Learning DNF by Statistical and Proper Distance Queries Under the Uniform Distribution
نویسنده
چکیده
We show that s-term DNF formulas can be learned under the uniform distribution in quasi-polynomial time with statistical queries of tolerance Ω(ε/s). The tolerance improves on the known tolerance Ω(ε/s) and is optimal with respect to its dependence on the error parameter ε. We further consider the related model of learning with proper distance queries and show that DNF formulas can be learned under the uniform distribution with quasi-polynomial queries, where the hypotheses are DNF formulas of polynomial size. Finally we consider the class of majorities over DNF formulas and provide polynomially related upper and lower bounds for the number of distance queries required to learn
منابع مشابه
Learning with Queries Corrupted by Classification Noise
Kearns introduced the “statistical query” (SQ) model as a general method for producing learning algorithms which are robust against classification noise. We extend this approach in several ways in order to tackle algorithms that use “membership queries”, focusing on the more stringent model of “persistent noise”. The main ingredients in the general analysis are: (1) Smallness of dimension of th...
متن کاملA Query Algorithm for Agnostically Learning DNF?
Motivation: One of the most celebrated results in computational learning theory is Jackson’s query algorithm for PAC learning DNF formulas with respect to the uniform distribution [3]. A natural question is whether DNF formulas can be learned (even with queries and with respect to the uniform distribution) in a highly noisy setting, i.e., the wellknown agnostic framework of learning [5]. Additi...
متن کاملA Subexponential Exact Learning Algorithm for DNF Using Equivalence Queries
We present a 2 ~ O(p n) time exact learning algorithm for polynomial size DNF using equivalence queries only. In particular, DNF is PAC-learnable in subexponential time under any distribution. This is the rst subexponential time PAC-learning algorithm for DNF under any distribution.
متن کاملAdvanced Topics in Computational Learning
The question of whether DNF formulas can be learned in the PAC model, is one of the most studied questions in the field of computational learning. Many negative results have been published since this question was first posed by Valiant in 1984, and until recently it was not known whether DNF can even be learned with the help of membership queries. This was the situation until Jackson made the b...
متن کاملEfficiency and Computational Limitations of Learning Algorithms
This thesis presents new positive and negative results concerning the learnability of several well-studied function classes in the Probably Approximately Correct (PAC) model of learning. Learning Disjunctive Normal Form (DNF) expressions in the PAC model is widely considered to be the main open problem in Computational Learning Theory. We prove that PAC learning of DNF expressions by an algorit...
متن کامل