Saliency-guided training combined with PACT quantization keeps MNIST and CIFAR-10 accuracy near parity with a quantized baseline, while the claimed efficiency and interpretability gains are not directly measured.
Conservative algebras of $2$-dimensional algebras, V
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
The notion of conservative algebras appeared in a paper by Kantor in 1972. Later, he defined the conservative algebra $W(n)$ of all algebras (i.e. bilinear maps) on the $n$-dimensional vector space. If $n>1$, then the algebra $W(n)$ does not belong to any well-known class of algebras (such as associative, Lie, Jordan, or Leibniz algebras). It looks like $W(n)$ in the theory of conservative algebras plays a similar role to the role of $\mathfrak{gl}_n$ in the theory of Lie algebras. Namely, an arbitrary conservative algebra can be obtained from a universal algebra $W(n)$ for some $n \in \mathbb{N}.$ The present paper is a part of a series of papers, which dedicated to the study of the algebra $W(2)$ and its principal subalgebras.
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Quantized and Interpretable Learning Scheme for Deep Neural Networks in Classification Task
Saliency-guided training combined with PACT quantization keeps MNIST and CIFAR-10 accuracy near parity with a quantized baseline, while the claimed efficiency and interpretability gains are not directly measured.