关键词：PAC学习模型；CNFs；参数f
摘 要：The thesis explores efficient learning algorithms in settings which are more restrictive thanthe PAC model of learning in one of the following two senses: (i) The learning algorithmhas a very weak access to the unknown function, as in, it does not get labeled samples for theunknown function (ii) The error guarantee required from the hypothesis is more stringent than the PAC model.