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。

The Case Against Geometric Algebra

25 点作者 wiktor-k大约 1 年前

2 条评论

Nevermark大约 1 年前
As a software developer, the point about distinguishing multi-vectors from operations they may be isomorphic to makes complete sense.<p>I.e. just because two data structures have the same values, and are related, doesn&#x27;t make them the same thing.<p>A great deal of clarity is had by not treating all two-field structures of real&#x2F;float x and y, as if they were all points and offsets and vectors, etc. Each of those things is a different type, with different expectations on how it will be legally used, and combined with the others.<p>The fact that two numbers over two dimensions allows for isomorphic relations between them, does not make them the same thing.<p>--<p>This is very similar to how pointers&#x2F;addresses (1-dimensional objects) are represented by an integer, and so are offsets&#x2F;sizes. But they are different. Adding two addresses makes no sense. De-referencing an offset makes no sense.<p>There is an isomorphism, between offsets from the zero address and the same numerical addresses, but the distinction between whether you have an offset or address is important if you want your type system to guarantee proper&#x2F;safe behavior.<p>Conceptually, an address consists of a RAM origin, and an offset. Where as an offset is just an offset.<p>You can add and subtract offsets. You can add and subtract offsets to addresses (i.e. to their offsets). And you can subtract addresses (the RAM origins cancel, leaving an offset).<p>But you can&#x27;t add two addresses, because adding two origins makes no sense.<p>(Generalizing from adding two values, to multiple additions = multiplication of by indices: multiplying offsets by indices makes sense. Multiplying addresses by offsets or indices does not. Adding and subtracting indices makes sense. Multiplying indices makes sense, but not as an offset (although there is an isomorphism from the number of elements n of an i x j array of objects, it objects have a size of 1.))
jvanderbot大约 1 年前
The basic complaint is that GA adds nothing to exterior algebra, which is well understood, even if under-utilized. That the concepts of EA are under-used is not up for debate: Everyone agrees. In fact, by reframing everything from the first introduction to linear algrebra and up in terms of GA, you <i>lose</i> intuition, compatibility, and gain nothing.<p><a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Exterior_algebra" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Exterior_algebra</a><p>Having read the GA textbook and vector calculus textbook, I can&#x27;t agree more. The whole thing made me consistently think &quot;Oh cool, so what?&quot;. Usually when I read through a grad &#x2F; undergrad textbook I gain a new way of thinking about problems I want to think about. With GA, it was more of contorting my problems to fit onto GA, and then ... nothing.<p>However, there were some claims that I did not verify in the GA textbook about being computationally faster and more stable. <i>Those</i>, if borne out, would be valuable.
评论 #39584273 未加载