Convert python array to neuron vector

When Python is the interpreter, what is a good
design for the interface to the basic NEURON
concepts.

Moderator: hines

Post Reply
Prokopiou

Convert python array to neuron vector

Post by Prokopiou » Thu May 03, 2012 2:45 pm

Hello,

I have been trying to convert a python array ( numpy.ndarray ) into a neuron vector.

I used the command:

Code: Select all

neuron_vector.from_python(python_array)
and it does work (dimension-wise) but doesn't get the correct values. The only value the neuron vector stores is 0 but I noticed when the sign changes in python so does in neuron vector. So I have a vector full of -0 and 0.

This makes me think that there must be an error in the bits the number is represented, each value in the python array is a 'numpy.float64'.

Is there an issue in transferring 64 bit numbers using the from_python() function?

Thanks

Prokopiou

Re: Convert python array to neuron vector

Post by Prokopiou » Thu May 03, 2012 3:29 pm

Hello,

Please ignore my previous post, it does convert the 64bit array when I test it with a different array which confuses me even more now...

Code: Select all

stim_vec = h.Vector()
time_vec = h.Vector()

time_vec.from_python(t)

i = 0
for sec in h.allsec():
    for seg in sec:
       stim_vec.from_python(external_voltage[:,i])
       stim_vec.play(seg._ref_e_extracellular, time_vec)
       i += 1
this is the code that seems to not work...
t is an array in python which converts just fine to a neuron vector.

the external_voltage[:,i] gives back 1D array (same length as t) but the stim_vec remains at zero with alternating sign (negative-positive) when the external_voltage changes value. I cant really see why it should do this.

Thanks

hines
Site Admin
Posts: 1575
Joined: Wed May 18, 2005 3:32 pm

Re: Convert python array to neuron vector

Post by hines » Fri May 04, 2012 8:08 am

Since there is only one instance of stim_vec, I would have expected every segment to get the same last external_voltage 1d array.
Perhaps you need an array of stim_vec as in
stim_vec = []
...........
stim_vec.append(Vector(external_voltage[:,i])
stim_vec[-1].play(seg....)
However, I would not have expected that last external voltage to be alternating +-0 so perhaps there is a bug in the
transfer of the numpy 1-d array to a Vector. Your fragment does not exhibit the creation of the 2-d numpy array. A complete
fragment would help me see the problem and work on a diagnosis. The following test

Code: Select all

import numpy
a=numpy.matrix([[1,2],[3,4]])
from neuron import h
v = h.Vector(a[:,0])
v.printf()
for i in range(0,2):
  v.from_python(a[:,i])
  v.printf()
produces the correct output

Code: Select all

1	3	
1	3	
2	4	

Prokopiou

Re: Convert python array to neuron vector

Post by Prokopiou » Fri May 04, 2012 12:38 pm

I seem to have resolved the problem...in a way that doesn't make sense to me so it is still a problem I guess..

Im using a play , record pair to stimulate the cell with an external voltage and this is the code I use:

Code: Select all

time_vec = h.Vector()
time_vec.from_python(t)

test_vec = []

i = 0
for sec in h.allsec():
    for seg in sec:
       stim_vec = h.Vector()
       stim_vec.from_python(external_voltage[:,i])
       test_vec.append(stim_vec)
       stim_vec.play(seg._ref_e_extracellular, time_vec)
       
       i += 1
            
results = []
for sec in h.allsec():

    for seg in sec:  
        rec_vec = h.Vector()
        rec_vec.record(seg._ref_v)
        results.append(rec_vec)        

 
neuron.init()
neuron.run(time)

results_list = []
for v in results:
    temp = v.to_python()
    temp.remove(temp[0])
    results_list.append(temp)
    
    
internal_voltage = np.array(results_list)
this seems to work just fine, what I don't quite understand is when I comment out the test_vec like this:

Code: Select all

time_vec = h.Vector()
time_vec.from_python(t)

#test_vec = []

i = 0
for sec in h.allsec():
    for seg in sec:
       stim_vec = h.Vector()
       stim_vec.from_python(external_voltage[:,i])
       #test_vec.append(stim_vec)
       stim_vec.play(seg._ref_e_extracellular, time_vec)
       
       i += 1
            
results = []
for sec in h.allsec():

    for seg in sec:  
        rec_vec = h.Vector()
        rec_vec.record(seg._ref_v)
        results.append(rec_vec)        

 
neuron.init()
neuron.run(time)

results_list = []
for v in results:
    temp = v.to_python()
    temp.remove(temp[0])
    results_list.append(temp)
    
    
internal_voltage = np.array(results_list)
The problem I was talking about before seems to pop up again. This confuses me and makes me wonder if im using the play record pair correctly in this python code.

Thanks,
Andreas.

hines
Site Admin
Posts: 1575
Joined: Wed May 18, 2005 3:32 pm

Re: Convert python array to neuron vector

Post by hines » Fri May 04, 2012 3:52 pm

by not putting the stim_vec instance in the test_vec, then when
stim_vec = h.Vector()
is executed, the reference count of the Vector previously referenced by stim_vec goes to 0
and that deletes that Vector.

Post Reply