 |
|
| FriendFeed is the easiest way to share online. Learn more » |
FACETS CodeJam #3
A group for live discussion/comments on the 3rd FACETS CodeJam meeting http://neuralensemble.org/meeting...
Eilif is banging on about silver bullets again (-;
Steve Ballmer's love of developers: http://www.youtube.com/watch...
- Andrew Davison
Mihai Petrovici is presenting the FACETS Demonstrator
probably a year to wait for the wafer-scale hardware, so need virtual hardware for testing/developing software
- Andrew Davison
"simulating the emulator"
- Andrew Davison
Eric Müller is presenting the FACETS neuromorphic hardware systems
will create a dummy PyNN module for virtual hardware
- Andrew Davison
in need of reset() support in PyNN, since setup (calculating the routing, etc.) can take up to an hour
- Andrew Davison
Andrew Davison is presenting the recent development in PyNN
upps. what happend to NeuroTools website?! http://neuralensemble.org/trac...
Stefan Lang is presenting NeuroDUNE - modeling and simulation of spatially-extended neuronal systems
also see http://frontiersin.org/confere...
- Andrew Davison
is working with Bert Sakmann
- Andrew Davison
Dejan Pecevski is presenting the latest news on the PCSIM simulator (http://www.igi.tugraz.at/pcsim...)
Bidirectional Python interface lets you, for example, run a simulation with multiple simulators (e.g. PCSIM and Brian) at the same time.
- Andrew Davison
Predicate-based connectivity patterns are very powerful
- Andrew Davison
The NeuralEnsemble server is getting "CodeJammed" ... i.e. there is an encouraging spike in client activity ... we're addressing the problem ...
Christian Garbers is presenting FIND (Finding Information in Neural Data) http://find.bccn.uni-freiburg.de/
Build a Swiss army knife, but not a cowsheepchickenpig
- Andrew Davison
We should hassle hardware vendors to provide Neuroshare (http://neuroshare.sourceforge.net) libraries for Linux, not just Windows
- Andrew Davison
There is a (Windows-only) Python interface for Neuroshare
- Andrew Davison
Samuel Garcia is presenting OpenElectrophy, and its coordination with NeuroTools
Samuel presents the unified core object model for OpenElectorphy, FIND, NeuroTools. Development will be hosted at http://neuralensemble.org/trac...
- emuller
There was some enthusiasm for improving the NeuroTools plotting module, supporting multiple backends (Chaco, etc) as well as matplotlib, having a "plot_everything()" function which intelligently decides the best way to visualise something.
Chris Ball is talking about extending Python attributes with Parameters and Traits
He's making a persuasive case for using Parameters within PyNN
- Andrew Davison
although he says we should actually use Traits rather than Parameters
- Andrew Davison
Stefan Behnel is talking about Cython
also develops lxml
- Andrew Davison
Cython is a fork of Pyrex
- Andrew Davison
Fortran support is coming
- Andrew Davison
He is now doing a very impressive demo using a Sage notebook to show the speed-up you can get - ended up 180 times faster
- Andrew Davison
Apple dropped support for Carbon in 64bit system, which stops Neuron GUI (InterViews) from compiling, but you can still switch to X11 ;)
It would be nice, if the ConPlotter could distinguish feed-forward and feedback connections
Hans Ekkehard Plesser is talking about visualizing neuronal network connectivity with ConnPlotter
connectivity pattern tables (CPTs), plain and aggregated
- Andrew Davison
ConnPlotter, a Python application for generating CPTs
- Andrew Davison
55th anniversary of first neural simulation (farley, clark, 1954)
Do we need new programming paradigm for parallel processing?
Marc-Oliver Gewaltig is talking about reproducible model descriptions
"models cannot be verified, they can only be falsified"
- Andrew Davison
Greg Wilson's blog is at http://pyre.third-bit.com/. His research seems to be about how scientists use computers and how to help them use them better.
- Andrew Davison
link to Donoho paper: http://ieeexplore.ieee.org/search...
- Andrew Davison
"model descriptions should go in the Results section, not Methods, since the model is a product of significant scientific analysis and synthesis"
- Andrew Davison
Nordlie article: http://dx.doi.org/10...
- Andrew Davison
LaTeX source code for model description tables: http://www.nest-initiative.org/images...
- Andrew Davison
Andreas Klöckner is talking about PyOpenCL (http://mathema.tician.de/softwar...)
OpenCL generalizes beyond GPUs to multicore CPUs, accelerator boards
- Andrew Davison
trade-off of performance against programmer time - like writing C instead of Python, or assembler instead of C.
- Andrew Davison
Matrix and vector operations are fast, but any control structure (branching) is expensive.
- Marc-Oliver Gewaltig
(Py)OpenCL is a rather thin wrapper around what the device offers. Could hide a lot of the book-keeping work by writing a higher-level interface.
- Andrew Davison
There is an intriguing note written on the blackboard: "Casino or Library". I can't think when that might be a likely choice of activities?!
This is just a leftover from a course we had in this lecture hall past evening :-) Forgot to destroy all evidence, unfortunately.
- ZYV
@neurotools Did anybody of you look at this: http://www.bionet.ee.columbia.edu/code...
Eilif Muller is presenting a tutorial on mpi4py (http://mpi4py.scipy.org/)
I think one problem is that different simulators prefer (can use/compile easily with) different MPI implementations.
- Andrew Davison
have to manually match up Numpy data types and MPI data types. Will be fixed in next release?
- Andrew Davison
are the slides going to be available online?
- btel
ipcluster http://ipython.scipy.org/doc...
- Andrew Davison
btel: yes, we plan to make the slides for all talks available online if their authors agree.
- Andrew Davison
If you want to read Herb Sutter's paper goto http://www.gotw.ca/publica...
Michael Hines is presenting recent developments for the NEURON simulator
supporting Python 3.x needs to be worked on
- Andrew Davison
the Section class constructor now accepts 'name' and 'cell' keyword arguments. The latter tells a Section which Cell object it belongs to. e.g. self.soma = h.Section(cell=self)
- Andrew Davison
it would be useful to have some guidance on how to integrate the event loop of Python GUI toolkits with NEURON
- Andrew Davison
Jochen Eppler presents the NEST live CD and proposes to add other free Neuroscience software. Maybe we can win the INCF to issue a live Software Repository (as CD and VMWare/VirtualBox Image.
Jochen Eppler is presenting recent developments in the NEST simulator
The new multimeter and a new init/reset framework look particularly interesting
- Andrew Davison
tests are structured (self tests, unit tests, regression tests, MPI tests) and run in a hierarchy from simple to complex
- Andrew Davison
Mikael Djurfeldt is presenting the MUSIC project http://software.incf.org/softwar...
options for creating a Python wrapper for the MUSIC C++ library? SWIG, SIP, Boost.Python, Cython, hand-rolled?
- Andrew Davison
benchmarks show that using MUSIC does not appreciably slow down simulations
- Andrew Davison
will soon be available in Debian/Ubuntu
- Andrew Davison
there was a discussion on the interaction between the buffering done by MUSIC and the buffering done by individual simulators.
- Andrew Davison
Welcome to the CodeJam meeting, and to the first NeuroTools session.
Marc-Oliver Gewaltig commented that the NeuroTools documentation needs to include an explanation of each algorithm, perhaps by having a wiki page for each non-trivial function/method.
- Andrew Davison
Bartosz Telenczuk commented that we need lots more examples of how to use NeuroTools
- Andrew Davison
It seems as though there is considerable overlap between the Spike2 and NeuroExplorer approaches to storing experiment data, so it is definitely worth the effort trying to find the common ground and integrate this fully into NeuroTools.
- Andrew Davison
There was a discussion on working with Neuroshare (http://neuroshare.sourceforge.net/index...), which unfortunately seems to be Windows-only
- Andrew Davison
If you look at he Neuroshare page, you see that they started from a similar initiative as NeuroTools. It may be worthwhile to contact them. Since they want to cater a wide audience, they may be willing to consider other platforms.
- Marc-Oliver Gewaltig
Luc Estabanez would like to have a persistent environment for data analysis, as with Origin or Igor Pro. He suggested using NeuroTools together with Qtiplot. Bartosz suggested trying Sage.
- Andrew Davison
Eilif Muller presented the stgen module. Future plans include an inhomogeneous Ornstein-Uhlenbeck process, and optimizations using Cython.
- Andrew Davison
Other ways to read this feed:
