If you get import errors about nlopt:

pip install nlopt

On some systems, you may need the system nlopt library:

# Debian/Ubuntu
sudo apt install libnlopt-dev

# macOS
brew install nlopt

Armadillo Not Found

If compilation fails with Armadillo errors:

1. Install Armadillo development headers (see System Requirements)
2. Ensure CMake can find Armadillo:

   pkg-config --cflags --libs armadillo

Verifying Installation

import smooth
print(smooth.__version__)

# Test basic functionality
import numpy as np
from smooth import ADAM

data = np.random.randn(100)
model = ADAM(model="ANN")
model.fit(data)
print(model.summary())