Work Planning
log in

Advanced search

Message boards : Science : Work Planning

Author Message
Profile jcastro
Volunteer developer
Volunteer tester
Project scientist
Avatar
Send message
Joined: 16 Mar 15
Posts: 218
Credit: 14,859
RAC: 0
Message 97 - Posted: 11 Apr 2015, 16:18:53 UTC

Hi!

Some of you have ask about the guidelines of our project. One part of them are explained in other post in that forum http://denis.usj.es/denisathome/forum_thread.php?id=9&postid=51#51 .

The other part is related to improve the way that we use BOINC to solve the calculus of that part. To do that we are first of all preparing the server. This is why we generate small tasks ( 2 minutes and 30 minutes ) to test the capacity of our server an problems that can occur.

We have operations that can cost 100 and 300 minutes. But before send them to you, we have to improve our program (using checkpoints) and improve our server (solve all the traffic problems). After that, we will generate a continuous workload that we hope that could be better to you.

Regards, Joel.

Dr Who Fan
Avatar
Send message
Joined: 8 Apr 15
Posts: 26
Credit: 101,303
RAC: 0
Message 101 - Posted: 12 Apr 2015, 3:14:57 UTC - in response to Message 97.

Thank you for posting this information and taking the time to let us know how things are progressing with the new project.
____________

[VENETO] boboviz
Send message
Joined: 9 Apr 15
Posts: 108
Credit: 403,294
RAC: 0
Message 158 - Posted: 30 Apr 2015, 14:57:56 UTC

Any preliminary results? Is the code working ok??

Dayle Diamond
Send message
Joined: 28 Apr 15
Posts: 12
Credit: 274,008
RAC: 0
Message 159 - Posted: 30 Apr 2015, 17:30:22 UTC

We seem to be churning through every task in the queue within a day.

I would love to get the big picture of how much work this project needs.

Profile jcastro
Volunteer developer
Volunteer tester
Project scientist
Avatar
Send message
Joined: 16 Mar 15
Posts: 218
Credit: 14,859
RAC: 0
Message 160 - Posted: 1 May 2015, 12:10:44 UTC

Hi!

The computation that we are made now is to test the server, and to see the impact of the size of the computation in the infrastructure. We are designing a continuous generation of work units, based on aleatory generation of simulation to fill a big database with the data of this simulations.

We are still defining the goals of our project and the way that we get the solution.

We think that the "aleatory simulation" could be the best way to generate a constant number of workunits and the database generated could be very useful for future heart cells research.

For the next week we will generate the new version of the app, with checkpoint, that make possible to solve longer problems, this will be our next step.

Thanks for your attention, is great to see people that are so interested in science!

Dayle Diamond
Send message
Joined: 28 Apr 15
Posts: 12
Credit: 274,008
RAC: 0
Message 161 - Posted: 1 May 2015, 16:15:53 UTC - in response to Message 160.

Thank you for your quick response.

I hope that this stress test is giving you an excellent preview.

As a laymen, I'm a little nervous about your answer.

Aleatory simulations rely on random number generators to make the work units, right? So the work could go on indefinitely? You seem to be proposing unlimited work for a project, but you're not sure about your goals or methodology?

Profile jcastro
Volunteer developer
Volunteer tester
Project scientist
Avatar
Send message
Joined: 16 Mar 15
Posts: 218
Credit: 14,859
RAC: 0
Message 162 - Posted: 2 May 2015, 9:46:12 UTC

Hi!

We are searching not just a way to solve the problem, that is generate simulations in a little piece of time, we are searching the best way to do this with a volunteer computing network like boinc.

The first attempt to do that, was the system that we are using now. Generate configuration files for each specific simulation. One scientist give us the parameters of the simulation and we do it.

But while we were testing the server, we see that for volunteer computing networks will be more interesting and useful a continuous workload. This is why we think that an aleatory generator of simulations could be useful. Generate in a determinate range of parameters aleatory variations of those configuration files, store the result in a database to work as seed for future investigations.

I'm not native english and this terms are hard to explain, so if you still have doubts we will delighted to answer all your questions.

Thanks for your interest, Joel.

Profile Chus Carro
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 18 Mar 15
Posts: 82
Credit: 424,346
RAC: 0
Message 165 - Posted: 4 May 2015, 8:49:20 UTC

Hi Dayly Diamond!
I want to give you some general ideas to help understanding the scope of the project.

The main goal of the project is to create a network where researchers in cardiac electrophysiology could simulate large amounts of variations of a model.

With this idea, we will lunch several subprojects for specific tasks, goals and expected results.

We have several problems to solve, most of them being solved in computation clusters until now. With this first tests we are learning and selecting wich ones are the best ones to run in DENIS.

The first one we will launch is the one explained by Joel. To do better models and to know how these models represent physiological variations of different parameters we need to simulate them. This problem is potentially infinite, for this reason, we want to generate a big number of simulations uniformly distributed in the space of solutions that will permit us to approximate the results with a response surface. Until now we have been working with 5 parameters, but with DENIS we want to modify 14 parameters at the same time (the ionic conductances). There is a simple relation: more simulations, better response surface, better decissions creating a new model.

With the results of this simulations we will create a new version of our human ventricular cell model. A good model is the base for good in silico research.

To modify a model is slow work and in which we must take many things into consideration. Concrete results will be forthcoming, but we will inform you of partial results (when the simulations of a marker are enough to model it, new markers, etc.)

We are planning other subprojects but we will talk about them when they will be closed.

We are preparing a web page explaining this ideas. With the new version we will also publish this information to give you feedback of our results.

Thank you for your interes, Jesús.

[VENETO] boboviz
Send message
Joined: 9 Apr 15
Posts: 108
Credit: 403,294
RAC: 0
Message 176 - Posted: 19 May 2015, 8:56:04 UTC

Any news about checkpoint??

Profile jcastro
Volunteer developer
Volunteer tester
Project scientist
Avatar
Send message
Joined: 16 Mar 15
Posts: 218
Credit: 14,859
RAC: 0
Message 181 - Posted: 19 May 2015, 23:22:09 UTC

Hi,

Yes, tomorrow we will launch the new version of the app with checkpoints and MAC OS X support.

Thanks for your attention, joel

[VENETO] boboviz
Send message
Joined: 9 Apr 15
Posts: 108
Credit: 403,294
RAC: 0
Message 443 - Posted: 2 Sep 2015, 15:54:34 UTC - in response to Message 165.

We are preparing a web page explaining this ideas. With the new version we will also publish this information to give you feedback of our results.


Thanks to optimized app, you crunch a lot of wu...any preliminary results?
A page with results is welcome!

Profile Chus Carro
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 18 Mar 15
Posts: 82
Credit: 424,346
RAC: 0
Message 446 - Posted: 3 Sep 2015, 8:22:00 UTC
Last modified: 3 Sep 2015, 8:22:25 UTC

Hi boboviz,
First of all, we have to apologize for not created yet the page with the information of the research in progress. We have to put it on the top of our ToDo list.

We have been studing for what Electrophysiological research problems DENIS is better. We have detected some problems that are perfect for volunteer computing (models populations and related problems), other ones that are good but they need some modifications in the algorithms (problems using surface response optimization) and others that if the number of parameters to study is too big are very good also (sensitivity analysis and experimental design). Now we are preparing a publication to show this results and to encourage other researchers to use DENIS.

At the same time we have been studing the behavior of the CRLP model under many different situations. This model has some limitations and we want to find what things should be modified to solve them. Thanks to the lots of wu you are able to do, we have added many more variables to study and many more situations (for example, in these days we are studing the behavior under different stimulation frequencies).

We are not very sure, but our initial results point to the Intracellular Calcium Current. In the next weeks we will lunch new experiments to analyze it in detail (we have to prepare some new codes).

Thank you for the interest and for your support.

Best!
Jesús.
____________
Jesús Carro
San Jorge University
@ChusCarro


Post to thread

Message boards : Science : Work Planning


Main page · Your account · Message boards


Copyright © 2019 Universidad San Jorge