no code implementations • 3 Apr 2024 • Weichao Lan, Yiu-ming Cheung, Qing Xu, Buhua Liu, Zhikai Hu, Mengke Li, Zhenghua Chen
In addition to the supervision of ground truth, the vanilla KD method regards the predictions of the teacher as soft labels to supervise the training of the student model.
1 code implementation • 12 Jun 2023 • Mengke Li, Zhikai Hu, Yang Lu, Weichao Lan, Yiu-ming Cheung, Hui Huang
To rectify this issue, we propose to augment tail classes by grafting the diverse semantic information from head classes, referred to as head-to-tail fusion (H2T).
1 code implementation • 18 May 2023 • Mengke Li, Yiu-ming Cheung, Yang Lu, Zhikai Hu, Weichao Lan, Hui Huang
Based on these perturbed features, two novel logit adjustment methods are proposed to improve model performance at a modest computational overhead.
no code implementations • 3 May 2022 • Weichao Lan, Yiu-ming Cheung, Juyong Jiang
To this end, this paper presents a new method termed TissueNet, which directly constructs compact neural networks with fewer weight parameters by independently stacking designed basic units, without requiring additional judgement criteria anymore.
no code implementations • 6 Oct 2020 • Weichao Lan, Liang Lan
One popular way to reduce the memory cost of deep CNN model is to train binary CNN where the weights in convolution filters are either 1 or -1 and therefore each weight can be efficiently stored using a single bit.