Linformer: Self-Attention with Linear Complexity

Review of paper by Sinong Wang, Belinda Z. Li, Madian Khabsa et al, Facebook AI Research, 2020

This paper suggests an approximate way of calculating self-attention in Transformer architectures that has linear space and time complexity in terms of the sequence length, with the resulting performance on benchmark datasets similar to that of the RoBERTa model based on the original Transformers with much less efficient quadratic attention complexity.


End-to-End Object Detection with Transformers

Review of paper by Nicolas Carion, Francisco Massa, Gabriel Synnaeve et al, Facebook AI Research, 2020

This paper describes a completely automated end-to-end object detection system combining convolutional networks and Transformers. The new model shows competitive performance on par with Faster R-CNN and can be generalized to other tasks such as panoptic segmentation.


Synthesizer: Rethinking Self-Attention in Transformer Models

Review of paper by Yi Tay, Dara Bahri, Donald Metzler et al, Google Research, 2020

Contrary to the common consensus that self-attention is largely responsible for the superior performance of Transformer models on various NLP tasks, this paper suggests that substituting outputs of self-attention layers with random or simply synthesized data is sufficient to achieve similar results with better efficiency.


Supervised Contrastive Learning

Review of paper by Prannay Khosla, Piotr Teterwak, Chen Wang et al, Google Research, 2020

The authors used contrastive loss, which has recently been shown to be very effective at learning deep neural network representations in the self-supervised setting, for supervised learning, and achieved better results than those obtained with the cross-entropy loss for ResNet-50 and ResNet-200.


ResNeSt: Split-Attention Networks

Review of paper by Hang Zhang1, Chongruo Wu2, Zhongyue Zhang1 et al, 1Amazon and 2UC Davis, 2020

The authors suggest a new ResNet-like network architecture that incorporates attention across groups of feature maps. Compared to previous attention models such as SENet and SKNet, the new attention block applies the squeeze-and-attention operation separately to each of the selected groups, which is done in a computationally efficient way and implemented in a simple modular structure.


No more pages to load