ALGORITHM OF PRUNING OF HYBRID NEURAL NETWORKS ENSEMBLE
Journal Title: Електроніка та системи управління - Year 2018, Vol 1, Issue 55
Abstract
Despite the fact that the ensemble is usually more accurate than a single network, existing ensemble techniques tend to create unreasonably large ensembles that increase the use of memory and computation costs. The ensemble's pruning solves this problem. The article analyzes the compromise between accuracy and diversity and it is proved that classifiers, which are more accurate and make more predictions in the minority group, are more important for the construction of the subensemble. A metric that takes into account accuracy and diversity is proposed to evaluate the contribution of a separate classifier that will help to allocate the required number of networks with the best results.
Authors and Affiliations
Olena Chumachenko, Anastasia Kuzmenko
Keywords
Related Articles
- EP ID EP523373
- DOI 10.18372/1990-5548.55.12772
- Views 125
- Downloads 0
VOICE CONTROL COMMAND RECOGNITION SYSTEM OF UAV BASED ON STEGANOGRAPHIC-CEPSTRAL ANALYSIS
The article suggests the developed voice command recognition system based on steganographic-cepstral analysis and the computational algorithm of mel-frequency cepstral coefficients which is used for creation the basic vo...
INERTIAL NAVIGATION ACCURACY AND ACCELEROMETERS WITH DAMPING
Nowadays an anxiously important problem is the creation of the low-cost gyro-free inertial navigation systems. That is why we focus primarily on linear accelerometer to avoid gyro using. The problem of inertial navigatio...
BASIC PROBLEMS OF DESIGN OF REDUNDANT NONCOLLINEAR CONFIGURATIONS WITH INERTIAL SENSORS
The paper deals with basics tasks of the design of redundant noncollinear inertial measuring units based on the technology of micrielectromechanical systems. Analysis of noncollinear configurations of inertial sensors i...
DIAGNOSTIC SYSTEM BASED ON AUTOENCODERS
It`s considered the problem of image processing which is used in diagnostic systems when it is necessary to process the results of ultrasound, computed tomography and magnetic resonance imaging. For the solution of this...
COMPARISON OF NUMERIC PROPERTIES OF OBJECTS OF DIFFERENT DATA DOMAINS IN RELATIONAL DATABASES
The problem of association of models of data domains is considered. It is offered to compare objects of data domains on the basis of values of properties of copies of those objects. Methods of benchmarking properties dif...