Blog
Here, some simple introductions to our research papers and AI popularization content are presented. This section is intended for science communication purposes only and has not undergone rigorous scientific validation. If there are any imperfections, please let us know.
2023
In this work, we propose a novel permutable hybrid network for Vol-MedSeg, named PHNet, which capitalizes on the strengths of both convolution neural networks (CNNs) and MLP. PHNet addresses the intrinsic isotropy problem of 3D volumetric data by employing a combination of 2D and 3D CNNs to extract local features.
大语言模型 (LLM) 压缩一直备受关注，后训练量化（Post-training Quantization) 是其中一种常用算法，但是现有 PTQ 方法大多数都是 integer 量化，且当比特数低于 8 时，量化后模型的准确率会下降非常多。想较于 Integer (INT) 量化，Floating Point (FP) 量化能更好的表示长尾分布，因而越来越多的硬件平台开始支持 FP 量化。这篇发表在 EMNLP 2023上的文章给出了大模型 FP 量化的解决方案。
In this paper, we propose an auxiliary and integrated network architecture, named Convolutional-Auxiliary Efficient Graph Reasoning Transformer (CAE-GReaT), which joints strengths of both Convolutional Neural Networks (CNNs) and Vision Transformer (ViT) into a uniform framework. The paper is published in the International Journal of Computer Vision (IJCV).
2022
本文收集了一系列医学图像分割的技巧，适用于不同的模型实现阶段。分别是预训练模型、数据预处理、数据增强、模型实现、模型推理和结果后处理)，并通过大量的实验结果探讨了这些技巧在一致性的基准模型上的有效性。