A general framework for hypercomplex-valued extreme learning machines

Vieira, Guilherme; Valle, Marcos Eduardo

doi:10.1016/j.jcmds.2022.100032

Cited by 12 publications

(2 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, coquaternion-and tessarine-based HvCNNs outperformed the HvCNN based on quaternions, the four-dimensional algebra most widely used in applications. At this point, we would like to recall that superior performance of neural networks based on non-usual hypercomplex algebras has been previously reported in the literature [26], [29], despite quaternion-based neural network yielding better performance in applications like controlling a robot manipulator [12]. Finally, we would like to recall that Genovese et al obtained average accuracy rates of 97.92% using a ResNet18 combined with histopathological transfer learning [7].…”

Section: A Computational Experimentsmentioning

confidence: 63%

See 1 more Smart Citation

Acute Lymphoblastic Leukemia Detection Using Hypercomplex-Valued Convolutional Neural Networks

Vieira¹,

Valle²

2022

Preprint

Self Cite

View full text Add to dashboard Cite

This paper features convolutional neural networks defined on hypercomplex algebras applied to classify lymphocytes in blood smear digital microscopic images. Such classification is helpful for the diagnosis of acute lymphoblast leukemia (ALL), a type of blood cancer. We perform the classification task using eight hypercomplex-valued convolutional neural networks (HvCNNs) along with real-valued convolutional networks. Our results show that HvCNNs perform better than the real-valued model, showcasing higher accuracy with a much smaller number of parameters. Moreover, we found that HvCNNs based on Clifford algebras processing HSV-encoded images attained the highest observed accuracies. Precisely, our HvCNN yielded an average accuracy rate of 96.6% using the ALL-IDB2 dataset with a 50% train-test split, a value extremely close to the stateof-the-art models but using a much simpler architecture with significantly fewer parameters.

show abstract

Section: A Computational Experimentsmentioning

confidence: 63%

“…The four non-commutative hypercomplex algebras include quaternions and coquaternions. Also, they are isomorphic to Clifford as well as Cayley-Dickson algebras [26], [29].…”

Section: Introductionmentioning

confidence: 99%