Incorrect behaviour in repair mode 1 (flip observed variations) for a simple toy network

[akittas]

The output of ingranalyze is not correct when analyzing a simple toy network with repair mode 1. More in specific:

Assume network:
A -> B +
B -> C +

with observations:
A = +
B = +
C = -

In this case the network is inconsistent.

Possible repair sets are:

1. {vFlip(A), vFlip(B)} and 2) {vFlip(C)}.
   So only one minimal repair set (2), which should become the predictions.

In contrast, running ingranalyze with he selected network gives no predictions and suggests as repair sets the following:

Computing all repair sets with size 1 ... done.
repair 1 :
vflip("Network.obs",gen("C"),-1)
repair 2 :
vflip("Network.obs",gen("B"),1)

Computing predictions that hold under all repair sets size 1 ... done.
0 predictions found:

It therefore gives 2 minimal repair sets (instead of 1) and 0 predictions (instead of 1). In contrast, repair mode 2 works correctly with this example.

[sthiele]

Hi,
there seems to be a problem indeed.
Please try our new version(still under development)
https://pypi.python.org/pypi/iggy
that should fix this problem.
It gives the correct answer one repair,
~/.local/bin/iggy.py net.sif obs.txt --show_predictions --show_colorings 10 --scenfit --autoinputs

Reading network net.sif ... done.
Reading observations obs.txt ... done.
Computing input nodes ... done.
Computing scenfit of network and data ... done.
The network and data are inconsistent: scenfit = 2.
Compute scenfit colorings... done.
Coloring 1:
gen("C") = 1
gen("B") = 1
gen("A") = 1
flip("obs.txt",gen("C"),2)

Compute predictions under scenfit ... done.

Predictions are still empty because we only count nodes which have initially no color as predictions.

Best regards and thanks for using our software,
Sven

[akittas]

Many thanks for the quick reply. We are using bioasp and ingranalyze as a back-end to a software package we are building, so I was wondering if this is an isolated case or a more general problem.

Is it possible to pinpoint the source of the problem and patch the behaviour of ingranalyze? Could it be in of the functions: get_minimum_of_repairs or get_minimal_repair_sets?

Thanks and best regards,
Telis

[sthiele]

Sure, i will have a look into it on monday. I keep you posted on the progress.

[sthiele]

Hi Telis,
I think I fixed the problem. Maybe you can test the current master branch to see if it works for you.
Best regards,
Sven

[akittas]

Hi, I tested the new version, it works as expected for the test case outline below. Looks like the bug is fixed. Many thanks for the quick response!