Porting fragment code Python to Julia

General Usage
#1 
selection = np.linspace(0,3,5000000)

eps = 0.9

# spikes numpy 1d array

ee_counter = np.array( [ ( ( spikes >= selection[j] ) & ( spikes < selection[j]+eps ) ).sum() for j in range( len( selection ) ) ] )

pdf = ee_counter / spikes.sum()

I have this fragment code on python. How can i porting this code on Julia with minimal number of cycles and rows of codes

#2

Can you please copy and paste the code as text instead of a screen shot? Also, remember to put it inside triple backticks.

#3

Please post editable code instead of images. Anyway, this might be a direct translation:

selection = range(0,3,5000000)
ϵ = 0.9

ee_counter = [sum(i->s<=i<s+ϵ, spikes) for s in selection]
pdf = ee_counter ./ sum(spikes)
#4

I edited, sorry.

1 Like
#6

I edited topic. I tried this and has next error

image
image1102×589 38.6 KB

Spikes is vector
image

#7

sum and count do the same thing here, but using count makes more sense, imho, and is more likely to discover bugs.

#8

True, but sometimes I prefer sum as it allows simd optimizations more than the other alternatives.

#9

This helped, but the calculation takes a long time. This code on python ran for 1 minute. Length of spikes 3098

After 5 minutes of calculations, I aborted ran code

#10

You must put the code in a function, and make sure to avoid global variables.

#11

The code you provided is still not self-contained, it isn’t a ‘minimal working example’ or MWE. Can you provide some example data for spikes?

Also, when developing and testing code, I suggest that you make smaller data sets. Waiting for minutes just to see if your code works seems like a waste of time.

#12

It takes 3s in Julia:

function sum_spikes(spikes)
    ϵ = 0.9
    ee_counter = [sum(i->s<=i<s+ϵ, spikes) for s in range(0,3,5000000)]
    pdf = ee_counter ./ sum(spikes)
end

spikes = randn(3098)
@btime sum_spikes($spikes)  # 3.237 s (4 allocations: 76.29 MiB)
4 Likes
#13

It helped, thank you