...SVMTorch1
SVMTorch is available at http://www.idiap.ch/learning/SVMTorch.html.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... case2
Note that this is the most common case. For instance for a Gaussian kernel with distinct examples xi, it is easy to see that it is always the case.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... 3
If we were really able to find such variables, this would mean that $(\boldsymbol{\alpha}, \boldsymbol{\alpha}^{\star})$ is a solution of our problem.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... here.4
Note also that testing whether the $\alpha_i$and $\alpha_i^{\star}$ are non-zero at the same time would be a waste of time.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... phase,5
This is verified in practice.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... authors.6
For 5000 examples, we used -clever -best -lazy, while for more than 5000 examples, we used -clever -best -lazy -ssz 200.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dataset7
Available on http://www.cs.toronto.edu/~delve/data/kin/desc.html.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dataset8
Available on ftp://ftp.ics.uci.edu/pub/machine-learning-databases/covtype/covtype.info.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dataset9
Available on http://www.research.att.com/~yann/ocr/mnist/index.html.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... kernel10
The kernel is $k(x,\, y) = \exp(-\Vert x -y\Vert^2 / \sigma^2)$.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... projection11
Actually, the original source code of this optimizer has been done by Leon Bottou.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Journal of Machine Learning Research