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。

A puzzle of two unreliable sensors

18 点作者 tibbar24 天前

4 条评论

sweezyjeezy21 天前
Let U~Uniform(0,1) Let sensor target measurement be x, so A ~ (x + U), B ~ x or U with probability 0.5. We draw a from A and b from B, we want the estimator to minimise mean absolute error - the bayes optimal rule is the posterior median of x over the likelihood function of L(x | a, b).<p>Note that if a = 0 and b = 1 -&gt; we KNOW b!=x because a is too small - there is no u with (u + 1) &#x2F; 2 = 0. I&#x27;ll skip the full calculation here, but basically if b could feasibly be correct its &quot;atomic weight&quot; ends up being as least as large as 0.5, so it is the posterior median, otherwise we know b is just noise, and the median is just a. So our estimator is<p>b if a in range [b&#x2F;2, (b+1)&#x2F;2]; a otherwise<p>This appears to do better than OPs solution running an experiment of 1M trials (MAE ~ 0.104 vs 0.116, I verify OPs numbers). The estimator to minimise the mean squared error (the maximum likelihood estimator) is more interesting - on the range a in [b&#x2F;2, (b+1)&#x2F;2] it becomes a nonlinear function of a of the form 1 &#x2F; (1 + piecewise_linear(a)).
评论 #43736659 未加载
defanor21 天前
Looks like A and B are swapped in the &quot;errorB vs. errorA&quot; plot&#x27;s description: in other places (and on the plot itself, actually) A is the noisy sensor, B is the occasionally-faulty one.
reedf124 天前
<a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Kalman_filter" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Kalman_filter</a>
评论 #43703600 未加载
echoangle24 天前
Is this just a thought exercise? Are there any sensors that actually work like the model described here?
评论 #43709550 未加载
评论 #43710908 未加载