A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Yuri Kuratov, A. Bulatov, Petr Anokhin, et al.
Recurrent memory augmentation enables GPT-2 to process sequences up to 11 million elements, far exceeding standard methods limited to 10,000 elements.
Amina Adadi
This survey systematically categorizes data-efficient ML methods into four strategies: non-supervised learning, data augmentation, transfer learning, and algorithm modification.
Yan Yan, Yuxing Mao, Bo Li
Proposes SECOND, a sparse convolutional network with angle loss regression and data augmentation for fast and accurate 3D object detection from LiDAR.
Ke Wang, Jiahui Zhu, Minjie Ren, et al.
This survey comprehensively reviews data generation techniques for LLMs, covering data augmentation and synthesis across the entire LLM lifecycle.
Unknown
SimCLR simplifies contrastive learning by optimizing data augmentation, adding learnable nonlinear transformations, and scaling batch size and training steps.
Liang Zeng, Liangjun Zhong, Liang Zhao, et al.
A 7B LLM fine-tuned on a new 2.5M instance math QA dataset generated via a two-stage pipeline with diverse seeds and hard problem augmentation.
Dhruv Grewal, Cinthia B. Satornino, Thomas H. Davenport, et al.
This paper presents a four-quadrant framework for marketers to evaluate trade-offs in generative AI inputs and human augmentation needed for outputs.
Alexander Buslaev, Vladimir I. Iglovikov, Eugene Khvedchenya, et al.
Albumentations is a fast, flexible open-source image augmentation library offering many transform operations and serving as an easy-to-use wrapper around other libraries.
Sebastian Raisch, Sebastian Krakowski
This paper argues that automation and augmentation in AI are not separable but paradoxical, requiring a balanced perspective to avoid negative outcomes.
Connor Shorten, Taghi M. Khoshgoftaar
A comprehensive survey of image data augmentation techniques for deep learning, covering geometric, color, kernel, mixing, erasing, adversarial, GAN, style transfer, and meta-learning methods.
Hao Liu, Chenghuan Huang, Ye Huang, et al.
FVAttn is a training-free sparse-attention system that uses runtime load balancing and slack-aware augmentation to improve distributed execution efficiency for video generation.
Connor Shorten, Taghi M. Khoshgoftaar, Borko Furht
This survey reviews text data augmentation for deep learning, covering motifs, frameworks, generalization, and practical tools.