Результаты исследований: Научные публикации в периодических изданиях › статья › Рецензирование
Bio-inspired machine learning: programmed death and replication. / Грабовский, Андрей Владимирович; Vanchurin, Vitaly.
в: Neural Computing and Applications, Том 35, № 27, 09.2023, стр. 20273-20298.Результаты исследований: Научные публикации в периодических изданиях › статья › Рецензирование
}
TY - JOUR
T1 - Bio-inspired machine learning: programmed death and replication
AU - Грабовский, Андрей Владимирович
AU - Vanchurin, Vitaly
N1 - V.V. was supported in part by the Foundational Questions Institute (FQXi) and the Oak Ridge Institute for Science and Education (ORISE).
PY - 2023/9
Y1 - 2023/9
N2 - We analyze algorithmic and computational aspects of biological phenomena, such as replication and programmed death, in the context of machine learning. We use two different measures of neuron efficiency to develop machine learning algorithms for adding neurons to the system (i.e., replication algorithm) and removing neurons from the system (i.e., programmed death algorithm). We argue that the programmed death algorithm can be used for compression of neural networks and the replication algorithm can be used for improving performance of the already trained neural networks. We also show that a combined algorithm of programmed death and replication can improve the learning efficiency of arbitrary machine learning systems. The computational advantages of the bio-inspired algorithms are demonstrated by training feedforward neural networks on the MNIST dataset of handwritten images.
AB - We analyze algorithmic and computational aspects of biological phenomena, such as replication and programmed death, in the context of machine learning. We use two different measures of neuron efficiency to develop machine learning algorithms for adding neurons to the system (i.e., replication algorithm) and removing neurons from the system (i.e., programmed death algorithm). We argue that the programmed death algorithm can be used for compression of neural networks and the replication algorithm can be used for improving performance of the already trained neural networks. We also show that a combined algorithm of programmed death and replication can improve the learning efficiency of arbitrary machine learning systems. The computational advantages of the bio-inspired algorithms are demonstrated by training feedforward neural networks on the MNIST dataset of handwritten images.
KW - Bio-inspired algorithms
KW - Classification
KW - Constructive algorithms
KW - Machine learning
KW - Neural networks
KW - Neuron correlations
KW - Pruning algorithms
UR - https://www.scopus.com/record/display.uri?eid=2-s2.0-85165286519&origin=inward&txGid=33fc24869a8dbc1ac624f533115995fe
UR - https://www.mendeley.com/catalogue/2f2fb520-2830-3781-a865-4f82b90b6f8b/
U2 - 10.1007/s00521-023-08806-4
DO - 10.1007/s00521-023-08806-4
M3 - Article
VL - 35
SP - 20273
EP - 20298
JO - Neural Computing and Applications
JF - Neural Computing and Applications
SN - 1433-3058
IS - 27
ER -
ID: 52756398