- 非自回归生成研究最新综述，近200篇文献揭示挑战和未来方向 (opens in new tab)
- ICLR 2022 | 微软亚洲研究院深度学习领域最新研究成果一览 (opens in new tab)
- 语音识别的快速纠错模型FastCorrect系列来了！ (opens in new tab)
- 如何亿点点降低语音识别跨领域、跨语种迁移难度？ (opens in new tab)
- 气候变化、流行病、发展鸿沟…… 应对这些挑战我们还要做些什么？ (opens in new tab)
- 你真的了解计算生物学和AI for Science吗？ (opens in new tab)
- 微软亚洲研究院副院长刘铁岩博士获选2021 ACM Fellow! (opens in new tab)
- Molecular Dynamics Simulation Accelerates Research of the Pathogenic Mechanism of COVID-19 (opens in new tab)
- 公开催化剂挑战赛冠军模型、通用AI分子模拟库Graphormer开源! (opens in new tab)
- 秦涛：以独立、深度的视角看世界，做有意义、创新的研究 (opens in new tab)
- 微软亚洲研究院成立理论中心，以理论研究打破AI发展瓶颈 (opens in new tab)
- 微软亚洲研究院推出时空预测开源工具FOST，应对各行业共性预测需求 (opens in new tab)
- R-Drop: A simple and effective regular method to correct the defects of Dropout (opens in new tab)
- NeurIPS 2021 一文洞悉因果机器学习前沿进展 (opens in new tab)
- NeurIPS 2021 | CyGen：基于概率论理论的生成式建模新模式 (opens in new tab)
- NTD的深度研究，为厘清新冠病毒机理提供新方向！ (opens in new tab)
- EMNLP 2021 | 微软亚洲研究院NLP领域最新研究一览 (opens in new tab)
- 精心设计的 GNN 只是“计数器”？ (opens in new tab)
- 微软翻译突破百种语言和方言大关 (opens in new tab)
- AI由“点”到“面”，逐个解锁传统行业 (opens in new tab)
- AI Helps Connect the Dots: Unlocking Traditional Industries One by One (opens in new tab)
- 可持续发展的人工智能 (opens in new tab)
- AI打通关键环节，加快物流行业数字化转型 (opens in new tab)
- 应对个性化定制语音合成挑战，微软推出AdaSpeech系列研究 (opens in new tab)
- Tech Minutes: Mastering Mahjong with AI (opens in new tab)
- Humana leverages Microsoft Cloud for Healthcare to develop advanced predictive models (opens in new tab)
- 讲堂 | 刘铁岩：科研到底怎么做？什么是高质量研究？ (opens in new tab)
- 如何利用深度学习优化大气污染物排放量估算？ (opens in new tab)
- IJCAI 2021 | 一文了解微软亚洲研究院机器学习方向前沿进展 (opens in new tab)
- KDD 2021 | Transformer、知识图谱等热点话题，微软亚洲研究院论文精选，速看！ (opens in new tab)
- KDD 2021 | 用NAS实现任务无关且可动态调整尺寸的BERT压缩 (opens in new tab)
- 系统调研450篇文献，微软亚洲研究院推出超详尽语音合成综述 (opens in new tab)
- ICML 2021 | 微软亚洲研究院精选论文一览 (opens in new tab)
- R-Drop：填补Dropout缺陷，简单又有效的正则方法 (opens in new tab)
- 谭旭：AI音乐，技术与艺术的碰撞 (opens in new tab)
- 刘铁岩：跨界共创AI的产业价值和科学价值 (opens in new tab)
- KDD Cup 2021 | 微软亚洲研究院Graphormer模型荣登OGB-LSC图预测赛道榜首 (opens in new tab)
- Transformer stands out as the best graph learner: Researchers from Microsoft Research Asia wins the KDD Cup’s 2021 Graph Prediction Track (opens in new tab)
- 一键部署分布式训练，微软“群策 MARO”上新集群管理助手 (opens in new tab)
- ICLR 2021 | 微软亚洲研究院精选论文一览 (opens in new tab)
- 机器学习隐私研究新进展：数据增强风险被低估，新算法“降服”维数依赖 (opens in new tab)
- AAAI 2021 | 微软亚洲研究院论文大礼包请查收！ (opens in new tab)
- AAAI 2021 | 微软与上交大最新研究，强化学习助力AI+金融 (opens in new tab)
- 快速上手微软 “群策 MARO” 平台，打造简易的共享单车场景 (opens in new tab)
- AAAI 2021 | 不依赖文本也能做翻译？UWSpeech语音翻译系统了解一下 (opens in new tab)
- NeurIPS 2020 | 微软亚洲研究院论文摘录之强化学习&GAN篇 (opens in new tab)
- 微矿Qlib：业内首个AI量化投资开源平台 (opens in new tab)
- NeurIPS 2020 | 微软亚洲研究院论文摘录之目标检测篇 (opens in new tab)
- 开源平台MARO：资源调度优化的任意门 (opens in new tab)
- 还在捞五条人？不如用AI自己组乐队 (opens in new tab)
- KDD 2020 | LRSpeech：极低资源下的语音合成与识别 (opens in new tab)
- 如何在机器学习的框架里实现隐私保护？ (opens in new tab)
- IJCAI 2020 | 微软亚洲研究院精选论文摘录 (opens in new tab)
- ICML 2020 | 摆脱warm-up！巧置LayerNorm使Transformer加速收敛 (opens in new tab)
- 微软与清华大学联合提出DeepRSM模型，以AI助力空气污染治理 (opens in new tab)
- Microsoft and Tsinghua University jointly propose the DeepRSM model to help control air pollution with AI (opens in new tab)
- FastSpeech语音合成系统技术升级，微软联合浙大提出FastSpeech2 (opens in new tab)
- FastSpeech 2: Fast and High-Quality End-to-End Text to Speech (opens in new tab)
- 集“百家”之长，成一家之言！微软提出全新预训练模型MPNet (opens in new tab)
- MPNet combines strengths of masked and permuted language modeling for language understanding (opens in new tab)
- 微软超级麻将AI Suphx论文发布，研发团队深度揭秘技术细节 (opens in new tab)
- 低耗时、高精度，微软提出基于半监督学习的神经网络结构搜索算法SemiNAS (opens in new tab)
- 从病毒到免疫， “科学地”揭开新冠病毒的神秘面纱 (opens in new tab)
- 2019盘点：机器学习更亲民，AI系统更精巧 (opens in new tab)
- NeurIPS 2020: Moving toward real-world reinforcement learning via batch RL, strategic exploration, and representation learning (opens in new tab)
- NeurIPS 2019 | 全参数化分布，提升强化学习中的收益分布拟合能力 (opens in new tab)
- NeurIPS 2019丨推敲网络+soft原型序列，带来轻便又精准的机器翻译 (opens in new tab)
- 微软超级麻将AI Suphx，破解非完美信息游戏 (opens in new tab)
- 游戏 AI 挑战进阶，即时策略游戏和非完美信息游戏成为热点 (opens in new tab)
- 哪类游戏AI难度更高？用数学方法来分析一下 (opens in new tab)
- 游戏 AI 的缘起与进化 (opens in new tab)
- 速度提升270倍！微软和浙大联合推出全新语音合成系统FastSpeech (opens in new tab)
- FastSpeech: New text-to-speech model improves on speed, accuracy, and controllability (opens in new tab)
- WMT 2019国际机器翻译大赛：微软亚洲研究院以8项第一成为冠军 (opens in new tab)
- Microsoft Research Asia (MSRA) leads in 2019 WMT international machine translation competition (opens in new tab)
- 刘铁岩谈机器学习：随波逐流的太多，我们需要反思 (opens in new tab)
- Finding the best learning targets automatically: Fully Parameterized Quantile Function for distributional RL (opens in new tab)
- More than a game: Mastering Mahjong with AI and machine learning (opens in new tab)
- 《分布式机器学习：算法、理论与实践》——理论、方法与实践的全面汇总 (opens in new tab)
- 成为机器学习大神，你不能不懂数学 (opens in new tab)
- 机器学习：未来十年研究热点 (opens in new tab)
- Machine Learning: Research hotspots in the next ten years (opens in new tab)
- ICML 2018 | 模型层面的对偶学习 (opens in new tab)
- 分布式深度学习新进展：让“分布式”和“深度学习”真正深度融合 (opens in new tab)
- ICML 2018 | 训练可解释、可压缩、高准确率的LSTM (opens in new tab)
- 邵斌：用符号学习生成精确、可解释模型 (opens in new tab)
- ICLR 2018论文 | Learning to Teach：让AI和机器学习算法教学相长 (opens in new tab)
- 微软人工智能又一里程碑：微软中-英机器翻译水平可“与人类媲美” (opens in new tab)
- 干货 | NIPS 2017：用于序列生成的推敲网络 (opens in new tab)
- NIPS 2017线上分享：利用价值网络改进神经机器翻译 (opens in new tab)
- 干货 | 手把手带你入门微软Graph Engine (opens in new tab)
- 到底什么是生成式对抗网络GAN？ (opens in new tab)
- 刘铁岩：人工智能的挑战与机遇 (opens in new tab)
- 秦涛：深度学习的五个挑战和其解决方案 (opens in new tab)
- 微软亚洲研究院开源图数据查询语言LIKQ (opens in new tab)
- ICML 2019 | 微软提出极低资源下语音合成与识别新方法，小语种也不怕没数据了！ (opens in new tab)
- ICML 2019 | 序列到序列自然语言生成任务超越BERT、GPT！微软提出通用预训练模型MASS (opens in new tab)
- Introducing MASS – A pre-training method that outperforms BERT and GPT in sequence to sequence language generation tasks (opens in new tab)
- 分布式图处理引擎Graph Engine 1.0 预览版正式发布 (opens in new tab)
- 微软亚洲研究院开源分布式机器学习工具包 (opens in new tab)
- 开源 | LightGBM：三天内收获GitHub 1000 星 (opens in new tab)
- LightRNN：深度学习之以小见大 (opens in new tab)
- 刘铁岩：对偶学习推动人工智能的新浪潮 (opens in new tab)
- 对偶学习：一种新的机器学习范式 (opens in new tab)
- 微软首席研究员刘铁岩：深度学习的推力与阻碍 (opens in new tab)
- 刘铁岩：博弈机器学习是什么？ (opens in new tab)