This means, that optimization algorithm detected that with high probability (not in the strict, mathematical sense) you can speed up your training by turning the -h 0
flag in your options. Basically, -h
is the shrinking heuristics, implemented in the libsvm package which for some data significantly reduces number of required computations, while in others - makes it slower. There is no general rule, but there are some heuristics regarding this heuristic, which "detect" when it could help - and this is exactly the reason for this message - one of such "metaheuristics" detected, that for your particular data and parameters - it could be more valuable to set the -h 0
flag.
If you are interested in more details regarding this possible optimization speed up you could refer to the source paper:
T. Joachims, Transductive Inference for Text Classification using Support Vector Machines. International Conference on Machine Learning (ICML), 1999.