torstenneher / gridtoplacemodel Goto Github PK

I added some patches to the code to get this at least producing some figues on Ubuntu 21.10. The forked version of the code is here. The following software versions were used:

python virtualenv (3.9.4-1build1)
python (2.7)
packages:
numpy (1.22.2)
matplotlib (3.5.1)
scipy (1.8.0)
pickle-mixin (1.0.2)
sklearn (1.0.2)

I did not create a pull request because I was only able to generate figures 1, 2, 3, and 1/2 of 5. Issues creating other figures were:

function: problematic()
Error: self.distance_matrix[int(np.subtract((self.distance_matrix <= bins[i]),(self.distance_matrix <= bins[i-1]),dtype=np.float32))] = bins[i] #every entry that is <= bins[i] but without beeing <= bins[i-1] becomes bins[i]
TypeError: only size-1 arrays can be converted to Python scalars

Figure 5 and 6A (hebbPaperNew()) and
Figure 7 (hebbPaperStabilityErr())
Error: self.distance_matrix[int(np.subtract((self.distance_matrix <= bins[i]),(self.distance_matrix <= bins[i-1]),dtype=np.float32))] = bins[i] #every entry that is <= bins[i] but without beeing <= bins[i-1] becomes bins[i]
TypeError: only size-1 arrays can be converted to Python scalars

Figure generating functions working:

MethodsPaper() (Fig. 1)
SVCPaper() (Fig. 2)
SVCPaperOneMod() (Fig. 3)
hebbPaperLesionsErr() (Fig. 5 panels C, D, and E)

I think the patches added may make the code run with equivalent results to the original version. One patch I made was changing if self.patterns_1.shape[-1] == 1 into self.patterns_1.shape[-1].all() == 1. I don't know if this affects results.

Another approach to getting the original code to work would be creating a virtual machine with I assume Ubuntu 15.04 and installing old Python packages. I preferred to work outside of a virtual machine so I patched the code.