www.neuron.yale.edu

We've started a process of redoing those tutorials, which are pretty old. The new versions use more modern idioms and come as Jupyter notebooks. We haven't moved these to the website because we don't have a full replacement, but here's a few: https://neuron.yale.edu/ftp/ramcdougal/pythontutorial.ipy...

Are you trying to debug NEURON itself or a NEURON simulation? For debugging NEURON simulations written in Python, you can use pdb as with any other Python code. A quick search will lead you to many pdb tutorials. For debugging NEURON simulations written in HOC... there's no equivalent. Going forward...

Calling h.fadvance() advances the simulation by one timestep. When using fixed-step integration, the time step size is controllable by setting h.dt ; by default it is 1/40 ms. When controlling a simulation in this way, before beginning your advances, you must explicitly initialize the simulation wit...

Internally, the Sections are stored in a linked list. There is thus no direct, consistent-across-runs section indexing scheme... but as you've observed, there's no advantage for NEURON storing such a list since you can create it quickly in the non-parallel case : seclist = list(h.allsec()) Once you ...

Unfortunately, this isn't in 7.7; we considered it an internal class and didn't realize people were using it otherwise. The general workaround is to use the corresponding section method trueparentseg (there's also a parentseg); e.g. >>> from neuron import h >>> soma = h.Section(name='soma') >>> dend...

Are you trying to simulate ephaptic coupling or extracellular diffusion of ion concentrations in the extracellular space? There are currently at least two models in ModelDB that use NEURON for ephaptic coupling: http://identifiers.org/modeldb/3676 and http://identifiers.org/modeldb/240957 (The rxd m...

The literal translation is adapex=h.ADEX(0.5) Since this is technically a point process, the 0.5 means it gets associated with the middle of some section, the so called currently accessed section. This is fine, but in Python we can be more explicit about which section, here mysec: adapex=h.ADEX(myse...

I've added a full working code (and the associated image output) to the calcium wave rxd tutorial page: https://neuron.yale.edu/neuron/docs/reaction-diffusion-calcium-wave There's not much in the way of explanation yet, but hopefully after working through the rest of the existing tutorials ( https:/...

Two questions: Does NEURON work from a regular python prompt (i.e. not Spyder) What is the full NEURON part of your PYTHONPATH? Because of reasons NEURON installs two versions of the library. Only one has the "missing" file. In particular, on my machine the two paths are: /usr/local/nrn/lib/python/n...

Ted raises good points. The underlying issue though is that if you run the straightforward translation of that mod file, you'll get a maximum value of 1 (assuming your concentrations are below 1). Why 1? It has to do with the order of events during initialization: NEURON sets concentrations to 1 unl...

To fix, add the following two paths to your system PATH environment variable (for instructions on how to modify PATH, see https://www.architectryan.com/2018/03/17/add-to-the-path-on-windows-10/ ): c:\nrn\bin c:\nrn\mingw\usr\bin We've identified the source of the 1024 limit, and this should be fixed...

This is likely due to a previously under-appreciated detail in the numerical methods used for 3D simulation. In particular, we use scipy.sparse.linalg.bicgstab with a tolerance of 1e-5 to solve a matrix problem. This should have been connected to the atol used elsewhere*, but it is not. For now, you...

Adaptive integration chooses timesteps to constrain error.

In particular, unlike fixed step integration, it will adapt its timesteps to handle events at the time received and not shift them or group them.

Conceptually, you are correct. For memory reasons, however, NEURON's hh.mod doesn't actually store those values since they are only useful for computing one location at one time point. Instead, a GLOBAL value is used, meaning that there is only one memory location used for every spatial point (diffe...

Globals are accessed globally; i.e. through the h object: