Complexity Science | Reality Bending Lab

About signal complexity

Mon, 05 Dec 2022 00:00:00 +0000

The signals recorded from the brain or the body are rich in information, and there are many ways to analyze them. For instance, for EEG, one can focus on Event Related Potentials (ERP), time-frequency analyses, microstates, etc.

An alternative framework, used to characterize the general characteristics of the signal, relies on the extraction of indices of “complexity” (a general term for constructs such as entropy, chaos, fractal dimension, predictability). However, that field is quite complex (no pun intended), drawing heavily onto mathematical concepts that psychologists or neuroscientists might not be familiar with.

In order to better understand the world of complexity indices as applied to neurophysiology, we have done some groundwork to help us make better decisions in our future usage of this type of analysis.

A gentle introduction to complexity indices applied to neuroscience.
A indices selection guide in which we compare how different indices relate to one another.

Additionally, we also provide an easy way to compute them in Python in our NeuroKit package (see here for the list of functions and here for an EEG application).

NeuroKit2 0.2.0 is out 🎉

Wed, 18 May 2022 00:00:00 +0000

NeuroKit2 0.2.0 is out! 🎉

What was supposed to be a small release turned out in a massive update. A big thanks - and a warm welcome - to An Shu and Max, the newest member of the Reality Bending Team, and thus maintainers of NeuroKit. They worked massively to update all of the examples and docstrings. New features include:

A BRAND NEW WEBSITE with a revamped documentation, now hopefully much more useful to navigate. Check-it out: https://neuropsychology.github.io/NeuroKit/ and let us know what you think!!
An overhaul of the Complexity Indices: With more than a 100 indices, NeuroKit is now the most comprehensive package to quantify chaos, entropy and fractal dimension of signals.
Tons of other improvements and fixes ☺️

Once again, a big thanks to all the contributors for their help in making NeuroKit an awesome open-source software for physiological signal processing!

NeuroKit2 0.1.5 'Complexity' is out 🎉

Fri, 12 Nov 2021 00:00:00 +0000

NeuroKit2 0.1.5 is out! 🎉

In the latest 0.1.5 release of NeuroKit2, our team has fixed several bugs in existing functionalities and in particular, overhauled the support for computing complexity measures of neurophysiological signals. We added a ton of new indices of entropy and fractal dimensions, including:

Petrosian’s, Katz’s and Sevcik fractal dimension
Differential, Permutation, Spectral, SVD entropy
Fisher information
Hjorth’s and Lempel-Ziv’s complexity
Relative Roughness
Hurst and Lyapunov exponent(s)
Detrended Fluctuation Analysis (as well as MFDFA)
…

You can compute them all using the new nk.complexity() function!

import neurokit2 as nk

signal = nk.signal_simulate(frequency=[5, 6], noise=0.5)

results, info = nk.complexity(signal, which="fast")
results

  DiffEn       FI    Hjorth       KFD  PEn  ...      PFD        RR       SFD
1.536573  0.01524  1.355543  4.720953  1.0  ... 1.017423  1.638357  1.691036

To understand more about complexity science, we recommend reading our preprint, which introduces the theoretical (and mathematical) meanings of complexity and reviews the existing studies of complexity analysis across multiple fields of psychology.

Of course, the NeuroKit2 Python package also includes tons of other useful features for physiological signal processing (see this quick example)!

Don’t forget to watch our repo to keep a look out for more complexity functionalities coming up! 👀