Libsvm model file format

Question

According to this FAQ, the format of the model in libsvm should be simple. And actually it is when I just call svm-train . As an example, the first SV for a1a is

  1 3:1 11:1 14:1 19:1 39:1 42:1 55:1 64:1 67:1 73:1 75:1 76:1 80:1 83:1 

On the other hand, if I use the easy.py script, my first SV ends:

  512 1:-1 2:-1 3:1 4:-1 5:-1 6:-1 7:-1 8:-1 9:-1 10:-1 11:1 13:-1 14:1 15:-1 16:-1 17:-1 18:-1 19:1 20:-1 21:-1 22:-1 23:-1 24:-1 25:-1 26:-1 27:-1 28:-1 29:-1 30:-1 31:-1 32:-1 33:-1 34:-1 35:-1 36:-1 37:-1 38:-1 39:1 40:-1 41:-1 42:1 43:-1 44:-1 45:-1 46:-1 47:-1 48:-1 49:-1 50:-1 51:-1 52:-1 53:-1 54:-1 55:1 56:-1 57:-1 58:-1 59:-1 61:-1 62:-1 63:-1 64:1 65:-1 66:-1 67:1 68:-1 69:-1 70:-1 71:-1 72:-1 73:1 74:-1 75:1 76:1 77:-1 78:-1 79:-1 80:1 81:-1 82:-1 83:1 84:-1 85:-1 86:-1 87:-1 88:-1 90:-1 91:-1 92:-1 93:-1 94:-1 95:-1 97:-1 98:-1 99:-1 100:-1 101:-1 102:-1 103:-1 104:-1 105:-1 106:-1 107:-1 108:-1 109:-1 110:-1 112:-1 113:-1 114:-1 115:-1 117:-1 118:-1 119:-1 

which is an instance that does not exist at all in my training set! In fact, if I do this:

  $ grep "119:" a1a -1 1:1 6:1 18:1 22:1 36:1 42:1 49:1 66:1 67:1 73:1 74:1 76:1 80:1 119:1 -1 1:1 6:1 18:1 26:1 35:1 43:1 53:1 65:1 67:1 73:1 74:1 76:1 80:1 119:1 -1 2:1 6:1 15:1 19:1 39:1 42:1 55:1 62:1 67:1 72:1 74:1 76:1 78:1 119:1 -1 4:1 6:1 16:1 21:1 35:1 44:1 49:1 64:1 67:1 72:1 74:1 76:1 78:1 119:1 -1 2:1 6:1 14:1 30:1 35:1 42:1 49:1 65:1 67:1 72:1 74:1 76:1 78:1 119:1 -1 2:1 6:1 17:1 20:1 37:1 40:1 57:1 63:1 67:1 73:1 74:1 76:1 80:1 119:1 -1 5:1 6:1 18:1 22:1 36:1 40:1 54:1 61:1 67:1 72:1 75:1 76:1 80:1 119:1 -1 5:1 6:1 17:1 26:1 35:1 42:1 53:1 62:1 67:1 73:1 74:1 76:1 80:1 119:1 

There is no instance with 119: -1 (and even if it just changes +1 to -1 , there is not a single instance with 119: 1 and 118: 1 either - the missing attributes are zeros)

If I make this change to the source code , I can clearly see that in the first case (only svm-train ) the first SV is also the first instance. But in the latter case (i.e., with the easy.py script), the output that that instance of SV should give me is grid.py

What is going on here?

Answer 1

I think the culprit here is probably the easy.py call does for svm-scale, which scales each attribute within [-1,1]. Training examples sent to svm-train will not be the same as in your training file.