Dear volunteers, today all WUs have been processed. There is time to analyse the results. Thanks for your support. The last batch was far to be easy to handle with its awful 10GB RAM requirements but your performance was outstanding. When new plans for Neurona@Home are available, they will be announced here.
We just processed all WUs from current batch. Congratulations and thanks you all. These WUs required not much RAM (about 2.2 GB each). However, once this has been done, we are moving to more demanding WUs, of about 10GB RAM or more. Time for real hardcore computing. Check your host before joining!
This project has been presented in the
“12th Granada Seminar for Physics, Computation, and the Mind” (http://ergodic.ugr.es/cp/)
A summary of the speaking is available here
A more extended paper with the complete content of the speaking is to appear in brief, I will keep you informed.
Looks like some of the WUs we are computing in this phase are able to sustain the neuronal activity during all the test period.
Here we have the results of a WU (problema 20557902) where the average connectivity is k=300, the average transition time from resting to firing is 10 and the transition time from refractory to standby is 3 discrete steps.
In the following WU (problema 20540903) the transition time from refractory to resting is of 2 steps and the other parameters are the same as the former figure:
During a server update something went wrong and as a result our database became corrupted, preventing to run not only Neurona@Home but our entire BOINC server.
Things became worse when we realised that our backup script contained a bug, and that our last valid database copy was done on November 2011. The problem was not the lose of some WUs, but the data regarding all users created since then, as well as the credit information.
We have spent a ton of hours and sleepless nights working to recover the corrupted database, and finally, and thanks to the invaluable help of a DDBB expert from http://www.canalboinc.com/, that is one of our project supporters (thanks, Mr. Poveda!), we recovered all the information.
Lessons learned, we have improved our backup system and now are much more aware of what and what not to do with our server without the appropriate technical support.
So, dear volunteers, computation is on the road again. Ensure that you are able to log in and that your account information is OK, but I hope no issues should arise.
Our BOINC server suffered a data base corruption during a system update. Our last valid backup is several weeks old so we are trying to recover all the information before having to use the backup and lose the last weeks of user information. We have had no success so far. I will keep you informed about this issue. Thank you for your understanding.
We have a problem in our server. The projects are going to be offline until we fix it. We are working on it. Sorry for the inconvenience.
Network change has been done, with quite less problems than expected. The change has given us also a new server IP. DNS has been updated, but may take a while to propagate. Hope this helps to improve the project performances.
Dear volunteers, in the last weeks you have computed 640000 WUs with peaks of about 10GB RAM, which is pretty high. Congratulations, not many projects can say they have such devoted high spec helpers. From the many results obtained, there is only one promising solution found, and this is the corresponding to the WU labelled Prueba10156690.pro, the only one able to sustain the neuronal activity (firing excitatory neurons) with 1000000 cellular automata neurons. The superimposed fluctuations should correspond to the electroencephalogram.
There is also a second second case (k=1030) where the activity finally fades off at t=1000 but several large peaks are found before the extinction.
We will continue with the analysis of the parameter space to determine the regions where the neural network behave as a real brain concerning the average behaviour of the collective of neurons. You are going now to compute WUs changing the inhibition recovery rate and time, instead of exploring networks with high connectivity. The first effect is that Phase IV requires quite less RAM, 1 or 2 GB should be enough. I expect this batch (which is composed of 680000 WUs) to be finished in just a few weeks.
We are going to migrate our laboratory’s Internet connection from the current ADSL line to the high speed dedicated University’s network. This means that the project will be off-line a few days.