How to describe the spatial near-far relations among concepts?

Zheng, Keyin; Qian, Yuhua; Cheng, Hao

doi:10.1016/j.ijar.2023.02.005

Cited by 3 publications

(1 citation statement)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The way the human brain processes, recognizes, and understands images does not focus on individual pixels but considers local and global feature regions within images, the relations among different concepts within images, and the concepts of interest associated with tasks. Second, Relevant studies have indicated that considering the potential underlying correlation structures among data features or variables holds significant value Cheng et al (2020;, leading to effective performance improvements in tasks such as category representation and data fusion Liang et al (2021); Zheng et al (2023). Third, symmetric transformations in representation learning have gained increased attention in the field of neuroscience Higgins et al (2022).…”

Section: Introductionmentioning

confidence: 99%

Centrosymmetric Constrained Convolutional Neural Networks

Zheng,

Qian,

Yuan

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Complex signals can be viewed as compositions of numerous sine waves with different frequencies and amplitudes. As the fundamental unit of perceiving image features, traditional convolutional neural networks (CNNs) typically employ convolutional kernels without direct constraints. However, human perception of the world is inherently structured, relevant studies have indicated that considering the potential underlying correlation structures among data features or variables holds significant value, with particular emphasis on symmetric transformations in representation learning gaining attention in the field of neuroscience. Inspired by the process of human visual concept perception and classic image feature detection operators, when the convolution kernels have a similar scale to the objects, they produce larger responses. This paper introduces constraints on model parameters using centrosymmetric convolutional kernels. Under the influence of nonlinear combinations and local connections in neural networks, these kernels can diversely represent or perceive local image features. Experimental comparisons are conducted, and it is observed that the approach proposed in this paper not only enhances the convergence and accuracy of model, but also reduces the model solution space, resulting in an interpretable perception of local concepts. Furthermore, comparative experiments with classic deep CNNs have demonstrated comparable performance.

show abstract