Back to Homepage of Kalle Rutanen
TIM is a cross-platform C++ / Matlab library for efficient non-linear time-series analysis.
This package contains 64-bit prebuilds of TIM 1.3.0 for Windows and Linux. TIM 1.3.0 can also be built under Mac OS X, but unfortunately I don’t have access to such a machine. Note that the interface has changed somewhat from 1.2.0.
Note: For Windows, you need to have the Visual C++ Redistributable for Visual Studio 2019 installed for the prebuilt TIM 1.3.0 to work.
This package contains 32-bit and 64-bit prebuilds of TIM 1.2.0 for Windows, Linux, and Mac OS X.
Note: For Windows, you need to have the Visual C++ Redistributable for Visual Studio 2012 Update 2 installed for the prebuilt TIM 1.2.0 to work.
In Matlab console, assuming TIM is on Matlab path:
% Generate lots of 10-dimensional gaussian distributed vectors.
A = randn(10, 100000)
% Estimate differential entropy for the underlying random vector.
tim.differential_entropy_kl(A);
The maximum number of threads Matlab uses for parallel computation is limited by maxNumCompThreads, which on my computer and Matlab version is set to 8. Check this value by executing on Matlab console:
maxNumCompThreads
If your computer has a larger amount of logical processors (e.g. 8 physical cores + hyperthreading = 16 logical processors), set this number higher to make full use of them:
maxNumCompThreads(16)
Employing all of the cores may make the computer unresponsive during the computation. While this is efficient, for multitasking it may be more productive to reserve some processors for other tasks (e.g. use 14 instead).
The source code for TIM is hosted on GitHub in a Git repository. Assuming you have Mercurial installed, the repository can be cloned from the command-line by
git clone https://github.org/kaba2/tim
The source code always reflects the current development; it is unstable. To obtain the source code corresponding to this version, write
git checkout 1.3.0
TIM is licensed under the MIT license.