1 code implementation • 16 Oct 2023 • Zhicheng Cai, Xiaohan Ding, Qiu Shen, Xun Cao
We propose Re-parameterized Refocusing Convolution (RefConv) as a replacement for regular convolutional layers, which is a plug-and-play module to improve the performance without any inference costs.
no code implementations • 2 Jul 2023 • Xinyue Wang, Zhicheng Cai, Chenglei Peng
However, existing vision MLP architectures always depend on convolution for patch embedding.
no code implementations • 18 Jun 2023 • Zhicheng Cai, Chenglei Peng, Qiu Shen
In this way, LENI can enhance the model representational capacity significantly while maintaining the original advantages of ReLU.
no code implementations • 10 Jun 2023 • Zhicheng Cai, Qiu Shen
To address these issues, we factorize the four vital components of light-weight CNNs from coarse to fine and redesign them: i) we design a light-weight overall architecture termed LightNet, which obtains better performance by simply implementing the basic blocks of other light-weight CNNs; ii) we abstract a Meta Light Block, which consists of spatial operator and channel operator and uniformly describes current basic blocks; iii) we raise RepSO which constructs multiple spatial operator branches to enhance the representational ability; iv) we raise the concept of receptive range, guided by which we raise RefCO to sparsely factorize the channel operator.
no code implementations • 23 May 2023 • Zhicheng Cai
Traditionally, different types of feature operators (e. g., convolution, self-attention and involution) utilize different approaches to extract and aggregate the features.
no code implementations • 15 Jul 2021 • Zhicheng Cai
Compared to the baseline models with traditional gradient descent algorithm, models with SA-GD algorithm possess better generalization ability without sacrificing the efficiency and stability of model convergence.
no code implementations • 26 Jun 2021 • Zhicheng Cai
However, it can be difficulty to settle the optimal network depth and make the middle layers learn distinguished features.
no code implementations • 25 Jun 2021 • Zhicheng Cai, Chenglei Peng, Sidan Du
As Jitter point acting as a random factor, we actually add some randomness to the loss function, which is consistent with the fact that there exists innumerable random behaviors in the learning process of the machine learning model and is supposed to make the model more robust.
no code implementations • 13 Jun 2021 • Zhicheng Cai, Kaizhu Huang, Chenglei Peng
This paper proposes a novel nonlinear activation mechanism typically for convolutional neural network (CNN), named as reborn mechanism.