TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Matrix Calculus (For Machine Learning and Beyond)

183 点作者 ibobev大约 2 个月前

9 条评论

sfpotter大约 2 个月前
If you want to get handy with matrix calculus, the real prerequisite is being comfortable with Taylor expansions and linear algebra.<p>In a graduate numerical optimization class I took over a decade ago, the professor spent 10 minutes on the first day deriving some matrix calculus identity by working out the expressions for partial derivatives using simple calculus rules and a lot of manual labor. Then, as the class was winding up, he joked and said &quot;just kidding, don&#x27;t do that... here&#x27;s how we can do this with a Taylor expansion&quot;, and proceeded to derive the same identity in what felt like 30 seconds.<p>Also, don&#x27;t forget the Jacobian and gradient aren&#x27;t the same thing!
评论 #43520321 未加载
评论 #43520067 未加载
i_am_proteus大约 2 个月前
Those looking for a shorter primer could consult <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;1802.01528" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;1802.01528</a>
评论 #43521039 未加载
westurner大约 2 个月前
&gt; <i>The class involved numerous example numerical computations using the Julia language, which you can install on your own computer following these instructions. The material for this class is also located on GitHub at <a href="https:&#x2F;&#x2F;github.com&#x2F;mitmath&#x2F;matrixcalc" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;mitmath&#x2F;matrixcalc</a> </i>
dandanua大约 2 个月前
The Matrix Cookbook [1] can be handy when learning this topic.<p><a href="https:&#x2F;&#x2F;www.math.uwaterloo.ca&#x2F;~hwolkowi&#x2F;matrixcookbook.pdf" rel="nofollow">https:&#x2F;&#x2F;www.math.uwaterloo.ca&#x2F;~hwolkowi&#x2F;matrixcookbook.pdf</a>
评论 #43524962 未加载
评论 #43522513 未加载
vismit2000大约 2 个月前
3b1b classic on this topic with beautiful visualizations: <a href="https:&#x2F;&#x2F;youtu.be&#x2F;O85OWBJ2ayo" rel="nofollow">https:&#x2F;&#x2F;youtu.be&#x2F;O85OWBJ2ayo</a>
windsignaling大约 2 个月前
Great course. I highly recommended anyone interested in this topic to check it out on the MIT website, taught by the same authors. They are great lecturers.
评论 #43520808 未加载
Koncopd大约 2 个月前
I have skimmed it, and it looks very good. It is actually not solely about matrix calculus, but shows a practical approach to differentiation in different vector spaces with many examples and intuitions.
revskill大约 2 个月前
What does calculus mean ?
评论 #43524613 未加载
评论 #43519371 未加载
FilosofumRex大约 2 个月前
wait what - another math textbook recommendation by academicians. ML and MLL are arts of tinkering not academic subjects.<p>Though Steven Johnson is the real deal and writes lots of code, Edelman is a shyster&#x2F;imposter who used to ride the coattails of G. Strang and now shills for Julia where he makes most of his money. You don&#x27;t need, and won&#x27;t understand ML&#x2F;LLM by reading textbooks.<p>1. If you want to have a little fun with ML&#x2F;LLM, fire up Google Collab and run one of tutorials on the web - Karpathy, Hugging Face or PyTorch examples.<p>2. If you don&#x27;t want to do, but just read for fun, Howard &amp; Parr&#x27;s essay as recommended by someone else here is much shorter and more succinct. <a href="https:&#x2F;&#x2F;explained.ai&#x2F;matrix-calculus&#x2F;" rel="nofollow">https:&#x2F;&#x2F;explained.ai&#x2F;matrix-calculus&#x2F;</a> this link renders better<p>3. If you insist on academic textbooks, Boyd &amp; Vandenberghe skips calculus and has more applications (engineering). Unfortunately, code examples are in Julia! <a href="https:&#x2F;&#x2F;web.stanford.edu&#x2F;~boyd&#x2F;vmls&#x2F;vmls.pdf" rel="nofollow">https:&#x2F;&#x2F;web.stanford.edu&#x2F;~boyd&#x2F;vmls&#x2F;vmls.pdf</a> <a href="https:&#x2F;&#x2F;web.stanford.edu&#x2F;~boyd&#x2F;vmls&#x2F;" rel="nofollow">https:&#x2F;&#x2F;web.stanford.edu&#x2F;~boyd&#x2F;vmls&#x2F;</a>. link to python version<p>4. If you Want to become a tensor &amp; differential programming ninja, learn Jax, XLA <a href="https:&#x2F;&#x2F;docs.jax.dev&#x2F;en&#x2F;latest&#x2F;quickstart.html" rel="nofollow">https:&#x2F;&#x2F;docs.jax.dev&#x2F;en&#x2F;latest&#x2F;quickstart.html</a> <a href="https:&#x2F;&#x2F;colab.research.google.com&#x2F;github&#x2F;exoplanet-dev&#x2F;jaxoplanet&#x2F;blob&#x2F;main&#x2F;docs&#x2F;tutorials&#x2F;introduction-to-jax.ipynb" rel="nofollow">https:&#x2F;&#x2F;colab.research.google.com&#x2F;github&#x2F;exoplanet-dev&#x2F;jaxop...</a>