Nowak, Martin A. 2006: Five Rules for the Evolution of Cooperation, Science 314: 1560–63

This paper provides a very easy to understand mathematical approach for the evolution of cooperation.

Cooperation is needed for evolution to construct new levels of organization. Genomes, cells, multicellular organisms, social insects, and human society are all based on  cooperation. Cooperation means that selfish replicators forgo some of their reproductive potential to help one another. But natural selection implies competition and therefore opposes cooperation unless a specific mechanism is at work. Here I discuss five mechanisms for the evolution of cooperation: kin selection, direct reciprocity, indirect reciprocity, network reciprocity, and group selection. For each mechanism, a simple rule is derived that specifies whether natural selection can lead to cooperation.

A nice discussion is going on in the NEAT User Group about the mechanism that lead to modular (artificial) neural networks, motivated by the recently published paper The evolutionary origin of modularity. The discussion around the questions, what modularity in a neural network actually is, how to measure it and how to setup evolutionary mechanisms to promote modularity in neural networks highlights the different views on that field that people with different backgrounds have.

I recently got the book Principles of Brain Evolution and can recommend it to everyone interested in evolutionary mechanisms that lead to the complexification of natural neural networks. The author writes in a subjective style that makes explicit why which topics are discussed in the book and which are just touched. Furthermore, he explains the categories in which brain evolution is studied and outlines the general limitations and problems with these categories. In this regard, the book provides a good and comprehensive introduction into the current notion of brain evolution from a cautious, distant perspective.

What a fantastic documentary. Truly inspiring vehicles created by the filmmakers.

I am still working on neural networks with heterogenous conduction delays, inspired by Izhikevich’s work about Polychronization. As I couldn’t find a Python implementation of his script file spnet.m (which is written in Matlab), I wrote it by myself, using the modules Numpy and Matplotlib. It is a more or less direct conversion of the Matlab code to Python thanks to Numpy. Maybe someone else can use this code as well.

spnet.py

Jorge Cham, the artist behind PhDComics gave a nice TEDx-talk at UCLA about the gap between the scientific world and its perception in the public. Enjoyed watching it:

In order to better understand Dynamic Causal Modeling I developed a while ago a small program that allows me to setup a model and to immediatly see how the neural parameters influence the hemodynamic and neural signal generated by the equations. The program is far from being complete and comfortable but I think it helps already a lot to understand the internal dynamics and how the signal generated by estimated parameters look under noise-free conditions. I don’t have time any more to continue the development on this program. But I release it here under a Creative Commons License.

The program is started with “dcm_create” which opens a window in which you can setup a new DCM.

After a click on “Create” another window opens in which you can modify the neural parameters of your model and see how the signals look like.

You can modify the parameters of the A, B and C matrix

And select whether you would like to see the hemodynamic signal (default) or the neural signal (identity).

After a click on “Update” you see the updated signal.

It is important that SPM 8 is in the search path of Matlab. The program was developed for the first SPM8 versions (e.g. spm_dcm_create.m 4310 2011-04-18) which uses the same structure and model than DCM in SPM5, before it was revised after this work.

Feel free to modify and extend the code for your purposes.

DCM_sim.zip