Install

Current install path

Install the latest released package from PyPI:

python -m pip install de-time

Optional multivariate backends for MVMD and MEMD install with:

python -m pip install "de-time[multivar]"

Optional EMD-family wrappers for EMD and CEEMDAN install with:

python -m pip install "de-time[emd]"

Optional torch-backed neural block support for NBEATS_INTERPRETABLE installs with:

python -m pip install "de-time[neural]"

Notebook tooling for the method gallery installs with:

python -m pip install "de-time[notebook]"

Names used by this project:

Do not install the unrelated detime package from PyPI when you want this project; that package is not DeTime.

To install the unreleased main branch directly from GitHub:

python -m pip install "git+https://github.com/systems-mechanobiology/DeTime.git"

Editable install

Use an editable install for local development, release preparation, or when you need the unreleased main branch directly from the repository.

python -m pip install --upgrade pip
python -m pip install -e .[dev,docs]

For optional multivariate wrappers during development:

python -m pip install -e .[dev,docs,multivar]

For the EMD and torch-backed neural extras during development:

python -m pip install -e .[dev,docs,emd,neural]

Platform Prerequisites

Most users will install from wheels and do not need a local compiler. If a wheel is unavailable and pip falls back to building from source, you need:

• Python 3.10+
• a working C/C++ toolchain
• CMake compatible with the scikit-build-core build

Typical source-build prerequisites by platform:

• Windows
• Visual Studio Build Tools with MSVC and CMake on PATH
• macOS
• Xcode Command Line Tools and a working C++17 compiler
• Linux
• GCC or Clang, Python headers, and CMake

Native extension behavior

DeTime ships native kernels for selected flagship methods:

• SSA
• STD
• STDR

Backend behavior:

• backend="auto"
• use native when available, otherwise fall back to Python
• backend="python"
• force the Python implementation
• backend="native"
• require the native kernel and raise if it is unavailable

For SSA:

• speed_mode="exact" uses the deterministic exact decomposition contract
• speed_mode="fast" uses the approximate iterative path

You can inspect the available native exports with:

from detime import native_capabilities

print(native_capabilities())

Optional multivariate backends

MVMD and MEMD are optional-backend methods, not core bundled implementations. They require the multivar extra and currently depend on the installed PySDKit backend. The full method literature and upstream package links live in Method References.

If you want reviewer-grade evidence that those wrappers work in your environment, run:

python examples/optional_multivariate_backends.py

Optional EMD and neural backends

EMD and CEEMDAN require the emd extra because they import PyEMD from the EMD-signal distribution at runtime. WAVELET, WAVEFORM_BLOCK, and WAVELETMIXER_BLOCK use PyWavelets, which is part of the default DeTime dependency set.

Most neural block methods are lightweight NumPy implementations. The exception is NBEATS_INTERPRETABLE, which requires torch and is installed through the neural extra.

Compatibility alias

The preferred import is detime. The compatibility scope retained through 0.1.x is limited to:

• top-level import tsdecomp
• the tsdecomp executable
• python -m tsdecomp

Transition-era submodules such as tsdecomp.backends, tsdecomp.leaderboard, and tsdecomp.methods.* are intentionally excluded from install artifacts.

Troubleshooting FAQ

python -m pip uninstall -y detime
python -m pip install de-time

python -m pip install --upgrade pip
python -m pip install --no-cache-dir de-time

python -m pip install "de-time[multivar]"
python -m pip install "de-time[emd]"
python -m pip install "de-time[neural]"

python -m pip install -e .[dev,docs]