输入:
X(patches 特征), Y(patches特征的“正负”标号), wIn(向量) ,weights(权重alphas,每个鉴别类的权重,要想只求一个,其他的设为0,要求的设为1),roundid(不用管了吧?)
输出:
wOut outLabels
weights (alphas):If the weights (alphas) aren't set, then set them to all ones
bias term b
nrmlambda
normquad
dall=nabla'*X