I trained iris dataset (including 3 classes) with one-vs-all svm classifier and print coef_ and intercept_ as bellow:

`[ [-0.04625854 0.5211828 -1.00304462 -0.46412978] // 1-vs-all coefficients [-0.00722313 0.17894121 -0.53836459 -0.29239263] // 2-vs-all coefficients [ 1.15034043 1.14954525 -3.53985244 -4.24622393]] //3-vs-all coefficients [ 1.4528445 1.50771313 13.63764975] //intercepts `

Then I created ovr iris datasets. I labeled specific class as 1 and the other two classes as 0. again I trained whole dataset with same classifire and print coef_ and intercept. here are results:

`1-vs-all: [[-0.04575352 0.52216766 -1.00294058 -0.46406882]] [ 1.44746413] 2-vs-all: [[-0.03070975 -2.38286314 1.13998914 -2.61285489]] [ 5.48399354] 3-vs-all: [[-1.15034043 -1.14954525 3.53985244 4.24622393]] [-13.63764975] `

As you can see absolute value of 1-vs-all and 3-vs-all results are same in both experiments but for 2-vs-all it is completely different. I can’t recognize why is this happening?