409 citations · 677 across the 54 of their papers we have counts for
7 papers · 1 filter
Improving Tail-Class Representation with Centroid Contrastive Learning
Anthony Meng Huat Tiong, Junnan Li, Guosheng Lin +3
In vision domain, large-scale natural datasets typically exhibit long-tailed distribution which has large class imbalance between head and tail classes. This distribution poses dif…
Noise-Resistant Deep Metric Learning with Probabilistic Instance Filtering
Chang Liu, Han Yu, Boyang Li +6
Noisy labels are commonly found in real-world data, which cause performance degradation of deep neural networks. Cleaning data manually is labour-intensive and time-consuming. Prev…
Initialization Matters: Regularizing Manifold-informed Initialization for Neural Recommendation Systems
Yinan Zhang, Boyang Li, Yong Liu +2
Proper initialization is crucial to the optimization and the generalization of neural networks. However, most existing neural recommendation systems initialize the user and item em…
Noise-resistant Deep Metric Learning with Ranking-based Instance Selection
Chang Liu, Han Yu, Boyang Li +6
The existence of noisy labels in real-world data negatively impacts the performance of deep learning models. Although much research effort has been devoted to improving robustness…
Latent-Optimized Adversarial Neural Transfer for Sarcasm Detection
Xu Guo, Boyang Li, Han Yu +1
The existence of multiple datasets for sarcasm detection prompts us to apply transfer learning to exploit their commonality. The adversarial neural transfer (ANT) framework utilize…
VisualGPT: Data-efficient Adaptation of Pretrained Language Models for Image Captioning
Jun Chen, Han Guo, Kai Yi +2
The ability to quickly learn from a small quantity oftraining data widens the range of machine learning applications. In this paper, we propose a data-efficient image captioning mo…
HYDRA: Hypergradient Data Relevance Analysis for Interpreting Deep Neural Networks
Yuanyuan Chen, Boyang Li, Han Yu +2
The behaviors of deep neural networks (DNNs) are notoriously resistant to human interpretations. In this paper, we propose Hypergradient Data Relevance Analysis, or HYDRA, which in…