V
主页
【2023春季全球午餐会第13期】Incorporating Diffusion Models into Conditional Text Generation
发布人
【2023春季全球午餐会第13期】Incorporating Diffusion Models into Conditional Text Generation Shansan (Sansa) Gong is an NLP researcher at Shark-NLP Shanghai AI Lab. Her current research interests include controllable text generation and large pretrain language models. Diffusion models have recently gained traction as a powerful approach to generative modeling. While these models have shown great success in domains such as vision and audio, applying them to natural language generation has been relatively unexplored, primarily due to the discrete nature of text data, especially in the context of conditional generation. In this talk, we will present DiffuSeq (https://github.com/Shark-NLP/DiffuSeq), a novel diffusion model specifically designed for sequence-to-sequence (Seq2Seq) text generation tasks. Moreover, we will also present some recent follow-up works, highlighting how diffusion models can be extended and adapted to various natural language generation tasks. Finally, we will provide a theoretical analysis that connects diffusion models with autoregressive and non-autoregressive models. This analysis will shed light on the similarities and differences between these models and provide insights into their strengths and limitations.
打开封面
下载高清视频
观看高清视频
视频下载器
【2021年春季第13场】上海交通大学计算机科学全球午餐讲座系列
计算机系学术讲座:Attacking Machine Learning Models
【2023春季全球午餐会第6期】图论与组合问题的机器学习求解
【2024春季全球午餐会第2期】Reinventing Design Creation
【2023秋季全球午餐会第4期】量子机器学习初探
【2024春季全球午餐会第3期】挥手“窥气”:气体感知普适化之初探
【2023春季全球午餐会第10期】Probabilistic Sports Analytics
【2023春季全球午餐会第4期】Adaptive and Effective Fuzzing
【2024春季全球午餐会第1期】排列型组合优化问题的机器学习求解和AI4Sci应用
【2024春季全球午餐会第4期】The continuous integration and deployment
【2023春季全球午餐会第1期】端到端自动驾驶算法设计思考
【2023秋季全球午餐会第13期】InfoBodied AI: Learning Mutual Information
【2023秋季全球午餐会第12期】从学术研究到工业落地
【2023春季全球午餐会第9期】通往AI安全之路
从迁移学习到图像合成_2021年春季第14场_上海交通大学计算机科学全球午餐讲座系列
【2023秋季全球午餐会第10期】医学影像可解释人工智能分析
【2024春季全球午餐会第13期】Federated Learning and Its Extension
【2023春季全球午餐会第8期】决策大模型:背景与前沿
【2023春季全球午餐会第14期】Stream Ciphers: Past, Present and Future
【2023春季全球午餐会第7期】基于光信号的移动端感知
【2023春季全球午餐会第16期】Security and Privacy of Bluetooth Low Energy
【2024春季全球午餐会第9期】Extracting Diagnostic Information from Atypical Speech
【2023春季全球午餐会第5期】推荐系统的公平性问题
【2023秋季全球午餐会第7期】Unveiling ‘Vision’
【2023春季全球午餐会第11期】Facilitating Software Security
【2023春季全球午餐会第12期】超大规模图数据:挑战和机遇
Talk to Data: Conversational Natural Language_【2021年春季第7场】上海交通大学计算机科学全球午餐讲座系列
【2024春季全球午餐会第11期】OceanBase Paetica - A Hybrid Database
【2024春季全球午餐会第17期】Safety Assessment of Large Generative Model
全球午餐会|回首2023,期待更精彩的2024!
【2024秋季全球午餐会第5期】面向AI时代的芯片物理设计创新方法
【2024春季全球午餐会第15期】Learning quantum properties
【2023春季全球午餐会第3期】阿里云的云计算数据中心软硬件问题与基础软件前沿技术
【YOLOv11】一小时速通版!知名博士逐一解读配置文件以及代码复现,环境安装+推理+自定义数据集搭建与训练,入门到精通!
我在B站上大学!【完整版-麻省理工-微积分重点】全18讲!学数学不看的微积分课程,看完顺滑一整年。_人工智能数学基础/机器学习/微积分/麻省理工/高等数学
【全集195集】深度学习必看圣经!李沐大神《动手学深度学习》最新版全套视频教程分享,小白也能信手拈来,看完直接跑通!(深度学习/神经网络/pytorch)】
深度双向智能系统介绍:理论、方法和应用_2021年春季第15场_上海交通大学计算机科学全球午餐讲座系列
【2023秋季全球午餐会第8期】LLM大模型的挑战及国产GPU的最新进展
【2023秋季全球午餐会第3期】理解和度量大模型的安全问题
杰出系友面对面|戴文渊讲座:求索之路