Can Neural Networks or any other supervised-learning algorithm learn special statistical methods? -
i understand neural networks number of hidden layers can approximate nonlinear function, can predict special functions, same statistical methods?
suppose statistical rule classification problem follows. training set input x_train , output y_train, calculate geometrical average (i.e. center of x_train per particular class) of x_train belonged each particular class. therefore, each class have center of x. test data, estimate class labels finding shortest euclidean distance trained centers. example, assuming training give centers following mapping: (-1,1,1)->0, (1,1,1)->1. test data (-0.8,0.5,1), since closer (-1,1,1), should belong class 0.
the problem not know if supervised learning method can above strategies. call 'supervised k-means'. knn method similar finds label based on n-nearest-points rather average of training points.
i wondering if neural networks can this. or miss other learning techniques can above strategy? if statistical strategy try learn more complex, example including both center , co-variance?
Comments
Post a Comment