“The output of different SVM classifiers can be combined simply by a weighted average of the estimates they produce. We combined in this way an SVM classifier which uses the set of color features F1 ∪ F2 ∪ F3 (with weight 1/3) and a second SVM classifier which uses the set of texture features G2 (wit…”— Philippe Golle, xenon.stanford.eduTagged: Machine Learning, Dogs vs Cats, Image Classification
“The success of our classifier does not come from the careful selection of a few colors with high predictive values, but rather from the combination of a large number of weakly predictive features.”— Philippe Golle, xenon.stanford.eduTagged: Machine Learning, Dogs vs Cats, Image Classification
“With a 60% accurate classifier, the probability of solving a 12-image Asirra challenge is only about 0.2%.”— Philippe Golle, xenon.stanford.eduTagged: Machine Learning, Dogs vs Cats, Image Classification