The Minimum Transfer Cost Principle for Model-Order Selection
نویسندگان
چکیده
The goal of model-order selection is to select a model variant that generalizes best from training data to unseen test data. In unsupervised learning without any labels, the computation of the generalization error of a solution poses a conceptual problem which we address in this paper. We formulate the principle of “minimum transfer costs” for model-order selection. This principle renders the concept of cross-validation applicable to unsupervised learning problems. As a substitute for labels, we introduce a mapping between objects of the training set to objects of the test set enabling the transfer of training solutions. Our method is explained and investigated by applying it to well-known problems such as singular-value decomposition, correlation clustering, Gaussian mixturemodels, and k-means clustering. Our principle finds the optimal model complexity in controlled experiments and in real-world problems such as image denoising, role mining and detection of misconfigurations in access-control data.
منابع مشابه
A Multi-objective Mathematical Model for Sustainable Supplier Selection and Order Lot-Sizing under Inflation
Recently, scholars and practitioners have shown an increased interest in the field of sustainable supplier selection and order lot-sizing. While several studies have recently carried out on this field, far too little attention has been given to formulating a multi-objective model for the integrated problem of multi-period multi-product order lot-sizing and sustainable supplier selection under i...
متن کاملOptimal Coding Subgraph Selection under Survivability Constraint
Nowadays communication networks have become an essential and inevitable part of human life. Hence, there is an ever-increasing need for expanding bandwidth, decreasing delay and data transfer costs. These needs necessitate the efficient use of network facilities. Network coding is a new paradigm that allows the intermediate nodes in a network to create new packets by combining the packets recei...
متن کاملA Multiple Objective Nonlinear Programming Model for Site Selection of the Facilities Based on the Passive Defense Principles
One of the main principles of the passive defense is the principle of site selection. In this paper, we propose a multiple objective nonlinear programming model that considers the principle of the site selection in terms of two qualitative and quantitative aspects. The purpose of the proposed model is selection of the place of facilities of a system in which not only it observes the dispersion ...
متن کاملA new method for fuzzification of nested dummy variables by fuzzy clustering membership functions and its application in financial economy
In this study, the aim is to propose a new method for fuzzification of nested dummy variables. The fuzzification idea of dummy variables has been acquired from non-linear part of regime switching models in econometrics. In these models, the concept of transfer functions is like the notion of fuzzy membership functions, but no principle or linguistic sentence have been used for inputs. Consequen...
متن کاملSupply chain optimization policy for a supplier selection problem: a mathematical programming approach
Most supplier selection models consider the buyer’s viewpoint and maximize only the buyer’s profit. This does not necessarily lead to an optimal situation for all the members of a supply chain. Coordination models have been developed to optimize the entire supply chain and align the decisions between its entities. Little research has been done on the application of these models in the supplie...
متن کامل