FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2182982932> ?p ?o ?g. }

• W2182982932 abstract "Artificial neural networks have been applied to a variety of classification and learning tasks. The success of error correction training algorithms such as backpropagation has meant that supervised learning, where the correct outcome is known, has been the most used. In addition, although the structure of an artificial neural network is a significant contributing factor to its performance, the structure is generally heuristically chosen. The use of evolutionary algorithms as search techniques has allowed different properties of artificial neural networks to be evolved. Several successful neural/evolutionary approaches such as the EPNet [35] algorithm are hybrid algorithms combining gradient descent methods such as backpropagation learning for connection weight training with evolutionary structure search. Generally the fitness measure used in these approaches is a mean or sum squared network output error over input training patterns. This type of fitness measure is applicable to supervised learning, but not necessarily to reinforcement learning tasks where the correct output is not generally known. Evolutionary search algorithms allow greater flexibility in the choice of a suitable fitness measure. As a result, neural/evolutionary algorithms that are not reliant on gradient descent techniques and not reliant on the standard fitness measure may have wider applicability outside the supervised learning domain. The goal of this work is the implementation of an evolutionary algorithm that simultaneously searches for optimum artificial neural network structures and weights without the use a gradient-descent process. Neural network structural modifications are alternated with connection weight search/training attempting minimum disruption between parent and offspring neural network individuals whilst maintaining an effective search. The implemented algorithm is heavily based on the successful EPNet algorithm [35], applying mutation operators in a step-wise process, but does not use a backpropagation weight training algorithm. Two different fitness functions are used a standard sum squared error calculation and an accuracy calculation based on correct pattern classifications and network complexity. The aim of this is to enable a comparison of the effectiveness of the two fitness functions on the same problems. A secondary aim of this work is to determine if the implemented algorithm will continue to find smaller, less complex neural networks that generalise better after a first solution network has been found. The algorithm is tested on Even Parity 4 and 5 and the IRIS data classification problems. The results show that solution networks with low complexity can be found for Parity 4 and 5 problems using the accuracy fitness. The mean squared output accuracy performance of the solution networks was worse in comparison to the results of EPNet. Where solution networks were found early in a run, the final solutions were always less complex than the first solutions. This indicates that the use of a network complexity penalty incorporated into the fitness function was successful. The results on the IRIS data classification problem were good, however the algorithm needs to be applied to more difficult classification problems to test its scalability. Experiments on the parity problems with the standard sum squared error function were disappointing with only one solution being found for parity 4." @default.
• W2182982932 created "2016-06-24" @default.
• W2182982932 creator A5010345608 @default.
• W2182982932 date "2003-01-01" @default.
• W2182982932 modified "2023-09-27" @default.
• W2182982932 title "Evolving Neural Network Architecture and Weights Using An Evolutionary Algorithm" @default.
• W2182982932 cites W123765585 @default.
• W2182982932 cites W1481963315 @default.
• W2182982932 cites W1505554058 @default.
• W2182982932 cites W1530317173 @default.
• W2182982932 cites W1570448133 @default.
• W2182982932 cites W1571671538 @default.
• W2182982932 cites W1653273292 @default.
• W2182982932 cites W1867736342 @default.
• W2182982932 cites W1885530469 @default.
• W2182982932 cites W1978004231 @default.
• W2182982932 cites W2002016471 @default.
• W2182982932 cites W2024060531 @default.
• W2182982932 cites W2037380487 @default.
• W2182982932 cites W2047094503 @default.
• W2182982932 cites W2100666687 @default.
• W2182982932 cites W2100877722 @default.
• W2182982932 cites W2116640126 @default.
• W2182982932 cites W2118276635 @default.
• W2182982932 cites W2124290836 @default.
• W2182982932 cites W2127011380 @default.
• W2182982932 cites W2132183903 @default.
• W2182982932 cites W2134514463 @default.
• W2182982932 cites W2136271046 @default.
• W2182982932 cites W2151047121 @default.
• W2182982932 cites W2157375065 @default.
• W2182982932 cites W2161782846 @default.
• W2182982932 cites W2167662333 @default.
• W2182982932 cites W2264803162 @default.
• W2182982932 cites W2505921950 @default.
• W2182982932 hasPublicationYear "2003" @default.
• W2182982932 type Work @default.
• W2182982932 sameAs 2182982932 @default.
• W2182982932 citedByCount "7" @default.
• W2182982932 countsByYear W21829829322012 @default.
• W2182982932 countsByYear W21829829322013 @default.
• W2182982932 countsByYear W21829829322023 @default.
• W2182982932 crossrefType "journal-article" @default.
• W2182982932 hasAuthorship W2182982932A5010345608 @default.
• W2182982932 hasConcept C105795698 @default.
• W2182982932 hasConcept C105902424 @default.
• W2182982932 hasConcept C11413529 @default.
• W2182982932 hasConcept C119857082 @default.
• W2182982932 hasConcept C136389625 @default.
• W2182982932 hasConcept C153258448 @default.
• W2182982932 hasConcept C154945302 @default.
• W2182982932 hasConcept C155032097 @default.
• W2182982932 hasConcept C159149176 @default.
• W2182982932 hasConcept C175202392 @default.
• W2182982932 hasConcept C2780598303 @default.
• W2182982932 hasConcept C33923547 @default.
• W2182982932 hasConcept C41008148 @default.
• W2182982932 hasConcept C41445625 @default.
• W2182982932 hasConcept C50644808 @default.
• W2182982932 hasConcept C97541855 @default.
• W2182982932 hasConceptScore W2182982932C105795698 @default.
• W2182982932 hasConceptScore W2182982932C105902424 @default.
• W2182982932 hasConceptScore W2182982932C11413529 @default.
• W2182982932 hasConceptScore W2182982932C119857082 @default.
• W2182982932 hasConceptScore W2182982932C136389625 @default.
• W2182982932 hasConceptScore W2182982932C153258448 @default.
• W2182982932 hasConceptScore W2182982932C154945302 @default.
• W2182982932 hasConceptScore W2182982932C155032097 @default.
• W2182982932 hasConceptScore W2182982932C159149176 @default.
• W2182982932 hasConceptScore W2182982932C175202392 @default.
• W2182982932 hasConceptScore W2182982932C2780598303 @default.
• W2182982932 hasConceptScore W2182982932C33923547 @default.
• W2182982932 hasConceptScore W2182982932C41008148 @default.
• W2182982932 hasConceptScore W2182982932C41445625 @default.
• W2182982932 hasConceptScore W2182982932C50644808 @default.
• W2182982932 hasConceptScore W2182982932C97541855 @default.
• W2182982932 hasLocation W21829829321 @default.
• W2182982932 hasOpenAccess W2182982932 @default.
• W2182982932 hasPrimaryLocation W21829829321 @default.
• W2182982932 hasRelatedWork W1552150563 @default.
• W2182982932 hasRelatedWork W1557075680 @default.
• W2182982932 hasRelatedWork W1565301672 @default.
• W2182982932 hasRelatedWork W2121839474 @default.
• W2182982932 hasRelatedWork W2163321856 @default.
• W2182982932 hasRelatedWork W2166241225 @default.
• W2182982932 hasRelatedWork W2316102783 @default.
• W2182982932 hasRelatedWork W2323585481 @default.
• W2182982932 hasRelatedWork W2412104792 @default.
• W2182982932 hasRelatedWork W2413005285 @default.
• W2182982932 hasRelatedWork W2483754710 @default.
• W2182982932 hasRelatedWork W2527007184 @default.
• W2182982932 hasRelatedWork W2726756416 @default.
• W2182982932 hasRelatedWork W2802636049 @default.
• W2182982932 hasRelatedWork W2890809735 @default.
• W2182982932 hasRelatedWork W2896230333 @default.
• W2182982932 hasRelatedWork W2912648029 @default.
• W2182982932 hasRelatedWork W2914317573 @default.
• W2182982932 hasRelatedWork W3201642682 @default.
• W2182982932 hasRelatedWork W70489736 @default.
• W2182982932 isParatext "false" @default.

• next