AI 与机器学习2014arxiv.org编辑精选
DeepSpeech：端到端语音识别的大规模扩展
DeepSpeech: Scaling up end-to-end speech recognition
进阶 · 约 55 分钟读完前置知识：深度学习基础、循环神经网络、语音识别基础
导读
百度硅谷 AI 实验室提出的 DeepSpeech，是一个端到端的语音识别系统：用深度学习（含双向循环层）直接从语音映射到文字，借助大规模数据训练，在完整测试集上取得 16.0% 的错误率，声称在嘈杂环境下表现优于 Google 和 Apple 的商用系统。论文展示了端到端深度学习方案在语音识别上的可扩展性。
HN 关注点：百度团队（含吴恩达）宣称用深度学习在语音识别上超越 Google 和 Apple，引发读者对结果可信度和与 IBM 等已有工作对比的讨论。
后来怎么样
该论文验证了端到端深度学习语音识别在工业规模上的可行性，成为这一范式的标志性工作之一；Mozilla 后来据此思路开源了同名项目 DeepSpeech，端到端方案此后逐步取代传统混合系统成为语音识别主流。
HN 评论精选
这几条评论里：支持 1 质疑 1 补充 2
质疑指出 IBM Watson 团队此前在 SWB 数据集上的错误率（10.4%）已优于百度报告的 13.1%
客观地说，我们 IBM Watson 团队此前已发表了更好的数字：在 SWB 数据集上 WER 为 10.4%，而百度是 13.1%。我们没在 CH 部分上跑过模型，所以无法在完整测试集上直接比较。论文见链接。
pesenti查看英文原文
To put it in perspective, my team in IBM Watson has already published better numbers (10.4% WER vs 13.1% WER for Baidu) on the SWB dataset. We haven't run our model on the CH part so we can't compare on the full test set. Paper here: http://www.mirlab.org/conference_papers/International_Conference/ICASSP%202014/papers/p5609-soltau.pdf.
支持感叹团队组建仅数月就产出论文，速度惊人
这是百度硅谷 AI 实验室非常快的进展！吴恩达 5 月才加入百度，而这篇论文的（几乎？）所有共同作者都是在那之后加入他团队的。祝贺 Carl、Sanjeev、Andrew 等人。
brandonb查看英文原文
This is very fast progress from Baidu's Silicon Valley AI lab! Andrew Ng only joined Baidu in May, and (nearly?) all of the co-authors of this paper have joined him since then: http://www.technologyreview.com/news/527301/chinese-search-giant-baidu-hires-man-behind-the-google-brain/ Congrats to Carl, Sanjeev, Andrew, and the others.
补充提出分阶段/无监督预训练以降低训练成本和数据需求的改进思路
很棒的工作。我在想能否通过分开训练模型各部分来简化训练：DeepSpeech 有三层前馈层加一个双向循环层再加一层前馈层，如果先用无监督任务（如自编码或完形填空）训练前几层再固定权重训练其余部分，可能缩短训练时间、减少所需标注数据，也便于把前几层迁移到翻译等其他任务。另外最终结果是六个独立训练网络的集成，能否只完整训练一次网络，然后固定前三层再训练其他网络？
greeneggs查看英文原文
Very nice. I wonder if training can be simplified by training pieces of the model separately, instead of training all together. For example, the DeepSpeech model has three layers of feedforward neurons (where the inputs to the first layer are overlapping contexts of audio), followed by a bi-directional recurrent layer, followed by another feedforward layer. What would the results be if we trained the first layers (perhaps all three) on a different problem, such as autoencoding or fill-in-the-blank (as in word2vec), and then fixed those network weights to train the rest of the network? Breaking the network up like this would reduce training time and perhaps reduce the needed training data. Since the first layers could be trained without supervision, less labeled data would be needed to train the last two layers. It would also facilitate transferring models between problems; the output of the first few layers, like a word2vec, could be fed into arbitrary other machine learning problems, e.g., translation. If this does not work, then how about training the whole model together, but only once? The final results are reported for an ensemble of six independently trained networks. What if started by training one network, and then fixed the first three layers to train other networks? (Instead of fixing the first layers, you could also just give them a slower training rate, although it isn't clear whether that would save you much.)
补充补充媒体报道链接
贴一下 arXiv 链接，另外还有一些媒体报道：Gigaom 和 Forbes 都报道了百度宣称的深度学习语音识别突破。
cbcase查看英文原文
Thought it best to post the arXiv link, but there's some press coverage as well: - https://gigaom.com/2014/12/18/baidu-claims-deep-learning-breakthrough-with-deep-speech/ - http://www.forbes.com/sites/roberthof/2014/12/18/baidu-announces-breakthrough-in-speech-recognition-claiming-to-top-google-and-apple/
历次讨论
- 2014Hacker News 讨论帖79 分 · 22 评
- 2015Hacker News 讨论帖63 分 · 8 评
同领域还值得读
DeepSeek-R1：通过强化学习激励 LLM 的推理能力
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL
研究如何用强化学习激励大语言模型的推理能力，提出 DeepSeek-R1，并探讨 RL 训练对模型推理表现的影响。
AI 与机器学习2025 · 1,351 分 · 1,056 评 · 引用 30进阶 · 约 80 分钟编辑精选1-bit LLM 时代：面向低成本计算的三值参数
The Era of 1-bit LLMs: ternary parameters for cost-effective computing
提出用三值参数 {-1, 0, 1} 替代 LLM 中的浮点权重，即所谓“1-bit LLM”。
AI 与机器学习2024 · 1,040 分 · 447 评 · 引用 19进阶 · 约 45 分钟编辑精选DeepSeek-V3.2：推进开放大语言模型的前沿
DeepSeek-v3.2: Pushing the frontier of open large language models
这是 DeepSeek-V3.2 的技术报告，延续了 DeepSeek 开源权重模型的路线，目标是让开放模型继续逼近甚至追平前沿闭源模型。
AI 与机器学习2025 · 982 分 · 465 评进阶 · 约 90 分钟编辑精选通过自我对弈与通用强化学习掌握国际象棋和将棋（AlphaZero）
Mastering Chess and Shogi by Self-Play with General Reinforcement Learning
DeepMind 的 AlphaZero 论文：用通用强化学习算法（自我博弈 + MCTS + 神经网络），从零开始仅数小时训练便在国际象棋和将棋上击败了世界顶级引擎 Stock…
AI 与机器学习2017 · 539 分 · 270 评 · 引用 13进阶 · 约 90 分钟编辑精选用强化学习发现更快的矩阵乘法算法
Discovering faster matrix multiplication algorithms with reinforcement learning
论文提出 AlphaTensor，用强化学习搜索矩阵乘法算法。
AI 与机器学习2022 · 518 分 · 111 评专业 · 约 90 分钟编辑精选