http:/parnec.nuaa.edu.cn A toy problem minSt dataset(2 vS. 5) as C(;, ui)=e 2o=lu,-uylIPa Figure 2: A toy example. Result weights of meta 05 features based prior vs. max-margin when training from two instances. (a, b Gray level of the two digits used for training. (c) The weight per pixel of max margin classifier(white is positive, black is negative) (d) The weights when using a Gaussian process prior o5 on meta-features to weight mapping. The accuracy improved from 82% to 88% relative to max-margin 2
Company name www.themegallery.com A toy problem MINIST dataset (2 vs. 5)
http:/parnec.nuaa.edu.cn Incorporating prior knowledge on features into learning (A/STATS'O7 ● Motivation KErnel design by meta-features ● A toy examp le o Handwritten digit recognition aided by meta-features O Towards a theory of meta-features
Company name www.themegallery.com Incorporating prior knowledge on features into learning (AISTATS’07) ⚫ Motivation ⚫Kernel design by meta-features ⚫ A toy example ⚫ Handwritten digit recognition aided by meta-features ⚫ Towards a theory of meta-features