Convergency of Learning Process
نویسندگان
چکیده
This paper presents a learning theoretic analysis on stability of learning in light of iterated belief revision. We view a learning process as a sequential belief change procedure. A learning policy is sought to guarantee every learning process leads to a complete knowledge about the world if the newly accepted information is the true fact on the world. The policy allows an agent to abandon the knowledge it has learned but requires a relatively moderate attitude to new information. It is shown that if new information is not always accepted in an extremely skeptical attitude and the changes of belief degrees follow the criterion of minimal change, any learning process for learning truth will converge to a complete knowledge state.
منابع مشابه
A Convengent Solution to Tensor Subspace Learning
Recently, substantial efforts have been devoted to the subspace learning techniques based on tensor representation, such as 2DLDA [Ye et al., 2004], DATER [Yan et al., 2005] and Tensor Subspace Analysis (TSA) [He et al., 2005]. In this context, a vital yet unsolved problem is that the computational convergency of these iterative algorithms is not guaranteed. In this work, we present a novel sol...
متن کاملNon-Negative Semi-Supervised Learning
The contributions of this paper are three-fold. First, we present a general formulation for reaping the benefits from both non-negative data factorization and semi-supervised learning, and the solution naturally possesses the characteristics of sparsity, robustness to partial occlusions, and greater discriminating power via extra unlabeled data. Then, an efficient multiplicative updating proced...
متن کاملNonlinear Blind Equalizers: NCMA and NMCMA
This paper proposes two nonlinear blind equalizers: the nonlinear constant modulus algorithm (NCMA) and the nonlinear modified constant modulus algorithm (NCMA) by applying a nonlinear transfer function (NTF) into constant modulus algorithm (CMA) and modified constant modulus algorithm (MCMA), respectively. The effect of the NTF on CMA and MCMA is theoretically analyzed, which implies that the ...
متن کاملStudy of Iterative Processes in Computer Graphic's Problems
The convergency of iterative processes can exhibit unexpected behaviours. In this paper an analysis of the convergency of the iterative processes is made through the application of the Newton method to the polynomial equations resolution. This leads to the deenition of a dynamic relaxation scheme applied to computer graphics
متن کاملAPPLICATION OF THE SINGULAR BOUNDARY VALUE PROBLEM FOR INVESTIGATION OF PISTON DYNAMICS UNDER POLYTROPIC EXPANSION PROCESS
In this paper a mathematical simulation of a simplified internal combustion engine is presented. To contribute engine kinematics and its geometry, simple relations are derived for constrained motions. The equation of motion for the piston forms a singular boundary value problem. The uniqueness of the solution was studied in the Banach space. For solving governing equations an iterative numerica...
متن کامل