2021/02/17 by Irwan Bello, Bello, Irwan · 1 voice · 48 citations
Computer Science · #Advanced Neural Network Applications #Artificial intelligence #Artificial neural network #Computer science #Convolutional neural network #Domain Adaptation and Few-Shot Learning #Image (mathematics) #Lambda #Machine learning #Multimodal Machine Learning Applications #Object detection #Pattern recognition (psychology) #Pixel #Range (aeronautics) #Segmentation #cs.CV #cs.LG
paper · pdf · doi:10.48550/arxiv.2102.08602
published in arXiv (Cornell University) (Cornell University) · Accepted for publication at the International Conference in Learning Representations 2021 (Spotlight)
arxiv created 2021/02/17 · openalex publication_date 2021/02/17 · arxiv published 2021/02/17 · arxiv updated 2021/02/18 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
We present lambda layers -- an alternative framework to self-attention -- for capturing long-range interactions between an input and structured contextual information (e.g. a pixel surrounded by other pixels). Lambda layers capture such interactions by transforming available contexts into linear functions, termed lambdas, and applying these linear functions to each input separately. Similar to linear attention, lambda layers bypass expensive attention maps, but in contrast, they model both content and position-based interactions which enables their application to large structured inputs such as images. The resulting neural network architectures, LambdaNetworks, significantly outperform their convolutional and attentional counterparts on ImageNet classification, COCO object detection and COCO instance segmentation, while being more computationally efficient. Additionally, we design LambdaResNets, a family of hybrid architectures across different scales, that considerably improves the speed-accuracy tradeoff of image classification models. LambdaResNets reach excellent accuracies on ImageNet while being 3.2 - 4.4x faster than the popular EfficientNets on modern machine learning accelerators. When training with an additional 130M pseudo-labeled images, LambdaResNets achieve up to a 9.5x speed-up over the corresponding EfficientNet checkpoints.