s******h
发帖数: 539 | 1 Good question. I agree with you that least favorable prior is to compare
priors on 'Proper Priors', i.e. the priors are probability distributions.
It's not we don't want to compare all 'priors including improper ones',
while I think it's very hard for us to compare them, for if we consider all
'priors+improper priors', 'least favorable' may not be that useful or
meaningful.
Consider
parameter space: \Theta={0,1}
In this case the risk set is a convex set on 2-dim space (Euclidean), say S.
For any |
|
g**r
发帖数: 425 | 2 Don't quite understand your question: if you already know, or at least you
are willing to assume your random variable has a beta distribution, then you
can use your conjugate beta prior.
For those you cannot find a conjugate prior, usually it is because your
distribution function is not in a "nice" format, and hence your prior*
likelihood function does not have the same kernal as your prior. So the key
is if you find a prior that has the same kernal as prior*likelihood, you
got your conjugate ... 阅读全帖 |
|
a***n
发帖数: 40 | 3 【 以下文字转载自 Statistics 讨论区 】
发信人: alvin (al), 信区: Statistics
标 题: a question about least favorable prior
发信站: BBS 未名空间站 (Tue May 29 23:34:29 2007)
Can a least favorable prior be an improper prior? My impression is that it
can, but it seems not true from the proof of the following theorem.
The theorem is: if Delta0 is a Bayes rule w.r.t. proper prior Pi0, and R(
theta, Delta0) <= r(Pi0, Delta0) for any theta. Then Pi0 is least favorable.
One of the steps of the proof is: integral of "R(delta0, the |
|
a***n
发帖数: 40 | 4 Can a least favorable prior be an improper prior? My impression is that it
can, but it seems not true from the proof of the following theorem.
The theorem is: if Delta0 is a Bayes rule w.r.t. proper prior Pi0, and R(
theta, Delta0) <= r(Pi0, Delta0) for any theta. Then Pi0 is least favorable.
One of the steps of the proof is: integral of "R(delta0, theta)*Pi(theta)"
over theta for any prior Pi, <= sup R(delta0, theta), sup is over theta.
That is, this step takes out sup R(), and it must be true |
|
a******e
发帖数: 119 | 5 Sorry，可能是我没说清楚，uninformative prior我是指diffuse prior/flat prior.
不是指conjugate prior |
|
v**********m
发帖数: 5516 | 6 U.S. Economy Expanded at 1.9% Pace in First Quarter, Above Prior Estimate
By Timothy R. Homan and Alex Kowalski - Jun 24, 2011 8:30 AM ET
U.S. Economy Grew 1.9% in First Quarter
Shoppers look at refrigerators at the ABT store during a promotion offering
a 15 percent rebate on certain Energy Star appliances in Glenview, Illinois,
U.S. Photographer: Tim Boyle/Bloomberg
Roger Altman Says Fed Growth Forecast Signals `Weakness'
General Motors employees assemble a GM crossover SUV on the assembly line... 阅读全帖 |
|
g**8
发帖数: 4951 | 7 我现在发现，舆论对民众的价值观和意见导向的作用是多么巨大。一个苹果告三星专利
的官司，媒体和大家围绕讨论的，简单了概括就是他们各自每个屏幕对角线的长度这样
的内容，然后大家都来阵阵有词的说苹果欺负人，再专业点的人也就是说什么“prior
art"前例。
我们就来说说这个prior art前例。以触摸屏的使用为例。注意苹果这里注册专利的不
是简单液晶显示触摸屏，偷换概念的讨论没有意义。苹果注册的应该是键盘由触摸屏上
显示产生，而非需要手机上再配一个物理键盘。
根据最近三星苹果专利官司的部分新闻，我注意到三星给出的证据之一就是强调苹果的
类似创意来自索尼，证据是他们（苹果）当时面试过一个索尼的员工。我不知道三星哪
里得到的如此具体的信息，其实侧面也证明了三星在类似事情（获取甚至盗窃其它公司
商业秘密方面）是非常有投入的，至少是很有SENSE的，所以都知道。其实那个消息至
少有一个简单逻辑：如果索尼没有提出对苹果的诉讼和证据说索尼员工在面试过程中泄
露了索尼的商业机密，那这个证据毫无法律效应。同样的，我相信苹果有更强的怀疑，
三星在给苹果产品提供零配件的过程中，窃取了更大量的苹果商业机密，涉... 阅读全帖 |
|
f***a
发帖数: 201 | 8 我接到一个很满意的offer。今天就要接受了。
在目前的学校工作了2年了，学校给办了H1B，I-140也在去年冬天办好了，今年年初
USCIS说已经接到申请材料了，这个就是已经在开始排期了，对吧？
新学校说给办H1B和绿卡，但是绿卡i485的部分要自己交钱，这个对么？
我今年夏天在现在的学校辞职，到新学校的话，我不太清楚现在的学校给办的I-140是
继续保持排期么？还是说他们的申请会取消，然后新学校的申请再弄新排期？
这换工作之间H1B的status和绿卡申请的交接问题，有什么需要我make sure的，还是说
学校同意给办那就take care of it了，我不用管了，没有任何过渡时期的问题？
还有就是，因为我在现在的学校已经2年了，到了新学校要不要申请prior service
credit呢？Dean说可以申请两年的，或者一年的。但是一般情况下大家都不申请，还有
人以前申请了，然后后悔了又撤销的，因为害怕在新学校3年或者4年之内达不到tenure
的要求。我现在有点犹豫要不要申请一年的credit。新学校比现在的学校对research的
要求稍微高一些，但是我现在的学校要求太低... 阅读全帖 |
|
n*********e
发帖数: 318 | 9 加入一个新公司之前， 被要求填写-
Prior invention disclosure
我自己曾写过一个Open Source 的package （网上可以免费下载） 属于GNU General
Public License。 只不过 implement 别人发表的 algorithm
请问这个算不算 Prior Invention？
谢谢 |
|
c**l
发帖数: 9003 | 10 Local Table Tennis Community,
Triangle Table Tennis, the new, dedicated Table Tennis Center conveniently
located in the middle of the Research Triangle, is getting ready to open
soon!
We are working hard to prepare for our Center’s Grand Opening on the last
weekend in May, to coincide with the arrival of our Head Coach, Fei-Ming
Tong.
In the meantime, we are pleased to offer several free Open-Play
opportunities prior to our Grand Opening. Although we’ll still be
implementing some of our plan... 阅读全帖 |
|
A*F
发帖数: 2272 | 11 I would say Apr 5 2014.
because his first patent did not become 102(a)(1) prior art until Apr 5,
2013
and he can use 102(b)(2)(C) to disqualify his first patent as 102(a)(2)
prior art
the |
|
R*******c
发帖数: 249 | 12 we know that normal distribution is usually chosen as the prior distribution
,
are there any other common prior distributions so that we can have a closed
form for the posterior distribution? |
|
s*******k
发帖数: 11 | 13 在网上查了很久 貌似一直没有对这个问题的专门研究
不知道版上有没有高手对了解一些的 求指教！
另外 从别的坛子上看来的 有人说 conjugate prior 是基于 sufficient statistics
来的
有时候可能出现 conjugate prior 不存在的情形
这时候只要抓住 sufficient statistics 及其 expectation 即可 但这招不是什么时
候都灵验 我现在需要对beta分布的random variable 进行 Bayesian parameter
estimation 就遇到困难了 |
|
g*******i
发帖数: 258 | 14 right,
some people do argue that inverse Wishart as a prior is not an ideal choice.
maybe you can try things like Jeffrey's prior, which I am not very sure |
|
i**z
发帖数: 194 | 15 这个用 flat prior 有时候挺危险的，最好的是用 cmp (conditional mean prior) 但
你得有历史数据，或者 expert idea。 或者自己的估算。 |
|
w*******y
发帖数: 60932 | 16 I went to Lowe's today and found out about a Class Action suit regarding
drywall from Lowe's. Please DO NOT ABUSE THIS. Only make a claim if you
indeed purchased the defective drywall from Lowe's.
Highlights:
- Drywall purchased or installed from Lowe's in your home prior to 7/27/10
is eligble
- $50 in Lowe's GC available for no proof of purchase or damages
- $250 in Lowe's GC available if you have proof of purchase prior to 7/27/10
- $2000 in Lowe's GC PLUS $2,500 in cash if you can independ... 阅读全帖 |
|
A****7
发帖数: 902 | 17 别人看到的世界都是不是颠倒的不一定，你看篇文章能只看到自己想看到的
Big online promotions and pre-holiday sales have lured in shoppers, some of
whom are becoming leery about the big crowds and in-store pandemonium that
often punctuates Black Friday. Anecdotal evidence suggests that consumers
shifted their purchases from physical stores to websites.
"This year, we saw Black Friday ads emerge before Halloween, as retailers
aimed to get at the shopper's wallet early," said Kevin Kearns, ShopperTrak
chief revenue officer. "And from ... 阅读全帖 |
|
d*****x
发帖数: 179 | 18 新的保险公司要proof of prior insurance，
直接寄原保险卡的复印件就行了吗？ |
|
p****n
发帖数: 109 | 19 一个project刚开始一个礼拜，根据NSF的guideline, 也要算到Results from prior
NSF support，但是还没有什么结果，该如何说？ |
|
发帖数: 1 | 20 跟不同人讨论得到的结果不一样呀
有人说要列出publications from prior NSF support, 显得没有浪费纳税人的钱；对
于合作的proposal，几个PI，如果都列出来好长呀
有人说不能超过半页，写多了没人看，不会影响评审；
有人说不写都行，但是听说有没写的直接被拒，没有送审；
大家都是怎么处理的？ |
|
l***d
发帖数: 1798 | 21 1. 如果有人prior nsf grant, 这个是一定要写的。
2. 一般1页之内
3. cite your publications , not list publications. |
|
G*****h
发帖数: 2272 | 22 Typically how many pages do you guys use for "results from prior NSF support
" for new NSF proposal? |
|
u**o
发帖数: 4652 | 23 Do not add any items to your cart on the merchant's site prior to obtaining
your OO reservation.
这个有关系么？ |
|
w*****g
发帖数: 186 | 24 以下这项是我在填州税（south Carolina）的时候碰到的，是什么东西啊？谢谢！
the carryover of unused non-refundable credits from prior years |
|
w*****g
发帖数: 186 | 25 查看了一下解释，还是不明白。
Enter the carryover of unused non-refundable credits from prior years. Refer
to the particular form applicable to your tax credit for the maximum credit
and period of time to carryover. Attach a break down of the credit by type
and year started. |
|
b*********h
发帖数: 240 | 26 2015因为股票收入hit AMT and 但今年没有， 可以填Prior year AMT credit for
2016 吗？好像很复杂。 有人用过吗?
谢谢 |
|
n*******e
发帖数: 2213 | 27 比Kershaw名头大啊。Prior当初被称为是年轻一代的Unit/Rocket。
可惜就废了。 |
|
c***d
发帖数: 756 | 28 写这么长没有一点是对的真是不容易啊。
1.palm早就有没有硬键盘的手机和pda了。
2.三爽知道苹果面试sony员工，是因为诉讼过程中双方都能检索对方文件。要不然你以
为苹果从哪找到的三爽的email, 还有iphone对三爽手机对比的ppt?
照你的逻辑，apple不黑进三爽的服务器是不行的吧，
这得sb成啥样才能得出这样的结论。
prior |
|
a***y
发帖数: 19743 | 29 写这么多有什么用？
还不如先把苹果的专利读清楚
prior
★ 发自iPhone App: ChineseWeb 7.3 |
|
k***i
发帖数: 662 | 30 September 3, 2007
Volume 85, Number 36
p. 9
POLYMERS
GE Plastics Adjusts Partnerships Prior to Sale to SABIC
Alex Tullo
GE Plastics is buying out partners in two joint ventures in advance of its sale to Saudi Basic Industries (SABIC), which is expected to be completed later this quarter.
GE is purchasing Bayer MaterialScience's 50% stake in Exatec, a joint venture formed in 1998 to promote the use of polycarbonate in automotive windows.
Exatec developed a plasma-enhanced vapor-deposition |
|
i*******p
发帖数: 297 | 31 sry i wasn't very clear, 102b printed publication type of prior art.
it. |
|
k*********r
发帖数: 362 | 32 XYZ filed first nonprovisional on Apr 1 2013, issued on Apr 5 2013.
XYZ wants to file second nonprovisional but does not want the first
nonprovisional to become prior art. What is the latest date XYZ can file the
second nonprovisional? Apr 1 2014 or Apr 5 2014?
some new posting shows this question and the answer does not look right |
|
k*********r
发帖数: 362 | 33 XYZ filed original nonprovisional application on March 20, 2014, a french
patent application was filed on Jun 18, 2013, is the french patent qualified
as prior art? |
|
u****g
发帖数: 90 | 34 hi DXJM.
in the myERAS current/prior training, can we fill up 实习 in here? Thanks! |
|
b*w
发帖数: 349 | 35 报名中遇到的问题：
I was required to complete an internship prior to receiving my medical
diploma
这个是不是应该选no, 因为我们那实习算是clerkship.
最近报过名的能给指点一下吗？
多谢了先 |
|
m*********g
发帖数: 646 | 36 MLE? Really ? I thought the op was asking prior probability in Bayesian. |
|
o****o
发帖数: 8077 | 37 noninformative prior may result in closed form posterior under some models
distribution
closed |
|
s**u
发帖数: 383 | 38 Negative Binomial distribution use Beta for "p", the probability. But for
the integer "r", what kind of prior should be used? Thanks, |
|
p*******g
发帖数: 53 | 39 i think he is looking for a conjugate prior for beta likehood |
|
s*******6
发帖数: 686 | 40 Yes,
I need to find conjugate prior for beta likelihood.
I googled a lot and can not find a good answer.
thx |
|
d******e
发帖数: 7844 | 41 容易计算这一点就足够了。
Parametric的Bayesian Model可以直接在Sampling的每一步得到closed-form solution
。不用Conjugate Prior，算积分算死。最终的直接结果就是Parametric的Bayesian
Model能很轻松的处理大规模数据，estimate几万几十万个甚至更多parameter都不在话
下。
我还没见过有一般的MCMC玩超过100个dimension以上的问题。 |
|
h******3
发帖数: 190 | 42 Bayesian hierarchical model里通常用inverse wishart作为covariance matrix的
conjugate prior. 我的问题是怎么选择inverse wishart的scale matrix. 通常就用
identity matrix吗? 对estimates有什么影响. 多谢. |
|
g*******i
发帖数: 258 | 43 Usually it should be (k+m)I
where k is the degree of freedom, m <= 3, and I is the identity
this gives you a diffuse prior. no other meaning |
|
h******3
发帖数: 190 | 44 If I want the prior to be as non-informative as possible, is there any
alternative?
thanks. |
|
h******3
发帖数: 190 | 45 er... wrapping my head around it...
Didn't the coefficient k makes influence of prior larger? since k > 1 |
|
a******e
发帖数: 119 | 46 RT,
怎么找Gamma distribution的 uninformative priors 啊？ |
|
r********n
发帖数: 6979 | 47 google gamma distribution conjugate prior |
|
P*****r
发帖数: 554 | 48 maximize K-L distance 的prior？
. |
|
r********n
发帖数: 6979 | 49 如果你不要求conjugate
那随便个啥prior都可以了
. |
|
